Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1–19 are rejected under 35 U.S.C. 101 because the claims are directed to a judicial exception (abstract idea) without integrating the exception into a practical application, and the additional elements do not amount to significantly more than the exception.
This analysis follows the USPTO’s eligibility framework in MPEP 2106 (Step 1; Step 2A Prongs One/Two; Step 2B).
Step 1 — Statutory category
Claim 1 recites a “mobility sharing system … comprising” computing components (memory, processor, functional units). Thus, the claim is to a machine/system, a statutory category. (Step 1: YES).
Step 2A, Prong One — Claim 1 recites an abstract idea (judicial exception)
A. Mathematical concepts
Claim 1 expressly recites multiple mathematical / algorithmic techniques, including:
dividing an area into unit areas “using a clustering model based on unsupervised learning”;
predicting departures/arrivals by inputting data into a “bidirectional recurrent neural network (RNN) model”;
generating an input feature set via “embedding a result value of subtracting” arrivals and departures, and bit-width constraints (information representation/compression constraints);
determining a mobility location route by applying a “minimum cost maximum flow (MCMF) algorithm,” including defined costs/capacities and proportional/inverse proportional relationships; and
training the bidirectional RNN model by inputting labeled training data.
These limitations fall within the “mathematical concepts” grouping of abstract ideas (e.g., mathematical relationships, formulas, calculations, and algorithmic processing), as reflected in USPTO eligibility guidance and AI eligibility examples discussing neural-network-based processing and embeddings as mathematical concepts when claimed at this level of generality.
B. Certain methods of organizing human activity (resource allocation / logistics)
Claim 1 is also directed to managing a one-way vehicle sharing service by:
using historical reservation data to determine how many vehicles should be located across stations/areas,
forecasting station-level demand (departures/arrivals),
prioritizing distribution based on predicted deltas and conditions, and
computing relocation routes to rebalance supply.
This is a form of commercial/logistics resource allocation and scheduling—a classic “organizing human activity” type abstraction, implemented using generic computing.
Therefore, claim 1 recites a judicial exception (abstract idea) at least in the form of mathematical concepts and organizing human activity. (Step 2A, Prong One: YES).
Step 2A, Prong Two — Claim 1 does not integrate the abstract idea into a practical application
After identifying the abstract idea, the additional elements are evaluated individually and in combination to determine whether the claim as a whole integrates the exception into a practical application (i.e., whether the claim applies the exception in a manner that imposes a meaningful limit such that it is more than drafting to monopolize the exception).
Additional elements beyond the abstract idea
Beyond the abstract idea(s), claim 1 recites:
a “mobility sharing system” environment with stations/vehicles,
a memory storing instructions and historical data, and
a processor executing the instructions (including model training).
Why these do not integrate into a practical application
These additional elements do not integrate the abstract idea into a practical application because:
Generic computer implementation. The recited memory/processor perform routine functions (store, execute, train, predict, compute). The claim does not recite a specific computing architecture, specialized hardware, or a specific improvement to computer functioning (e.g., improved training technique tied to a specific technical effect in the computer itself).
Field-of-use limitation. Limiting the computations to a “mobility sharing” context (stations/vehicles/one-way reservations) is merely a technological environment/field-of-use that does not meaningfully limit the abstract idea; it is applying generic ML/optimization to a particular business/logistics domain.
No particular machine / transformation. The claim does not recite controlling a particular machine in a technological sense (e.g., vehicle control, sensor-actuation loops, or a transformation of matter). It computes predictions and routes and “sets” numbers of vehicles—i.e., information processing and planning outputs.
No technical improvement recited as claimed. Even though the claim uses advanced techniques (BiRNN, embeddings, MCMF), the claim does not recite how these produce a technological improvement (e.g., improved network security, improved signal processing, improved sensor fidelity, reduced computational complexity in a non-generic way). At this level, the claim is directed to using established ML/optimization techniques as tools to perform prediction and scheduling in a service operation. This aligns with Federal Circuit treatment of claims that apply generic machine learning in a particular environment without a technological improvement.
Accordingly, claim 1 as a whole does not integrate the abstract idea into a practical application. (Step 2A: YES, the claim is directed to the abstract idea).
Step 2B — Claim 1 lacks an inventive concept (“significantly more”)
Because claim 1 is directed to an abstract idea, the claim is evaluated for an inventive concept—i.e., whether additional elements, individually or as an ordered combination, amount to significantly more than the abstract idea itself.
Claim 1 does not provide an inventive concept because:
The additional elements amount to generic computing components (processor, memory) executing instructions to perform the abstract idea. The claim recites conventional steps of collecting/using historical data, training a model with labeled data, performing predictions, and running an optimization algorithm to produce an allocation/route plan.
The recited ML/optimization techniques are claimed at a result-oriented/functional level (use clustering; input into bidirectional RNN; embed a subtraction result; apply MCMF), without claiming a specific technological implementation that is unconventional in computer operation. That is, the claim uses the computer as a tool to implement the abstraction.
Recentive Analytics v. Fox (Apr. 18, 2025) is directly instructive: the court affirmed ineligibility where claims were “directed to the abstract idea of using a generic machine learning technique in a particular environment, with no inventive concept,” holding that applying established ML methods to a new data environment is not enough. The present claim similarly applies generic ML/optimization techniques to the mobility-sharing allocation/routing environment without a recited technological improvement.
Therefore, claim 1 does not recite additional elements that amount to significantly more than the abstract idea. (Step 2B: NO inventive concept).
Dependent claims 2–19
Claims 2–19 depend (directly or indirectly) from claim 1 and thus include the same abstract idea(s) and the same generic computer implementation. The additional limitations in claims 2–19 merely further specify the abstract mathematical/logistical rules (or add conventional ML model variants), and therefore do not integrate the exception into a practical application and do not add an inventive concept.
Claim 2: restates use of an unsupervised clustering model → mathematical concept.
Claim 3: sets areas based on probability of same-area departure/arrival → mathematical/statistical rule.
Claim 4: minimizes sum of station costs proportional to squared distances → explicit mathematical optimization.
Claim 5: minimizes absolute area differences → mathematical optimization.
Claim 6: nearest-station grouping constraint → mathematical rule/heuristic.
Claims 7–9: Bi-GRU / Bi-LSTM / stacked RNN layers → generic ML model selection (mathematical concept), not a technical improvement as claimed.
Claims 10–12: adds periodic time functions proportional to sin/cos with T=24 or T=7 → explicit mathematical relationships.
Claims 13–15: alternative subtraction embedding and representations → mathematical data transformation/encoding.
Claims 16–17: prioritization rules based on predicted deltas and tie-breakers → abstract decision rules for resource allocation.
Claims 18–19: further defines the MCMF graph/cost/capacity setup → mathematical optimization formulation.
Thus, claims 2–19 fall with claim 1 for the same reasons under Step 2A/2B.
Examiner 101 Conclusion
For the reasons set forth above, claims 1–19 are rejected under 35 U.S.C. 101 as being directed to an abstract idea (mathematical concepts and organizing human activity) without integration into a practical application and without significantly more.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Murakami et al. (US 6701300 B1), herein after will be referred to as Murakami, in view of Karamanis et al., herein after will be referred to as Karamanis, in view of Ara et al., herein after will be referred to as Ara, and in view of Crawford et al. (US 20150293.232 A1), herein after will be referred to as Crawford.
Regarding Claim 1
Disclosure by Murakami
Murakami discloses:
A mobility sharing system
See at least: "The present invention relates generally to a vehicle allo cation System; more particularly, to a vehicle allocation System capable of Supplying (or allocating) vehicles with Stability within a desired wait time Set in the System according to a ride demand in each port inside a certain area including a plurality of ports." (Col. 1, ll. 6-11)
Rationale: Murakami expressly discloses a system for allocating shared vehicles based on user ride demands between ports, which is a mobility sharing system.
for providing a one-way vehicle sharing service
See at least: "The ride demand includes the information about destination." (Col. 4, ll. 35-36); "The terminal 2 of port P in which the vehicle carrying the user arrives sends the... arrival recognition information... to the host 1." (Col. 4, ll. 43-48)
Rationale: The system manages trips where a user inputs a destination and the vehicle's arrival at a different port is recognized, constituting a one-way vehicle sharing service from a departure port to an arrival port.
comprising: a first vehicle determination unit
See at least: "The host 1 has a computing means (CPU) 10 for performing the computation" (Col. 4, ll. 52-53); "vehicle allocation setting section 108" (Fig. 2)
Rationale: Murakami's computing means (CPU) 10 and vehicle allocation setting section 108 perform the function of determining vehicle allocation. To a PHOSITA, these components constitute a first vehicle determination unit.
configured to divide an entire area
See at least: "a certain area including a plurality of ports" (Col. 1, ll. 11-12); "Ports P1, P2, P3, P4, and P5... are predetermined parking spaces arranged in i.e., a golf course, an airport, or a hotel." (Col. 3, ll. 57-61)
Rationale: Murakami's system is designed to operate within an entire area that is already divided into discrete ports. To a PHOSITA, the system is inherently configured to work with this pre-divided area, which is a necessary precursor to its allocation function.
into a plurality of unit areas
See at least: "a certain area including a plurality of ports" (Col. 1, ll. 11-12); "Ports P1, P2, P3, P4, and P5" (Col. 3, ll. 57-58)
Rationale: The area contains a plurality of unit areas, explicitly named as ports P1, P2, etc.
and to determine the number of vehicles
See at least: "determining the number of deployed vehicles on the basis of the number of forecast occurrence trips per average travel time between ports" (Col. 13, ll. 12-14)
Rationale: The system performs the function of determining the number of vehicles to deploy.
to be located in each of the plurality of unit areas
See at least: "deployment of 18 vehicles for each of 5 ports" (Col. 12, ll. 41-42)
Rationale: The determination results in a number of vehicles to be located in each of the plurality of unit areas (e.g., 18 vehicles per port).
based on historical data
See at least: "The memory 11 stores forecast ride demand data... as one of the ride demand statistical data of all ports." (Col. 4, ll. 57-60); "The forecast occurrence trip is an accumulation of the past ride demand results" (Col. 5, ll. 41-42)
Rationale: The determination is made based on historical data ("past ride demand results", "statistical data").
extracted from a one-way reservation history
See at least: "The ride demand includes the information about destination." (Col. 4, ll. 35-36); "contract user travel data" (Col. 4, ll. 47)
Rationale: The historical data is derived from trip records where users specify a destination, forming a history of one-way trips. To a PHOSITA, this trip history is a reservation history for the vehicle sharing service.
with different departure and arrival stations
See at least: "according to ride demands occurring in a plurality of ports" (Col. 1, ll. 14-15)
Rationale: The trip history is for journeys between different departure and arrival stations (ports).
a prediction unit
See at least: "forecast occurrence demand count computing section 102... forecast occurrence vehicle count computing section 104" (Fig. 2)
Rationale: These computing sections perform predictive calculations and collectively form a prediction unit.
configured to
Rationale: The functional description of the computing sections (102, 104) inherently means they are configured to perform their stated operations. This is an inherent characteristic of a system component to a PHOSITA.
by inputting a prediction data set
See at least: "The forecast occurrence trip is read by a forecast occurrence trip count detector 101" (Col. 5, ll. 46-47)
Rationale: The prediction unit operates by inputting a prediction data set (the stored "forecast occurrence trip" data).
generated based on the number of vehicles departing from each of stations included in a first unit area of the plurality of unit areas for each of a plurality of unit time intervals
See at least: "The forecast occurrence trip storage section 110 stores a ride demand result of one day in the form of time-series data as a forecast occurrence trip (forecast ride demand) for each port P." (Col. 5, ll. 38-41)
Rationale: The prediction data set (forecast occurrence trip) is generated based on historical ride demand results. A ride demand corresponds to a vehicle departure. This data is stored as time-series data, which inherently comprises data for a plurality of unit time intervals, and is stored for each port P (stations included in a... unit area).
and the number of vehicles arriving at each of the stations
See at least: "The forecast arrival trip count detector 103 detects a forecast arrival trip count" (Col. 5, ll. 53-54)
Rationale: The prediction data set also includes data related to the number of vehicles arriving at each of the stations ("forecast arrival trip count").
into a model predict the number of vehicles departing from each of the stations and the number of vehicles arriving at each of the stations
For predicting departures (forecast occurrence demand count):
See at least: "The forecast occurrence trip is read by a forecast occurrence trip count detector 101 to be supplied to a forecast occurrence demand count computing section 102." (Col. 5, ll. 46-49 ); "The forecast occurrence demand count computing Section 102 and the forecast occurrence vehicle count computing Section 104 compute the number of demands and the number of vehicles in this SD time" (Col 7, ll. 3-6)
Rationale: This describes inputting data (the forecast occurrence trip) into a computing section (model) to predict the number of vehicles departing (forecast occurrence demand count).
For predicting arrivals (forecast occurrence vehicle count):
"The forecast arrival trip count detector 103 detects a forecast arrival trip count on the basis of the above-mentioned forecast occurrence trip and inputs the detection into a forecast occurrence vehicle count computing section 104." (Col. 5, ll. 53-57); "The forecast occurrence vehicle count computing section 104…compute a forecast occurrence vehicle count." (Col. 5, ll. 57-60)
Rationale: This describes inputting data (forecast arrival trip count) into a computing section (model) to predict the number of vehicles arriving, which is part of the forecast occurrence vehicle count.
The predictive model and its formula:
"An evaluation value computing section 105 computes an evaluation value for determining the urgency of the number of vehicles in each port P by use of relation (f3) on the basis of a forecast occurrence demand count and a forecast occurrence vehicle count: Evaluation value=(forecast occurrence vehicle count- forecast occurrence demand)+forecast occurrence demand count^1/2 (f3)" (Col. 6, ll. 18-27)
Rationale: This explicitly discloses a computational model (relation f3) that processes the predicted counts (demand and vehicle counts) to output an evaluation value. This model is a core part of the prediction and relocation logic.
Rationale: Murakami explicitly discloses a system where historical and forecast trip data is input into specific computing sections (102, 104, 105). These sections implement algorithms or models (including the evaluation value formula, f3) that process this input data to generate predicted outputs: the forecast occurrence demand count (predicted departures) and the forecast occurrence vehicle count (which includes predicted arrivals). Therefore, Murakami teaches inputting data into a model to predict the number of vehicles departing from and arriving at each station. The limitation not disclosed by Murakami is the specific type of model claimed: a bidirectional recurrent neural network (RNN) model. The general concept of using a predictive model is present.
during a target time interval
See at least: "within a scheduled time period" (Col. 2, ll. 14); "SD time" (Col 7, ll. 6)
Rationale: The predictions are made for a specific future period, the target time interval, referred to as the "scheduled time period" or "SD time".
after the plurality of unit time intervals
Rationale: Murakami's system predicts for a future SD time based on historical time-series data. To a PHOSITA, it is inherent that the predicted target time interval occurs after the plurality of unit time intervals used to generate the input data.
wherein the prediction unit generates the prediction data set
See at least: "The forecast occurrence trip is read by a forecast occurrence trip count detector 101" (Col. 5, ll. 46-47)
Rationale: The prediction unit generates the prediction data set by reading and processing the stored forecast occurrence trip data.
by processing at least one of history data or external data
See at least: "a value obtained by multiplying a forecast arrival trip count by the above-mentioned ratio (referred to as a reduction coefficient) is used as a forecast arrival trip count" (Col. 6, ll. 5-6)
Rationale: The prediction data set (the adjusted forecast arrival trip count) is generated by processing... history data (currently owned vehicle count, forecast demand count) using a reduction coefficient.
wherein the prediction unit generates the prediction data set by subtracting the number of vehicles departing from each of the stations from the number of vehicles arriving at each of the stations,
See at least: "Evaluation value=(forecast occurrence vehicle count- forecast occurrence demand)+..." (Col. 6, Eqn. (f3); "the number of forecast occurrence demands is subtracted from the number of forecast occurrence vehicles of that port P" (Col. 9, ll. 55-57, Fig. 5, Step S7)
Rationale: The system calculates a value by subtracting the number of vehicles departing from each of the stations (forecast occurrence demand) from the number of vehicles arriving at each of the stations (forecast occurrence vehicle count).
a second vehicle determination unit
See at least: "vehicle relocation setting section 107" (Fig. 2)
Rationale: This section performs vehicle relocation decisions and acts as a second vehicle determination unit.
configured to
Rationale: The functional description of the vehicle relocation setting section 107 inherently means it is configured to perform its stated operations. This is an inherent characteristic to a PHOSITA.
based on the predicted number of departing vehicles,
See at least: "vehicle relocation setting section 107, on the basis of the vehicle excess or shortage count of each port P" (Col. 6, ll. 40-42).
Rationale: The relocation setting is performed based on the predicted number of departing vehicles (which contributes to the excess/shortage count).
the predicted number of arriving vehicles,
See at least: "vehicle relocation setting section 107, on the basis of the vehicle excess or shortage count of each port P" (Col. 6, ll. 40-42); “An excess/shortage count computing section computes excess or shortage from the difference between the number of demands and the number of vehicles” (Abstract).
Rationale: The relocation setting is performed based on... the predicted number of arriving vehicles (which contributes to the excess/shortage count).
and a set condition,
See at least: "the vehicles to be relocated are sequentially set for ports P in increasing order of the calculated evaluation values." (Col. 8, ll. 34-35)
Rationale: The relocation logic follows a set condition, such as operating in a defined order (increasing evaluation value).
set the number of location vehicles to be located at each of the stations
See at least: "gives an instruction of relocation for moving the excess vehicle 4 to another port P." (Col. 6, ll. 43-44)
Rationale: By instructing relocation from ports with excess vehicles to ports with shortages, the unit effectively sets the number of location vehicles to be located at each of the stations.
during the target time interval,
See at least: "relocation can be made between all ports within an SD time of 30 minutes or longer." (Col. 10, ll. 52-53)
Rationale: The relocation is planned for execution within the target time interval (the SD time).
wherein the second vehicle determination unit prioritizes distributing a vehicle to a station
See at least: "relocation is executed from the ports P in the order of more serious vehicle shortage" (Col. 10, ll. 22-24)
Rationale: The system prioritizes distributing a vehicle to a station based on the severity of vehicle shortage.
having a larger prediction result value
See at least: "Evaluation value=(forecast occurrence vehicle count- forecast occurrence demand)+..." (Col. 6, Eqn. (f3)). Ports are arranged in increasing order of evaluation values for service (Fig. 5, Step S4).
Rationale: The system prioritizes ports based on the evaluation value. A less negative (i.e., larger) evaluation value indicates a less severe shortage. The evaluation value incorporates the prediction result value of (vehicle count - demand). Therefore, the system prioritizes stations having a larger prediction result value (a less negative difference between arrivals and departures).
of subtracting the predicted number of departing vehicles from the predicted number of arriving vehicles
See at least: "Evaluation value=(forecast occurrence vehicle count- forecast occurrence demand)+..." (Col. 6, Eqn. (f3))
Rationale: The evaluation value, which governs prioritization, includes a term calculated by subtracting the predicted number of departing vehicles (forecast occurrence demand) from the predicted number of arriving vehicles (forecast occurrence vehicle count).
among stations satisfying the set condition
See at least: "the ports P are checked for vehicle excess or shortage in the increasing order of evaluation values." (Fig. 5, Step S5)
Rationale: The prioritization occurs among stations satisfying the set condition of having a vehicle shortage (which are the ones checked in Step S5).
a route setting unit
See at least: "vehicle relocation setting section 107" (Fig. 2)
Rationale: The vehicle relocation setting section 107 performs the function of determining relocation routes, constituting a route setting unit to a PHOSITA.
configured to determine a mobility location route
See at least: "gives an instruction of relocation for moving the excess vehicle 4 to another port P." (Col. 6, ll. 43-44)
Rationale: The system logic is configured to determine a mobility location route by issuing a relocation instruction from one port to another.
to locate the vehicles at each of the stations
See at least: "for moving the excess vehicle 4 to another port P." (Col. 6, ll. 43-44)
Rationale: The purpose of the route is to locate the vehicles at each of the stations, specifically to move them to stations (ports) with a shortage.
by applying an algorithm
See at least: "the vehicle relocatable ports P are checked in the increasing order of distances from the port P running short of vehicles." (Col. 9, ll. 62-64, Fig. 5, Step S9); "relocation setting processing" (Fig. 5, Steps S1-S13)
Rationale: The process for deciding which vehicle to relocate from which port involves a step-by-step procedure—checking ports in order of distance and following the logic outlined in the flowchart. To a PHOSITA, this procedural logic constitutes an algorithm.
to the number of location vehicles to be located at each of the stations,
See at least: "on the basis of the vehicle excess or shortage count of each port P" (Col. 6, ll. 41-43)
Rationale: The relocation algorithm is applied to the computed vehicle excess or shortage count for each port. This count directly defines the number of location vehicles to be located at each of the stations (i.e., a negative shortage indicates how many vehicles need to be located there, and a positive excess indicates how many are available to relocate).
and a cost
See at least: "the vehicle relocatable ports P are checked in the increasing order of distances from the port P running short of vehicles." (Col. 9, ll. 62-64), Step S9)
Rationale: The system uses distance as a criterion for selecting relocation sources, which is a form of cost minimization.
to be proportional to distances between stations corresponding to the first nodes and stations corresponding to the second nodes,
See at least: "checked in the increasing order of distances" (Page 18, col. 2, Step S9)
Rationale: The selection heuristic implies that the cost of relocating from one station to another is proportional to distances between stations.
a memory
See at least: "a storage device (memory) 11" (Col. 4, ll. 55-56)
Rationale: The system includes a memory.
configured to store instructions to operate the prediction unit
See at least: "The host 1 has a computing means (CPU) 10 for performing the computation" (Col. 4, ll. 55-56).
Rationale: It is inherent to a PHOSITA that the memory of a computing system stores program instructions that, when executed by the processor, operate the prediction unit (computing sections 102, 104, etc.).
and the historical data including at least one of a service reservation time, a departure time, an arrival time, a departure station of a vehicle, an arrival station of a vehicle, a cost, and positions of the stations,
See at least: "The memory 11 stores... contract user travel data. The contract user travel data includes travel distance and travel time" (Col. 4, ll. 57-66)
Rationale: The memory stores historical data ("contract user travel data"). Travel time data inherently encompasses departure time and arrival time. The travel is between a departure station and an arrival station. Travel distance/time can be used to calculate a cost. The positions of the stations are fixed and known to the system. Therefore, the stored data includes the enumerated information.
and a processor
See at least: "The host 1 has a computing means (CPU) 10" (Col. 4, ll. 52)
Rationale: The system includes a processor (CPU 10).
configured to execute the instructions to operate the prediction unit
See at least: "The forecast occurrence trip is read by a forecast occurrence trip count detector 101 to be supplied to a forecast occurrence demand count computing section 102." (Col. 4, ll. 46-49)
Rationale: The processor executes the system's programmed logic, causing the prediction unit (detector 101, computing sections 102, 104, 105) to operate.
Claim Limitations Not Explicitly Disclosed by Murakami
Murakami does not explicitly disclose:
using a clustering model
based on unsupervised learning
into a bidirectional recurrent neural network (RNN) model
to reduce a size of the prediction data set
by embedding a result value
wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations
by applying a minimum cost maximum flow (MCMF) algorithm
sets a capacity between first nodes and second nodes to be infinite
and inversely proportional to at least one of the number of location vehicles of the stations corresponding to the first nodes in the target time interval and the number of location vehicles of the stations corresponding to the second nodes in the target time interval
wherein the processor trains the bidirectional RNN model
by inputting a dataset for training labeled with correct answer data
including the number of vehicles departing from each of the stations and the number of vehicles arriving at each of the stations during the target time interval
Disclosure by Karamanis
Karamanis discloses:
using a clustering model
See at least: "K-means clustering was used to split the road-network into twenty clusters" (Page 9, Section III. A)
Rationale: Karamanis explicitly discloses the use of a clustering model (K-means) to divide the operational area into discrete clusters, which are analogous to unit areas or stations.
based on unsupervised learning
See at least: "K-means clustering" (Page 9, Section III. A)
Rationale: K-means clustering is a canonical example of an unsupervised machine learning method. To a PHOSITA, its use constitutes applying unsupervised learning to partition the network.
into a model
See at least: "Agent-Based Model" (Page 9, Section III. A)
Rationale: The model used for demand prediction and redistribution is formulated as a non-linear minimum cost flow problem and solved using convex optimization and network flow algorithms, not a bidirectional RNN.
to reduce a size of the prediction data set
See at least: "we created typical demand profiles... K-means clustering was used to split the road-network into twenty clusters" (Page 9, Section III. A, Lines 11-13)
Rationale: Clustering aggregate trip data from a large number of individual points (e.g., NYC taxi data) into a smaller number of representative clusters inherently reduces the size and dimensionality of the prediction data set used by the model.
by embedding a result value,
See at least: "We model the derived vehicle redistribution problem as a non-linear minimum cost flow problem" (Page 2, Section I, Lines 23-24); The formulation uses functions like
N
i
j
t
(
x
)
(Equation 4) which embed the result of the customer choice model (ridership prediction) into the network flow cost structure.
Rationale: The core methodology embeds the predicted outcome of vehicle supply on customer demand and revenue (a result value) directly into the cost functions of the network edges.
by applying a minimum cost maximum flow (MCMF) algorithm,
See at least: "we formulate the vehicle redistribution problem as a non-linear minimum cost flow problem" (Abstract); "we introduce an edge splitting algorithm to solve a transformed convex minimum cost flow problem" (Abstract, Lines 7-8); "Algorithm 1 CMCF Edge-Splitting Algorithm" (Page 8, Section II. D)
Rationale: The entire solution methodology is built upon solving a Minimum Cost Flow (MCF) problem. While termed "Convex Minimum Cost Flow (CMCF)", the core is an MCF algorithm adapted for convex costs, fulfilling the claimed function of applying an MCMF algorithm. The "maximum flow" aspect is implicit in satisfying demand (flow) at minimum cost.
sets a capacity between first nodes and second nodes to be infinite
See at least: "
u
i
j
=
∞
∀
(
i
,
j
)
∈
E
" (Page 4, Equation 18)
Rationale: In the network flow graph
G
=
(
V
,
E
)
, the capacity
u
i
j
for edges is explicitly set to infinity (
∞
).
and inversely proportional to at least one of the number of location vehicles of the stations corresponding to the first nodes in the target time interval and the number of location vehicles of the stations corresponding to the second nodes in the target time interval;
See at least: "
w
i
j
t
(
x
)
=
β
i
t
I
i
j
t
(
x
)
" (Page 3, Equation 2) and its relaxed version "
w
i
j
t
(
x
)
=
α
i
t
Z
i
j
t
x
+
1
" (Page 6, Equation 22).
Rationale: The wait time
w
, a key driver of cost and demand in the model, is formulated to be inversely proportional to a function of vehicle supply
x
(which represents the number of location vehicles) in a cluster (station). This relationship is embedded in the cost functions of the flow network.
including the number of vehicles departing from each of the stations and the number of vehicles arriving at each of the stations during the target time interval
See at least: "we assume the fleet operator has estimates of the total demand
Z
i
j
t
from cluster
i
to cluster
j
" (Page 3, Section II. A, Lines 1-2).
Z
i
j
t
represents the total number of trip requests from
i
to
j
, which corresponds to desired departures from
i
and arrivals at
j
.
Rationale: The core demand input to the model includes data on trips between clusters, which encapsulates the number of vehicles departing from and arriving at each cluster (station) for future time intervals.
Motivation to Combine Murakami and Karamanis
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami and Karamanis before them, to integrate Karamanis's advanced network optimization model into Murakami's vehicle redistribution framework to improve economic efficiency and predictive accuracy. Murakami discloses a basic system for relocating shared vehicles based on forecasted surplus and shortage. Karamanis directly addresses the core limitations of such systems by teaching that optimal redistribution must account for customer choice behavior and the diminishing returns of adding vehicles. A skilled artisan would have been motivated to enhance Murakami's system by applying Karamanis's clustering model to create meaningful operational zones, and by replacing Murakami's heuristic logic with Karamanis's minimum cost flow algorithm to optimize redistribution for profit rather than just distance. Incorporating Karamanis's principle that wait time is inversely proportional to vehicle supply would further provide Murakami's system with the realistic economic model it lacks. The combination is straightforward because Karamanis's methodologies directly address the same technical problem and would seamlessly enhance the predictive and optimization core of Murakami's system.
Claim Limitations Not Mapped by the Combination of Murakami and Karamanis
After combining the teachings of Murakami and Karamanis, the following are not explicitly disclosed:
into a bidirectional recurrent neural network (RNN) model
wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations
wherein the processor trains the bidirectional RNN model
by inputting a dataset for training labeled with correct answer data
Disclosure by Ara
Ara discloses:
into a bidirectional recurrent neural network (RNN) model
See at least: "This paper proposes a deep learning method combining convolutional neural network and bidirectional long short term memory (biLSTM) for predicting travel demand of ride hailing services." (Page 1, Section I); "The proposed model integrates convolutional and recurrent neural networks..." (Abstract); "Our work...is combining bidirectional LSTM with CNN..." (Page 1, Section I); "CNN-biLSTM architecture" (Page 4, Section D, and Figure 5).
Rationale: Ara explicitly discloses the use of a bidirectional Long Short-Term Memory (biLSTM) network, which is a specific type of bidirectional Recurrent Neural Network (RNN), integrated with a CNN to form a predictive model.
wherein the processor trains the bidirectional RNN model
See at least: "The proposed model forecasts travel demand...by observing the demand over the past two weeks for that particular hour." (Abstract, Lines 7-9); "Our experiments with a real-world hire vehicle dataset...showed that the proposed model outperforms..." (Abstract); The entire "RESULTS AND DISCUSSION" section (Page 5) details the training and evaluation of the model, including parameter tuning ("parameters are tuned and result is observed").
Rationale: The described experimental process, where the model's architecture is defined, parameters are set and tuned, and its performance is evaluated on historical data, constitutes the standard process of training a machine learning model. To a PHOSITA, this explicitly discloses that the processor executes instructions to train the bidirectional RNN (biLSTM) model.
by inputting a dataset for training labeled with correct answer data
See at least: "We have chosen 2018 (January- December) for hire vehicles trip data." (Page 2, Section III); "Each row represents a single trip..." (Page 2, Section III); "The demand prediction problem aims to predict the demand at time interval t+k, where t is current timestamp and k is the lag size." (Page 3, Section III).
Rationale: The model is trained on historical trip data (the "dataset for training"). In a supervised prediction task, the historical data point for a given past time interval (e.g., demand at time *t*) serves as the "correct answer data" or label for the input context (e.g., data from the two weeks prior to *t*). This is the standard paradigm for time-series forecasting and is inherent to Ara's described methodology.
Motivation to Combine Murakami, Karamanis, Ara
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, and Ara before them, to incorporate Ara's advanced deep learning prediction model into the combined vehicle redistribution system of Murakami and Karamanis, in order to achieve superior forecast accuracy. A skilled artisan, having already been motivated to combine Murakami's redistribution framework with Karamanis's economic optimization model, would recognize that the predictive accuracy of the demand input is critical to the performance of the overall system. While Murakami uses basic time-series forecasting and Karamanis formulates the optimization problem, neither discloses the use of state-of-the-art neural networks specifically designed for spatiotemporal demand prediction in ride-hailing. Ara directly teaches that a bidirectional RNN model (biLSTM) integrated with a CNN provides the highest accuracy for forecasting origin-destination ride-hailing demand between neighborhood zones. Given that accurate prediction of departures and arrivals is the foundational input for both Murakami's relocation logic and Karamanis's minimum cost flow network, a skilled artisan would be led to substitute or augment the existing prediction modules in the combined system with Ara's superior CNN-biLSTM model. This substitution is obvious because Ara's model is specifically designed for the same type of data (historical trip records between zones) and the same technical goal (accurate future demand prediction) required by the vehicle redistribution system. Thus, integrating Ara’s teaching to utilize a bidirectional RNN for the core prediction task represents a logical and expected enhancement to improve the input data quality for the already-combined redistribution and optimization framework.
Claim Limitations Not Mapped by the Combination of Murakami, Karamanis, and Ara
After combining the teachings of Murakami, Karamanis, and Ara, the following is not explicitly disclosed:
wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations
Disclosure by Crawford
Crawford renders obvious:
wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations
See at least: "The resulting delta is bit shifted by two places to the right... A truncation is then performed... saving only the first (i.e., least significant) 16 bits. A reduced precision delta... is thus saved." ([0045]); "reducing all but the initial GPS position from 64 bits to 32 bits, a significant amount of bandwidth can be saved." ([0045]).
Rationale: Crawford discloses a specific, operative technique for data compression: a computed value (a positional delta) is represented in a reduced-bit format (16 bits) relative to its original, full-precision source. This directly teaches the general principle of encoding a result value (e.g., a computed delta, index, or other derived numerical output) such that the number of bits required for its representation is less than or equal to the number of bits used for the original data from which it was derived. While Crawford does not specifically discuss vehicle arrival counts, its teaching of a general method for reducing the bit-length of derived numerical data makes it obvious to implement the claimed bit-compression inequality within the combined vehicle-sharing and prediction system.
Motivation to Combine Murakami, Karamanis, Ara, and Crawford
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, and Crawford before them, to apply Crawford's bit-reduction compression technique to the numerical result values produced by the combined vehicle-sharing, prediction, and optimization system. The system resulting from the combination of Murakami, Karamanis, and Ara inherently processes substantial numerical data—including cluster indices, prediction outputs, and vehicle counts—where efficiency in storage and transmission is a recognized concern, as evidenced by Karamanis's explicit teaching of data size reduction through clustering. Crawford directly addresses this efficiency concern by teaching a well-established method for compressing sequential numerical values through bit-shifting and truncation to reduce their bit-length. Faced with the practical implementation of the combined system, a skilled artisan would have been motivated to employ Crawford's compression technique as a routine optimization, thereby naturally achieving the claimed condition where the result value is represented with a number of bits equal to or less than that of the original vehicle arrival data.
Regarding Claim 18,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 16.
Disclosure by Murakami
Murakami discloses
wherein the route setting unit
see Col. 6, ll. 40-44: a vehicle relocation setting section 107 … gives an instruction of relocation;
Rationale: Murakami’s vehicle relocation setting section 107 issues relocation instructions—i.e., decides routing of excess vehicles—therefore embodies the claimed route-setting unit that computes movement routes between stations in the mobility sharing network).
Disclosure by Karamanis
Karamanis discloses
configures the MCMF algorithm
see at least Pg. 3: we therefore propose a minimum cost flow formulation in the form of resource allocation to solve the vehicle redistribution problem;
Rationale: Karamanis “proposes a minimum cost flow formulation,” confirming the route-setting module configures an MCMF algorithm to compute optimal relocations, thus meeting the requirement of configuring the algorithm within system logic)
with a start point,
see at least Pg. 3: We classify vertices with b(v)>0 as source vertices;
Rationale: Vertices having positive balance are declared sources; that designation establishes the network’s single start point from which supply originates, satisfying limitation)
first nodes connected to the start point
see at least Pg. 3: vertices in A have directed edges which connect to the vertices in B;
Ratioanle: Karamanis links every source vertex in set A to downstream decision vertices, showing that first-tier nodes (representing station origins) are directly connected to the start point via explicit edges there. Also, the sentence states those edges emanate from A’s sources toward subsequent nodes; direction confirms connectivity “to the start point,” fulfilling the requirement of explicit linkage orientation within flow graph
and corresponding to the stations,
see at least Pg. 3: We associate vertices in each set with the set of road network clusters J such that A = K = L =M = C = J;
Rationale: Every vertex in A maps one-to-one onto road-network clusters (station groups); thus nodes explicitly correspond to the stations, satisfying the limitation linking model vertices with physical station entities precisely)
an end point,
see at least Pg. 4: We also introduce sink vertex t;
Rationale: Karamanis introduces a unique “sink vertex t,” representing the terminal node collecting outbound flow; this singular sink functions as the claimed network end point in the MCMF configuration, effectively ensuring closure)
second nodes connected to the end point
see at least Pg. 4:…directed edges from all vertices in C to the sink vertex t;
Rationale: Karamanis edges connect every vertex in set C to sink t, confirming that second-tier nodes (C, representing destination stations) are connected to the end point as required explicitly. Also, Karamanis notes “directed edges … to the sink t,” establishing direct connectivity from the second nodes to the end point specified by the claim)
and corresponding to the stations,
see at least Pg. 3: We associate vertices in each set with the set of road network clusters J such that A = K = L =M = C = J;
Rationale: Equating |C| with |J| ties each C-vertex to one physical cluster (station set), so second nodes unmistakably correspond to stations, matching this limitation’s correspondence requirement with exact mapping)
wherein the route setting unit sets a cost
see at least Pg. 5, Eq. 14: cij (xij)=0 ∀(i,j)∈E;
Rationale: Karamanis defines a class of “zero cost edge[s]” with cost 0, explicitly specifying cost values; the route-setting unit thus sets edge costs for implementation as the claim requires)
between the start point and the first nodes to be zero
see at least Pg. 3: vertices in A have directed edges … and Pg. 5, Eq. 14: cij (xij)=0 ∀(i,j) ∈ E;
Rationale: The excerpt confirms an edge exists from each source vertex (start point) to its decision vertex, establishing the specific arc “between the start point and the first nodes” required. The same sentence pairs every A-vertex with its B-node partner, illustrating the start-to-first-node linkage; thus, first nodes receive edges directly from the start sources within flow representation. The zero-cost declaration applies precisely to E-class edges linking start sources to control vertices, making the cost between start point and first nodes equal zero by definition)
and a capacity
see at least Pg. 5 Eq. 19; b(i)=Si ∀i∈A
Rationale: In a minimum-cost-flow network, the balance b(i) at a source vertex is the maximum flow that can leave that vertex; therefore sets the edge-capacity numerically)
to be proportional
see at least Pg. 5 Eq. 19; b(i)=Si ∀i∈A
Rationale: Giving a direct linear proportionality (k = 1) between capacity and vehicle count)
to the number of location vehicles of stations corresponding to the first nodes
see at least Pg. 6: Parameter Si in (19) and (20) denotes the available vehicles Si in cluster n(i) at the start of the current period;
Rationale: Explicitly the count of vehicles located at station/cluster i (the first-node’s station). Because capacity is Si, it is tied to that vehicle number)
in a time interval immediately preceding the target time interval,
see at least Pg. 6: …at the start of the current period;
Rationale: “Current period” is the epoch that ends when the optimization determines routes for the next (target) interval. Thus, Si represents vehicles counted immediately before the target interval)
wherein the route setting unit sets a capacity between the first nodes and the second nodes to be infinite
see at least Eq. 18: uij=∞ ∀(i,j)∈E;
Rationale: Infinite-capacity relocation arcs: Equation (18) sets u_{ij} equals infinity for all edges. First-to-second arcs belong to set E; therefore, their capacity is infinite, fully meeting limitations)
and a cost to be proportional to distances between the stations corresponding to the first nodes and the stations corresponding to the second nodes,
see at least Eq. 11: cij(xij)=r1 n(i)n(j)CMxij ∀(i,j)∈C;
Rationale: Distance-weighted edge cost: Equation (11) multiplies CM by r^{1}_{n(i)n(j)}, where that term equals average travel time between clusters. Cost therefore rises directly with inter-station distance, satisfying the proportionality requirement)
and wherein the route setting unit sets a cost between the second nodes and the end point to be zero
see at least Eq. 14: cij(xij)=0 ∀(i,j)∈E;
Rationale: Zero-cost sink edges: Equation (14) states c_{ij} equals zero for every E-edge. Since C→t edges reside in E, cost from each second node to the end point is zero, fulfilling limitation)
and a capacity to be proportional to the number of location vehicles at stations corresponding to the second nodes in the target time interval
see at least Pg. 3: for resulting states, we consider the numbers of available vehicles at the beginning of epoch t + 1 at each cluster and Pg. 4: Furthermore, … directed edges from all vertices in C to the sink vertex t;
Rationale: Each C-vertex stores its epoch t+1 vehicle count; the only outgoing C → t edge transmits exactly that amount, so edge capacity is linearly proportional to second-node vehicles during the target interval. A PHOSITA would modify Karamanis et al. to include explicit station capacity constraints, proportional to location vehicles. This is crucial for practical feasibility in real-world systems with finite parking/charging spaces. It ensures optimal redistribution plans are implementable, aligning with industry best practices).
Motivation to Combine Murakami, Karamanis, Ara, and Crawford
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, and Crawford before them, to implement Murakami’s vehicle relocation setting section 107 (route setting unit that “gives an instruction of relocation”) using Karamanis’s minimum-cost flow / minimum-cost maximum-flow formulation (with a source, sink, station-corresponding nodes, edge costs, and capacities) driven by the predicted station-level departures/arrivals produced by the prediction unit, because Murakami and Karamanis address the same shared-mobility redistribution technical problem (deciding which vehicles should be moved between stations and along which routes), and Karamanis provides a conventional, mathematically structured MCMF solver that predictably yields lower relocation cost (distance-weighted edge costs) and feasible rebalancing plans (capacity constraints) when integrated into Murakami’s existing relocation-instruction framework using the same station/port demand and inventory quantities already produced by the combined forecasting components.
Claims 2 and 3 are rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Mahony et al., herein after will be referred to as Mahony.
Regarding Claim 2,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 2.
Claim Elements not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
However, Murakami, Karamanis, Ara, and Crawford do not explicitly disclose:
wherein the first vehicle determination unit sets the plurality of unit areas using a clustering model that is an unsupervised learning model.
Disclosure by Mahony
Mahony discloses:
wherein the first vehicle determination unit sets the plurality of unit areas
See at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster; Each cluster groups multiple stations into a single analytic region.
Rationale: Creating several such clusters divides the service region into numerous distinct unit areas, satisfying the limitation’s “sets plurality of unit areas.”)
using a clustering model
See at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster; relied on clustering stations based on their observed usage;
Rationale: Explicitly states that a clustering model underpins the approach; consequently, the first vehicle-determination unit employs a clustering model exactly as the limitation requires)
that is an unsupervised learning model
see at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: clustering stations based on their observed usage; relied on clustering stations based on their observed usage;
Rationale: Clustering partitions data solely from un-labeled usage vectors, inherently unsupervised; because Mahony’s method clusters without labels or targets, the employed clustering model is an unsupervised learning model. Mahony discloses an unsupervised clustering model to analyze bike-share station behavior and inform rebalancing strategies.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Mahony
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Mahony before them, to modify the mobility sharing system’s first vehicle determination unit (as established by Murakami/Karamanis/Ara/Crawford) to set the plurality of unit areas by clustering stations based on observed usage (i.e., assigning stations exhibiting similar rush-hour behavior to the same cluster), and to then perform the vehicle-number determination/prediction operations at the cluster (unit-area) level, because Mahony expressly teaches that clustering stations by observed usage is used to inform operations “to anticipate user demand” and is “highly successful” in tailoring rebalancing decisions, yielding the predictable benefit of improving demand-aware allocation/rebalancing decisions while reducing the complexity/noise of per-station modeling.
Regarding Claim 3,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 2.
Claim Elements not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
However, Murakami, Ara, and Karamanis do not explicitly disclose
wherein the first vehicle determination unit sets the plurality of unit areas such that, based on historical data, there is a high probability that a departure station and an arrival station of a vehicle using a one-way service are included in the same unit area.
Disclosure by Mahony
Mahony discloses
wherein the first vehicle determination unit sets the plurality of unit areas
See at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster;
Rationale: Each cluster groups multiple stations into a single analytic region; creating several such clusters divides the service region into numerous distinct unit areas, satisfying the limitation’s “sets plurality of unit areas.”
such that, based on historical data,
See at least Pg. 4 of 8, Sec. 4: Using historical data that indicate which stations accumulate bikes (producers) and which lose bikes (consumers);
Rationale: Clustering relies on archived ride histories (“historical data”) to decide station roles, so every unit-area boundary is derived from prior-period evidence, directly fulfilling the limitation’s historical-data premise)
there is a high probability
See at least Pg. 4 of 8, Sec. 4: The behavior of stations…is very consistent;
Rationale: Consistent, repeatable station behavior denotes a high probability that observed origin-destination patterns persist; the consistency statement therefore meets the claim’s probabilistic requirement)
that a departure station
See at least Pg. 3 of 8, Computing Demands for Bikes: number of bikes out on day i at time j;
Rationale: “Bikes out” denotes every departure event, supplying the departure-station element needed for the probability statement)
and an arrival station
see at least Pg. 4 of 8, expected number of bikes in and out of the station for each minute of the rush hour;
Rationale: Bikes in” captures each arrival event, giving the arrival-station side of the origin-destination pair)
of a vehicle using a one-way service
see at least Pg. 689: Computing Demand for Bikes…Consider a bike-share station at Penn Station in the evening rush hour. A huge number of commuters want to return bikes and take a train from the station;
Rationale: depicts riders returning bikes at a different station from where they began, illustrating the system’s one-way rental nature, thereby satisfying the claim’s one-way-service requirement)
are included in the same unit area
See at least Pg. 688: stations that experience similar behavior during rush-hours will belong to the same cluster;
Rationale: Clustering groups origin-and-destination stations with like rush-hour flows into the same cluster (unit area), making it highly probable departure and arrival lie within one unit area, satisfying the limitation).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Mahony
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Mahony before them, to set the plurality of unit areas by clustering stations using historical trip/usage data so that stations exhibiting strong, consistent intra-group trip behavior are grouped into the same unit area, thereby increasing the likelihood that one-way trip departures and arrivals fall within the same unit area, because Mahony teaches forming operational clusters from historical station behavior to support rebalancing, and this clustering predictably improves the tractability and accuracy of Ara’s zone-level demand prediction and Karamanis’s cluster-to-cluster flow optimization when integrated into Murakami’s relocation framework, while also reducing cross-unit transfers and associated computation/communication burden.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, in view of Mahony, and in view of Lahoorpoor et al., herein after will be referred to as Lahoorpoor.
Regarding Claim 4,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 1.
Claim Elements not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein the first vehicle determination unit sets the plurality of unit areas such that a sum of station costs of each of all stations included in the entire area is minimized, and wherein the station costs are proportional to a sum of a square of distances from one station included in one unit area to the remaining stations included in the one unit area.
Disclosure by Mahony
Mahony discloses:
wherein the first vehicle determination unit sets the plurality of unit areas
See at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster;
Rationale: Each cluster groups multiple stations into a single analytic region; creating several such clusters divides the service region into numerous distinct unit areas, satisfying the limitation’s “sets plurality of unit areas.”.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Mahony
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Mahony before them, to implement the clustering/“unit area” setting step using a standard distance-minimizing clustering objective that minimizes an overall within-cluster dispersion metric—i.e., selecting unit areas that minimize an aggregate station cost proportional to summed squared inter-station distances within each unit area— because Murakami and Karamanis both treat distance as a primary relocation cost driver, and a PHOSITA would predictably choose a squared-distance within-cluster cost (a conventional clustering criterion) to produce compact unit areas that reduce relocation travel cost and improve downstream prediction/optimization performance.
Claim Elements not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, Crawford, and Mahony
such that a sum of station costs of each of all stations included in the entire area is minimized, and wherein the station costs are proportional to a sum of a square of distances from one station included in one unit area to the remaining stations included in the one unit area.
Disclosure by Lahoopoor
Lahoorpoor discloses:
such that a sum of station costs of each of all stations included in the entire area is minimized,
see at least Pg. 9 of 12: utilizes the agglomerative hierarchical clustering algorithm using the Ward method that minimizes the total within-cluster variance;
Rationale: The Ward method aims to minimize the total within-cluster variance (total within-cluster sum of squares (TWSS)). TWSS is one scalar that aggregates the cost term for every station; the algorithm explicitly minimizes that summed cost across the entire network. Lahoorpoor models the rebalancing problem as an "optimization problem, which aims to minimize the tour length." While this is the overall optimization objective for rebalancing, the clustering itself is based on a "similarity measure based on the trips between stations" to discover "groups of correlated stations". More specifically, Lahoorpoor uses the "agglomerative hierarchical clustering algorithm using the Ward method that minimizes the total within-cluster variance." )
and wherein the station costs are proportional to a sum of a square of distances from one station included in one unit area to the remaining stations included in the one unit area
See at least Pg. 9 of 21, Equation 7: The initial cluster distances in Ward’s minimum variance method are defined to be the squared Euclidean distance between points, which are shown in Equation (7), in which dq p stands for the squared Euclidean distance between two stations q and p;
Rationale: The Ward method's objective is to minimize the total within-cluster variance. While not explicitly stated as "station costs proportional to a sum of a square of distances from one station to the remaining stations in the unit area", the mathematical basis of Ward's method directly involves minimizing the sum of squared distances of points (stations) within a cluster to their cluster's centroid, which is functionally equivalent to minimizing the variance. Minimizing within-cluster variance in a clustering algorithm aims to group data points close to each other in terms of squared Euclidean distance. If the "station cost" is defined as a sum of squared distances within a cluster, then a clustering algorithm that minimizes within-cluster variance (like Ward's method) would inherently be minimizing that sum).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Mahony, and Lahoopoor
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Mahony, and Lahoopoor before them, to incorporate Lahoopoor’s Ward-method hierarchical clustering objective (minimizing total within-cluster variance based on squared Euclidean distance between stations) as the specific criterion for setting the “unit areas” (clusters) in Mahony’s station-clustering step, because Murakami and Karamanis expressly treat inter-station distance as a primary driver of relocation/redistribution cost, and Ward’s squared-distance minimization predictably yields compact station groupings that reduce travel distance/tour length for rebalancing and improve the efficiency and stability of the downstream demand-prediction and flow-optimization operations without changing the fundamental functionality of the combined system.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, in view of Mahony, and in view of Li et al., herein after will be referred to as Li.
Regarding Claim 5,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 1.
Claim Limitations not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein the first vehicle determination unit sets the plurality of unit areas such that a sum of absolute values of area differences between the plurality of unit areas is minimized.
Disclosure by Mahony
Mahony discloses:
wherein the first vehicle determination unit sets the plurality of unit areas
See at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster;
Rationale: Each cluster groups multiple stations into a single analytic region; creating several such clusters divides the service region into numerous distinct unit areas, satisfying the limitation’s “sets plurality of unit areas.”.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Mahony
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Mahony before them, to implement the clustering/“unit area” setting step using a standard distance-minimizing clustering objective that minimizes an overall within-cluster dispersion metric—i.e., selecting unit areas that minimize an aggregate station cost proportional to summed squared inter-station distances within each unit area— because Murakami and Karamanis both treat distance as a primary relocation cost driver, and a PHOSITA would predictably choose a squared-distance within-cluster cost (a conventional clustering criterion) to produce compact unit areas that reduce relocation travel cost and improve downstream prediction/optimization performance.
Claim Limitations not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, Crawford, and Mahony
such that a sum of absolute values of area differences between the plurality of unit areas is minimized.
Disclosure by Li
Li, discloses
such that a sum of absolute values of area differences between the plurality of unit areas is minimized
See at least Pg. 92, Sec. 2.1 Service Area and Demand: The service area is divided into n zones with length L and width W/n;
Rationale: Li discloses "that a sum of absolute values of area differences between the plurality of unit areas is minimized" by describing a system where all "unit areas" (which they call "zones") are designed to have identical dimensions and thus identical areas. Li partitions the service rectangle into n zones of fixed length L and width W/n, guaranteeing that every zone's area equals L·W/n. Consequently, |Ai − Aj| is zero for all pairs, driving the total absolute area-difference metric to its mathematical minimum, zero. A PHOSITA would adopt this equal-area rule to balance workload and ensure equitable coverage with routine design effort and no inventive insight).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Mahony, and Li
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Mahony, and Li before them, to constrain or post-process the “plurality of unit areas” produced by the clustering-based region-setting of Mahony/Karamanis so that the resulting unit areas are defined as equal-area zones as taught by Li, because equal-area partitioning predictably minimizes the sum of absolute values of area differences between unit areas (driving the objective toward its minimum) and yields balanced geographic coverage and workload distribution across zones, which in turn improves the operational stability of Murakami’s allocation/relocation decisions and the reliability and tractability of the downstream demand prediction and network-flow optimization in Karamanis/Ara, using only routine design choice and known zoning practice.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, in view of Mahony, and in view of Yu et al., herein after will be referred to as Yu.
Regarding Claim 6,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 6.
Claim Limitations not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
Murakami, Karamanis, Ara, and Crawford do not explicitly disclose:
wherein the first vehicle determination unit sets the plurality of unit areas such that a station closest to one station included in the entire area is included in the same unit area with the one station.
Disclosure by Mahony
Mahony discloses:
wherein the first vehicle determination unit sets the plurality of unit areas
see at least Pg. 2 of 8, Sec. 3: Data Analytics for Bike Share Systems: stations that experience similar behavior during rush-hours will belong to the same cluster
Rationale: Each cluster groups multiple stations into a single analytic region; creating several such clusters divides the service region into numerous distinct unit areas, satisfying the limitation’s “sets plurality of unit areas.”
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Mahony
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Mahony before them, to incorporate Mahony’s station-clustering approach into Murakami’s mobility sharing system (as economically optimized by Karamanis and forecast-enhanced by Ara) to set the plurality of unit areas as clusters of stations exhibiting similar historical usage behavior, because doing so is a predictable and well-known way to define operational regions that stabilize rebalancing decisions, reduce unnecessary relocation, and improve the tractability and accuracy of the downstream prediction/optimization workflow.
Claim Limitations not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, Crawford, and Mahony
However, Murakami, Karamanis, Ara, Crawford, and Mahony do not explicitly disclose:
such that a station closest to one station included in the entire area is included in the same unit area with the one station.
Disclosure by Yu
Yu discloses:
such that a station closest to one station included in the entire area is included in the same unit area with the one station
See at least Pg. 4 of 17, Sec. 2.3.1. Algorithm Framework: First, each location point and its nearest neighbors are grouped into one cluster;
Rationale: Yu’s QTNA algorithm defines each station’s nearest-neighbor set NN(Oᵢ) and builds a provisional cluster from all points within radius θᵣ (Alg. 1, Step 2). θᵣ expands until the cluster meets a minimum-size test, so it can never shrink below the distance to Oᵢ’s closest station; that neighbor therefore remains in the same final cluster. Because accepted clusters are never split, this nearest-neighbor contiguity persists throughout QTNA. A PHOSITA would readily swap Murakami’s ad-hoc port zones for Yu’s proximity-based clusters, improving rebalancing efficiency with existing pairwise-distance data and no new technical hurdles).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Mahony, and Yu
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Mahony, and Yu before them, to implement the unit-area (cluster) setting step using Yu’s nearest-neighbor grouping rule (grouping each station with its nearest neighbor(s) into one cluster) as a design constraint on Mahony’s clustered unit areas, because nearest-neighbor contiguity is a conventional clustering principle that predictably yields compact, proximity-consistent unit areas, which directly reduces inter-station travel distance/cost for Murakami’s relocation actions and provides more coherent spatial zones for Ara’s demand prediction inputs and Karamanis’s flow-based redistribution optimization.
Claims 7-9 are rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Harrou et al., herein after will be referred to as Harrou.
Regarding Claim 7,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 7.
Claim limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein the bidirectional RNN model is configured with a bidirectional gated recurrent unit (Bi-GRU) consisting of at least two stacked layers of RNN algorithm
Disclosure by Harrou
Harrou discloses:
wherein the bidirectional RNN model is configured
See at least: “the bi-directional recurrent neural networks (BiLSTM and BiGRU) … processing data in the forward and backward direction” (19/22).
Rationale: Harrou’s express description of “bi-directional recurrent neural networks (BiLSTM and BiGRU)” and “forward and backward direction” corresponds to wherein the bidirectional RNN model is configured.
with a bidirectional gated recurrent unit (Bi-GRU)
See at least: “bidirectional GRU (BiGRU)” (p. 2/22).
Rationale: Harrou expressly names “bidirectional GRU (BiGRU),” which corresponds to with a bidirectional gated recurrent unit (Bi-GRU).
consisting of at least two stacked layers of RNN algorithm
See at least: “deep recurrent neural networks are built by stacking two recurrent layers” (Harrou, p. 11/22); “Similarly, the same architecture is used for BiLSTM and BiGRU models” (p. 11/22).
Rationale: Harrou’s “stacking two recurrent layers” disclosure, together with Harrou’s express statement that “the same architecture is used for BiLSTM and BiGRU models,” corresponds to consisting of at least two stacked layers of RNN algorithm.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Harrou
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Harrou before them, to configure the bidirectional RNN model of the mobility sharing system using Harrou’s bidirectional GRU (BiGRU) deep bidirectional recurrent architecture with stacked recurrent layers, because Harrou provides a conventional, well-understood bidirectional recurrent implementation (BiGRU with stacked layers) for improving sequence forecasting, and a PHOSITA would expect predictable performance benefits when applying such bidirectional gated recurrent architectures to spatiotemporal demand prediction in shared-mobility systems.
Regarding Claim 8,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 8.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein the bidirectional RNN model is configured with a bidirectional long short-term memory (Bi-LSTM) consisting of at least two stacked layers of RNN algorithm
Disclosure by Harrou
Harrou discloses:
wherein the bidirectional RNN model is configured with a bidirectional long short-term memory (Bi-LSTM)
see at least Pg. 12 of 21: …we compared the performance of the proposed GAHD-VAE approach to that of GRU, LSTM, BiLSTM, BiGRU, CNN, and ConvLSTM models because of their popularity in modeling and forecasting time-series data
Explicitly evaluates a “BiLSTM” architecture, demonstrating a bidirectional long-short-term-memory network configuration. This direct disclosure shows the underlying bidirectional RNN can be instantiated with Bi-LSTM cells, satisfying claim limitation)
consisting of at least two stacked layers of RNN algorithm
See at least: “deep recurrent neural networks are built by stacking two recurrent layers” (Harrou, p. 11/22); “Similarly, the same architecture is used for BiLSTM and BiGRU models” (p. 11/22).
Rationale: Harrou’s “stacking two recurrent layers” disclosure, together with Harrou’s express statement that “the same architecture is used for BiLSTM and BiGRU models,” corresponds to consisting of at least two stacked layers of RNN algorithm.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Harrou
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Harrou before them, to configure the bidirectional RNN model of the parent claim's system using Harrou's BiLSTM implemented with at least two stacked recurrent layers, because Harrou teaches that such a stacked BiLSTM architecture is a known and conventional model for improving sequence forecasting accuracy, and a PHOSITA would have been motivated to apply this predictable and effective deep learning architecture to the analogous problem of demand forecasting in a shared-mobility system to achieve improved prediction performance.
Regarding Claim 9,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 9.
Claim limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein the bidirectional RNN model consists of at least two stacked layers of RNN algorithm.
Disclosure by Harrou
Harrou discloses:
wherein the bidirectional RNN model
See at least: “Similarly, the same architecture is used for BiLSTM and BiGRU models: Deep bidirectional temporal feature extractors … Generally, the bidirectional models allow the input to be processed in the forward and backward direction …” (Page 12/22)
Rationale: Harrou’s “BiLSTM” and “BiGRU” are expressly identified as bidirectional models, and Harrou expressly characterizes these models as bidirectional processing in forward and backward directions, which corresponds to the bidirectional RNN model recited in the claim.
consists of at least two stacked layers of RNN algorithm.
See at least: “Here, deep recurrent neural networks are built by stacking two recurrent layers …” (Page 12/22)
Rationale: Harrou expressly teaches “stacking two recurrent layers” to build deep recurrent neural networks, which meets the requirement that the bidirectional RNN model consists of at least two stacked layers of RNN algorithm.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Harrou
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Harrou before them, to implement the bidirectional RNN demand-prediction component in the Claim 1 mobility sharing system using Harrou’s expressly-described deep recurrent architecture formed by stacking two recurrent layers (and applied to bidirectional recurrent models), because Harrou provides a conventional deep bidirectional recurrent configuration for time-series forecasting and a PHOSITA would have expected predictable improvements in sequence-modeling capacity when applying that known architecture to shared-mobility demand prediction.
Claims 10-12 are rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Yin et al., herein after will be referred to as Yin.
Regarding Claim 10,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 10.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
However, Murakami, Karamanis, Ara, and Crawford do not explicitly disclose
wherein the prediction unit combines a time function with the prediction data set to input the combined time function with the prediction data set into the bidirectional RNN model, wherein the time function being combined with the prediction data set is proportional to sin (2πi/T) or cos (2πi/T), wherein T is a period of the plurality of unit time intervals, and i is an order of the unit time intervals corresponding to the prediction data set among the unit time intervals included in the period.
Disclosure by Yin
Yin discloses:
Wherein the prediction unit combines a time function with the prediction data set
See at least Pg. 1, Sec II Approaches: transform the feature hour into cos (2π/24 · hour) and sin (2π/24 · hour), since hour is intuitively a periodic function;
Rationale: Explicitly adds cyclical sine/cosine terms—i.e., time functions—to other features, satisfying the requirement to combine a time function. The cyclical functions are merged with the historical-demand dataset, meeting the limitation of combining with the prediction dataset)
to input the combined time function with the prediction data set into the bidirectional RNN model
See at least Pg. 1, Sec II Approaches:…feeding…inputs into the learning algorithms… feature engineering approaches …;
Rationale: Yin states the engineered (time-augmented) features are what is fed into the forecasting model; a PHOSITA would feed the identical vector into Ara’s Bi-LSTM),
wherein the time function being combined with the prediction data set is proportional to sin (2πi/T) or cos (2πi/T),
See at least Pg. 1, Sec II Approaches: transform the feature hour into cos (2π/24 · hour) and sin (2π/24 · hour), since hour is intuitively a periodic function;
Rationale: Explicitly adds cyclical sine/cosine terms—i.e., time functions—to other features, satisfying the requirement to combine a time function. The cyclical functions are merged with the historical-demand dataset, meeting the limitation of combining with the prediction dataset. Yin’s exact formulas match the claimed proportional sin/ cos structure)
wherein T is a period of the plurality of unit time intervals,
See at least Pg. 1, Sec II Approaches: transform the feature hour into cos (2π/24 · hour) and sin (2π/24 · hour)
Rationale: In the formula, the denominator “24” is the daily cycle length. T = 24 hours—the period covering the recurrent hourly intervals used),
and i is an order of the unit time intervals corresponding to the prediction data set among the unit time intervals included in the period
See at least Pg. 1, Sec II Approaches: transform the feature hour into cos (2π/24 · hour) and sin (2π/24 · hour);
Rationale: The variable “hour” iterates 0 – 23 each day, serving as the ordered index i within the 24-hour period. Hour’s value corresponds to the claim’s indexed interval variable i).
wherein each of the plurality of unit time interval is set to one hour and the T is set to 24, or each of the plurality of unit time intervals is set to one day and the T is set to 7.
See at least Pg. 1, Sec. I: “contains 17379 rows of hourly count of rental bikes (CRB) … These features will be manipulated and used to predict the hourly CRB.”; Pg. 1, Sec. II: “transform the feature ‘hour’ into cos (2π/24 · hour) and sin (2π/24 · hour), since hour is intuitively a periodic function.”
Rationale: Yin expressly defines the prediction granularity as “hourly,” which sets each unit time interval to one hour, and Yin’s time-function transform uses “24” as the period divisor for the hour feature, which sets T=24 for that hourly periodicity. Because the claim is an OR, Yin’s explicit disclosure of this first branch is sufficient without needing the alternative one-day/T=7 branch.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Yin
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Yin before them, to incorporate Yin’s periodic time-function feature encoding (sin/cos of a period-normalized index) into the prediction data set supplied to Ara’s bidirectional RNN within the shared-mobility framework of Murakami and the zone-based modeling of Karamanis, because Yin expressly teaches that time-of-day is periodic and should be transformed into sine/cosine components before being fed as model inputs to improve forecasting of demand that exhibits daily cycles, and a PHOSITA would have expected predictable gains in departure/arrival prediction accuracy—and thus more reliable allocation/relocation decisions—by adding these known, standard temporal features to the existing trip-history based prediction pipeline.
Regarding Claim 11,
The combination of Murakami, Karamanis, Ara, Crawford, and Yin establishes the mobility sharing system of Claim 10, which is the basis for Claim 11.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein each of the plurality of unit time intervals corresponds to one hour, and the T corresponds to 24 hours.
Disclosure by Yin
Yin discloses:
wherein each of the plurality of unit time intervals corresponds to one hour
See at least Pg. 1, Sec I: 17379 rows of hourly count of rental bikes (CRB) … the dataset … contains … hourly count of rental bikes;
Rationale: Dataset is discretized in one-hour slots; every unit interval the model handles is exactly one hour),
and the T corresponds to 24 hours
See at least Pg. 1, Sec II Approaches: transform the feature ‘hour’ into cos(2π / 24 × hour) and sin(2π / 24 × hour), since hour is intuitively a periodic function;
Rationale: Using 2π/24 fixes the period at 24 hours, explicitly setting T = 24 within the temporal encoding).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Yin
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Yin before them, to incorporate Yin’s periodic time-function feature encoding (sin/cos of a period-normalized index) into the prediction data set supplied to Ara’s bidirectional RNN within the shared-mobility framework of Murakami and the zone-based modeling of Karamanis, because Yin expressly teaches that time-of-day is periodic and should be transformed into sine/cosine components before being fed as model inputs to improve forecasting of demand that exhibits daily cycles, and a PHOSITA would have expected predictable gains in departure/arrival prediction accuracy—and thus more reliable allocation/relocation decisions—by adding these known, standard temporal features to the existing trip-history based prediction pipeline.
Regarding Claim 12,
The combination of Murakami, Karamanis, Ara, Crawford, and Yin establishes the mobility sharing system of Claim 10, which is the basis for Claim 12.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
wherein each of the plurality of unit time intervals corresponds to one day, and the T corresponds to seven days.
Disclosure by Yin
Yin discloses:
wherein each of the plurality of unit time intervals corresponds to one day,
See at least Pg. 1 Introduction: The dataset … contains date/hour, seasons, holiday/working-day/ weekday…
Rationale: Claim 12’s “unit time interval” is expressly a day-long segment. Yin’s use of the “weekday” attribute to characterize demand over individual days meets that requirement: every training instance is associated with—and predictions are conditioned on—its specific day in the weekly cycle. No additional programming is needed to extract daily intervals; the source data and model features already provide them)
and the T corresponds to seven days
See at least Pg. 1, Sec. II Approaches: Digitization: Some features are “discrete” by their nature. For instance, there are 4 seasons in any year, and they are supposed to be independent….the digitalization can be applied to weekday, season, and weather;
Rationale: Claim 10 (from which Claim 12 depends) introduces the sin/cos time-function f(i) = sin (2π i /T) or cos (2π i /T). Claim 12 specifies T=7 days, producing a weekly sine/cosine cycle. Yin accomplishes the same weekly periodic encoding by digitizing the “weekday” feature into seven one-hot components, each fired once per seven-day cycle. Functionally, this embeds the identical seven-day periodicity that Claim 12 recites).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Yin
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Yin before them, to implement the periodic time-function feature on a daily prediction granularity using a weekly period (seven days), because Yin expressly teaches that demand exhibits systematic day-of-week effects by using “date”/“weekday” information as model inputs, and a PHOSITA would have predictably incorporated that same seven-day periodicity into the time-function encoding (i.e., a weekly cycle) when aggregating trip-history features into one-day unit intervals for improved demand forecasts that drive Murakami’s allocation/relocation decisions within Karamanis’s zone-based framework.
Claims 13-15 are rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, in view of Yin, and in view of Rodrigues.
Regarding Claim 13,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 10, which is the basis for Claim 13.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
However, Murakami, Karamanis, Ara, and Crawford do not explicitly disclose
wherein the prediction unit generates the prediction data set by embedding a result value of subtracting the number of vehicles arriving at each of the stations from the number of vehicles departing from each of the stations, and wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations.
Disclosure by Yin
Yin discloses:
wherein the prediction unit generates the prediction data set
See at least Pg. 1, Sec. II. Approaches: Before feeding the raw feature inputs into the learning algorithms, we first study some possible ways to engineer the raw data…;
Rationale: Yin’s feature-engineering phase transforms variables and assembles them into an input matrix, squarely constituting generation of the prediction dataset by the system’s forecasting module, which functions as the claimed prediction unit).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Yin
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Yin before them, to implement Ara’s bidirectional RNN demand-prediction module using Yin’s disclosed feature-engineering pipeline for “generat[ing] the prediction data set” (i.e., transforming and assembling “raw feature inputs” prior to “feeding” them into the learning algorithm), because Yin teaches an established, field-compatible preprocessing step that predictably improves learning stability/accuracy for mobility-demand forecasting, and the combined Murakami/Karamanis system inherently benefits from more accurate predicted departures/arrivals used for allocation/redistribution decisions.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, Crawford, and Yin
Murakami, Karamanis, Ara, Crawford, and Yin do not explicitly disclose:
by embedding a result value of subtracting the number of vehicles arriving at each of the stations from the number of vehicles departing from each of the stations, and wherein the number of bits required to represent the result value is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations.
Disclosure by Rodrigues
Rodrigues discloses:
by embedding a result value
see at least Pg. 2: … net flow: the difference between the number of arrivals and departures;
Rationale: Rodrigues explicitly inserts the net-flow integer into the optimization inputs, thereby embedding a derived result value within the dataset exactly as the limitation requires for downstream predictive or planning analysis)
of subtracting the number of vehicles
see at least Pg. 2: … difference between the number of arrivals and departures
Rationae: Net flow is mathematically defined by Rodrigues as departures minus arrivals, therefore each embedded value originates from subtracting the arrival figures for every station, satisfying this subtraction-operand requirement comprehensively here)
arriving at each of the stations
see at least Pg. 2: … difference between the number of arrivals and departures;
Rationale: The examiner interprets Rodrigues’s phrase, naming arrivals explicitly confirming the metric treating arrivals per station as individual terms, thus aligning squarely with the limitation focusing on vehicles arriving at each station)
from the number of vehicles
see at least Pg. 2: …difference between the number of arrivals and departures;
Rationale: Rodrigues’s same sentence references departures, demonstrating that departures serve as the minuend in the subtraction, thereby meeting the limitation requiring subtraction from the number of vehicles departing per station counts)
departing from each of the stations,
see at least Pg. 2: …difference between the number of arrivals and departures;
Rationale: Because departures are explicitly called out by Rodrigues when defining net flow, the per-station departure counts exactly fulfill the clause concerning vehicles departing from each of the stations as claimed)
and wherein the number of bits required to represent the result value
see at least Pg. 2: "Figure 1 displays each station's net flow: the difference between the number of arrivals and departures."
Rationale: The net flow, as a derived integer value representing departures minus arrivals, is embedded in the dataset for optimization. It is a well-known principle in computer science, as evidenced by general data representation practices, that the difference between two integers, such as net flow, often requires fewer bits to represent than the original values when the difference is small in magnitude, as is common in balanced mobility sharing systems where arrivals and departures are close in number, as also noted by Rodrigues et al., Pg. 2: "a station with relatively big numbers of absolute outflow and inflow does not necessarily tend to become unbalanced")
is equal to or less than the number of bits required to represent the number of vehicles arriving at each of the stations
see at Pg. 2: "Figure 1 displays each station's net flow: the difference between the number of arrivals and departures"
Rationale: the net flow, being the difference between arrivals and departures, can be represented with fewer bits than the number of arrivals when the net flow is small, a common scenario in bike-sharing systems as described in Rodrigues et al., Pg. 2, where stations like 105 have near-zero net flow due to balanced usage. This is consistent with standard data compression techniques, where encoding differences reduce the bit requirement compared to raw values, ensuring that the number of bits for the net flow is equal to or less than that for arrivals in cases where the net flow’s magnitude is smaller or comparable to the arrival count). A PHOSITA, motivated by the problem of fleet imbalance (as highlighted by Rodrigues) and the desire for efficient prediction (as disclosed by Yin et al.), would have found it obvious to calculate the "net flow" (difference between departures and arrivals) and use this derived value as an input feature for a prediction model. The resulting representation, which inherently requires an equal or fewer number of bits, is a predictable outcome of such a conventional data preprocessing and feature engineering step.
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues before them, to generate the bidirectional-RNN input prediction data set by embedding Rodrigues’s expressly defined “net flow” (the per-station difference between arrivals and departures) as the claimed embedded “result value” derived from subtracting arrivals and departures, because that net-imbalance feature is a direct predictor of rebalancing need in shared-mobility systems and therefore predictably improves the downstream redistribution/optimization performed in Murakami/Karamanis; and to satisfy the claimed bit-length condition, a PHOSITA would further apply Crawford’s taught bit-reduction encoding of computed deltas/differences (bit-shifting/truncation to a reduced-precision representation) to the embedded net-flow result values to reduce storage/bandwidth while preserving functional utility, yielding a routine and predictable optimization within the same data-pipeline used to feed Ara’s bidirectional RNN.
Regarding Claim 14,
Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues establish the mobility sharing system of Claim 13, which is the basis for Claim 14.
Disclosure by Murakami
Murakami discloses
wherein the prediction unit generates the prediction data set
see at least Pg. 1, Sec. II. Approaches: Before feeding the raw feature inputs into the learning algorithms, we first study some possible ways to engineer the raw data…; Yin’s feature-engineering phase transforms variables and assembles them into an input matrix, squarely constituting generation of the prediction dataset by the system’s forecasting module, which functions as the claimed prediction unit).
Disclosure by Rodrigues
Rodrigues discloses:
such that the prediction data set represents the result value
see at least Pg. 2, Data analysis of a real-world case: Complementing this information, Figure 1 displays each station’s net flow: the difference between the number of arrivals and departures.
Rationale: Explicitly embeds the “net flow”—the exact result value defined in claim 13—into the dataset it analyses and feeds to the optimization model)
and the number of vehicles departing from each of the stations
see at least Pg. 3: Figure 1: Net flow of bicycles in each station for the period of January and February 2019.
Rationale: Rodrigues’s input table (built from operator data) contains a net-flow column and a departures column for every station, so the generated prediction dataset “represents the result value and the number of vehicles departing from each of the stations” exactly as Claim 14 specifies).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues before them, to generate, via Yin’s disclosed feature-engineering step, a prediction data set for input to Ara’s bidirectional RNN in which the data set represents (i) Rodrigues’s expressly defined per-station net flow “result value” (difference between arrivals and departures) and (ii) the per-station departures counts used to compute that result value, because Murakami/Karamanis already require station-level departure/arrival information to predict imbalance and execute redistribution, and Rodrigues teaches that net flow is a standard, informative station-imbalance descriptor derived from departures/arrivals that predictably improves rebalancing/forecasting features; further, a PHOSITA would implement the dataset representation efficiently using Crawford’s taught reduced-bit encoding of computed difference values as a routine bandwidth/storage optimization compatible with the same numeric feature vectors.
Regarding Claim 15,
Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues establish the mobility sharing system of Claim 13, which is the basis for Claim 15.
Disclosure by Murakami
Murakami discloses
wherein the prediction unit generates the prediction data set
see at least Pg. 1, Sec. II. Approaches: Before feeding the raw feature inputs into the learning algorithms, we first study some possible ways to engineer the raw data…; Yin’s feature-engineering phase transforms variables and assembles them into an input matrix, squarely constituting generation of the prediction dataset by the system’s forecasting module, which functions as the claimed prediction unit).
Disclosure by Rodrigues
Rodrigues discloses:
such that the prediction data set represents the result value
see at least Pg. 2, Data analysis of a real-world case:
Rationale: Complementing this information, Figure 1 displays each station’s net flow: the difference between the number of arrivals and departures.; Explicitly embeds the “net flow”—the exact result value defined in claim 13—into the dataset it analyses and feeds to the optimization model).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Yin, and Rodrigues before them, to have the prediction unit generate the prediction data set such that the prediction data set represents the result value (i.e., Rodrigues’s expressly defined per-station “net flow” difference between arrivals and departures) as an engineered input feature for Ara’s bidirectional RNN, because Murakami and Karamanis already operate on station-level imbalance derived from arrivals/departures to drive redistribution, Yin teaches assembling engineered/derived features into the model input dataset, and Rodrigues teaches using net flow as the standard station-imbalance result value—so incorporating that result value into the generated prediction data set would have been a predictable, routine design choice; further, Crawford’s reduced-bit encoding of computed deltas would have been applied as a conventional efficiency optimization for storing/transmitting such derived result values.
Claims 16 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Legros.
Regarding Claim 16,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 16.
Disclosure by Murakami
Murakami discloses:
wherein the second vehicle determination unit
see at least Fig. 2, Col. 6, ll. 40-43: A vehicle relocation setting section 107 … gives an instruction of relocation;
Rationale: Vehicle relocation setting section 107 determines relocation actions using excess-shortage calculations, functioning as the second vehicle determination unit that finalizes port-specific placement numbers after forecasts are generated for balanced service)
prioritizes distributing a vehicle
see at least Col. 8, ll. 34-35: vehicles to be relocated are sequentially set for ports P in increasing order of the evaluation values;
Rationale: Ports are visited sequentially by priority order; the first action is sending a vehicle to the highest-priority port—an explicit prioritized distribution)
to a station
see at least Col. 2, ll. 3-4: each of a plurality of ports set in an area;
Rationale: The "ports" in Murakami function as locations for vehicles, directly corresponding to the claimed "stations)
having a larger predicted result value
see at least Abstract: computing an evaluation value indicative of the degree of seriousness of vehicle shortage;
Rationale: Murakami calculates an "evaluation value" to quantify shortage severity, where a lower value indicates a more serious shortage for prioritization)
of subtracting the predicted number of arriving vehicles from the predicted number of departing vehicles
see at least Col. 6, formula f3: (forecast-occurrence vehicle count – forecast-occurrence ride demand count)
Rationale: Formula f3 is literally a subtraction operation, fulfilling the “of subtracting” clause. Also, “Forecast arrival trip count” is the predicted arriving-vehicle number used in the subtraction, and Forecast ride-demand count represents predicted departures; it is the minuend in Murakami’s difference, satisfying limitation)
among the stations
see at least Fig. 5, Step S4: Arrange Ports In The Increasing Order Of Evaluation Values;
Rationale: Instruction to process every port sequentially “in increasing order” shows the algorithm ranks and compares all stations together, explicitly choosing among them, thereby satisfying limitation)
satisfying the set condition,
see at least Abstract: In relocating vehicles to a port running short of vehicles … a vehicle relocation setting section sets relocation by considering the evaluation value, namely the degree of seriousness of vehicle shortage;
Rationale: The phrase “a port running short of vehicles” defines the very condition (vehicle shortage) that must be met before relocation is set. Thus, the Abstract discloses a relocation decision only after that shortage condition is satisfied—exactly what the limitation requires)
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
and wherein the set condition is a complaint occurrence due to a lack of vehicles, a full of vehicles, or an inclusion of an electric vehicle charging station.
Disclosure by Legros
Legros discloses
and wherein the set condition is a complaint occurrence
see at least Abstract: minimize the rate of arrival of unsatisfied users ;
Rationale: Unsatisfied-user arrivals are explicit complaints; Legros uses them as the governing trigger, matching “complaint occurrence)
due to a lack of vehicles, a full of vehicles, or an inclusion of an electric vehicle charging station
see at least Abstract: unsatisfied users who find their station empty or full;
Finding an empty station generates complaints caused by lack of vehicles, fulfilling the first enumerated complaint type verbatim. Station-full complaints correspond to “a full of vehicles.” At least one enumerated complaint (empty or full) is disclosed, so the OR condition is satisfied).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Legros
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Legros before them, to modify Murakami’s set condition for selecting/prioritizing stations for redistribution by incorporating Legros’s expressly taught “unsatisfied users”/complaint-driven triggers for when a station is “empty or full” (i.e., complaint occurrence due to a lack of vehicles or a full of vehicles), because both references address the same shared-mobility operational problem of when and where to reposition vehicles, and using complaint/unsatisfied-user events as an additional triggering/selection condition is a predictable, compatible control input that would improve service quality by directing redistribution to the stations most likely to generate user dissatisfaction.
Claims 17 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Legros, and in view of Schuijbroek et al., herein after will be referred to as Schuijbroek.
Regarding Claim 17,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 1, which is the basis for Claim 17.
Disclosure by Murakami
Murakami discloses
wherein the second vehicle determination unit
see at least Fig. 2, Col. 6, ll. 40-43: A vehicle relocation setting section 107 … gives an instruction of relocation;
Rationale: Vehicle relocation setting section 107 determines relocation actions using excess-shortage calculations, functioning as the second vehicle determination unit that finalizes port-specific placement numbers after forecasts are generated for balanced service)
prioritizes distributing a vehicle to a station
see at least Fig. 5, Steps S4 and S5:
Rationale: Murakami sorts ports in ascending evaluation value, checks them sequentially for shortages, and, upon finding a deficit, reallocates vehicles from the nearest surplus port—ensuring the most urgently undersupplied station receives service first. That is a priority-based relocation
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
However, Murakami, Karamanis, Ara, and Crawford do not explicitly disclose
distributing a vehicle to a station having a larger predicted result value of subtracting the predicted number of arriving vehicles from the predicted number of departing vehicles, and, among stations having the same predicted result value, prioritizes distributing a vehicle to a station having a larger sum of the predicted number of departing vehicles and the predicted number of arriving vehicles.
Disclosure by Legros
Legros discloses:
and among stations having the same predicted result value,
see at least Abstract: the rule of thumb … is to prioritize the closer, the more active … and the more imbalanced stations;
Rationale: Legros lists ranking drivers sequentially; after imbalance equivalence, the next criterion is “more active,” demonstrating that when several stations share the same net-imbalance value, ties are resolved subsequently among them),
prioritizes distributing a vehicle to a station
see at least Abstract: decide at any point of time (i) which station should be prioritized;
Rationale: Explicitly states the algorithm must choose which station should be prioritized, embodying a decision to dispatch the truck—i.e., distribute bikes—thus fulfilling the claimed ‘prioritizes distributing a vehicle’ action requirement)
having a larger sum
see at least Pg. 742: the intensity of the users’ activity;
Rationale: Legros names ‘intensity of users’ activity’ as a ranking factor; intensity is defined in the model by arrival plus departure rates, satisfying the need for a larger cumulative prediction sum. Legros explicitly identifies "the more active" (i.e., higher intensity of users' activity) as one of the key factors for prioritizing stations during repositioning. Within his mathematical model, this intensity is quantitatively defined as the sum of the bike arrival rate (λ_i,t) and the bike departure rate (μ_i,t) at a given station i and time t. This sum (λ_i,t + μ_i,t) captures the total rate of user interactions (returns + rentals) at the station, directly measuring its activity level)
of the predicted number of departing vehicles
see at least Table 1:Departure process in the zone μ(t) = μ(1 + sin…);
Rationale: Departure process μ(t) explicitly models forecasted bike pickups, providing the predicted number of departing vehicles required for computing the net-imbalance and subsequent priority rankings in the second determination unit logic)
and the predicted number of arriving vehicles
see at least Table 1 Arrival process in the zone λ(t) = λ…;
Rationale: Arrival process λ(t) models forecasted bike returns, delivering the predicted arriving vehicle count, which together with departures supplies the net-imbalance metric and cumulative activity measure used in the prioritization rules).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, and Legros
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, and Legros before them, to implement the station-selection/prioritization logic of the second vehicle determination unit such that, when multiple stations present the same predicted imbalance value, the second vehicle determination unit resolves the tie by prioritizing the station with the greater predicted activity level, because Murakami’s relocation framework already relies on ranking stations for redistribution, and Legros expressly teaches using “more active” stations as a ranking driver; a PHOSITA would predictably compute that activity directly from the already-available predicted departures and predicted arrivals as a summed measure to improve service impact of each redistribution action.
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, Crawford, and Legros
distributing a vehicle to a station having a larger predicted result value of subtracting the predicted number of arriving vehicles from the predicted number of departing vehicles
Disclosure by Schuijbroek
Schuijbroek, discloses
distributing a vehicle to a station having a larger predicted result value
see at least Pg. 5, Sec. 3 Service Level Requirements: observe the total net demand … at each station i ∈ S;
Rationale: Total net demand is a scalar value per station; stations with the larger net-demand figure (positive deficit) are those needing inventory, satisfying the ‘larger predicted result value’ concept)
of subtracting
see at least Pg. 5, Sec. 3: total net demand (pickups minus returns) during [the] observation period;
Rationale: The phrase “pickups minus returns” is an explicit subtraction operation, providing the required ‘of subtracting’ language)
the predicted number of arriving vehicles
see at least Pg. 5, Sec. 3: …minus returns;
Rationale: In bike-share parlance a return is an arriving bicycle; thus “returns” maps directly to the “predicted number of arriving vehicles)
from the predicted number of departing vehicles
see at least Pg. 5, Sec. 3: …(pickups minus returns);
Rationale: A pickup represents a bicycle departing the station; the subtraction “pickups minus returns” therefore matches “departing vehicles minus arriving vehicles” exactly).
Motivation to Combine Murakami, Karamanis, Ara, Crawford, Legros, and Schuijbroek
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford, Legros, and Schuijbroek before them, to define and apply the predicted result value used for selecting which station receives redistributed vehicles as a net-demand/net-imbalance quantity formed by subtracting predicted arrivals (returns) from predicted departures (pickups), and to prioritize distributing a vehicle to the station having the larger such predicted result value, because Schuijbroek expressly teaches using station-level net demand (pickups minus returns) as the operative deficit/surplus metric for service-level targeting, and a PHOSITA would straightforwardly substitute or incorporate this known net-demand metric into Murakami’s relocation ranking using the same predicted departures/arrivals already produced by the combined prediction components, yielding predictable prioritization of the stations most in need.
Claims 19 is rejected under 35 U.S.C. 103 as being unpatentable over Murakami, in view of Karamanis, in view of Ara, in view of Crawford, and in view of Lu et al., herein after will be referred to as Lu.
Regarding Claim 19,
The combination of Murakami, Karamanis, Ara, and Crawford establishes the mobility sharing system of Claim 18, which is the basis for Claim 19.
Disclosure by Karamanis
Karamanis discloses:
wherein the route setting unit sets a cost
see at least Eqns. 10 – 14: we define the cost functions for the edges in the graph as follows:…;
Rationale: Explicitly declares that the route algorithm assigns cost coefficients c_{ij} to every edge, exactly the act of “setting a cost.”)
between the first and second nodes to be proportional to the distance
see at least Eqn. 11; c_{ij}(x_{ij}) = r^{1}_{n(i)n(j)} CM x_{ij} ∀ (i,j) ∈ C;
Rationale: Because r^{1}_{n(i)n(j)} is “the average travel time between the clusters n(i) and n(j),” this cost term is directly proportional to the inter-station distance. Hence, cost scales directly with inter-station distance for every first-to-second-node edge),
and inversely proportional to at least one of the number of location vehicles of the stations corresponding to the first node in the target time interval
see at least Eqn. 2:
PNG
media_image1.png
49
325
media_image1.png
Greyscale
Rationale: From equation (2), a larger Itij (idle-vehicle count) increases the denominator and therefore reduces the term, giving the required inverse relationship.It also defines Itij(x) as “the average number of idle vehicles in cluster i,” so the inverse weighting depends on at least one station’s vehicle count. Because cluster i is the origin of edge (i,j), that vehicle count is tied directly to the first node. Finally, the phrase “on period t” indexes Itij to epoch t, confirming the count, and thus the inverse cost effect applies within the target time interval).
Claim Limitations Not Explicitly Disclosed by the Combination of Murakami, Karamanis, Ara, and Crawford
Murakami, Ara, and Karamanis do not explicitly disclose
and the number of location vehicles of the stations corresponding to the second nodes in the target time interval.
Disclosure by Lu
Lu discloses
and the number of location vehicles of the stations
See at least Pg. 3, Sec. 1.1: the number of vehicles available for use at that zone;
Rationale: Explicitly names the quantity of vehicles available at a zone; this exactly matches the “number of location vehicles of the stations” phrase, satisfying limitation
corresponding to the second nodes
See at least Pg. 9, Sec. 3.1: …being returned to zone j;
Rationale: Describes flow termination at destination “zone j”, which is the network’s C-layer or “second node”. Thus, the cited text ties the vehicle count to the correct node set
in the target time interval
see at least Pg. 9, Sec. 3.1: …and in period s;
Rationale: “Period s” is the arrival epoch immediately following routing, i.e., the target interval. Linking node j’s vehicle count to this period meets limitation’s timing constraint precisely).
Motivation to Combine Murakami, Karamanis, Ara, Crawford and Lu
Therefore, given the teachings as a whole, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having Murakami, Karamanis, Ara, Crawford and Lu before them, to incorporate Lu’s explicit modeling of the “number of vehicles available for use at [a] zone” in the destination zone/period into Karamanis’s MCMF relocation graph by parameterizing each destination-side station node (second node) with its corresponding vehicle-availability count for the target time interval, because Lu and Karamanis address the same shared-mobility redistribution setting, and using destination-interval vehicle availability as a node/edge capacity constraint is a conventional, predictable way to ensure Karamanis’s distance-weighted routing outputs remain feasible and operationally implementable in the mobility sharing system.
Response to Arguments
Rejection Under 35 U.S.C. 101
Applicant's arguments filed 10/01/2025 have been fully considered but they are not persuasive. For the reasons set forth in the 101 analysis of the amended claims in this Office Action, claims 1–19 are remain rejected under 35 U.S.C. 101 as being directed to an abstract idea (mathematical concepts and organizing human activity) without integration into a practical application and without significantly more.
Response to Applicant’s Arguments (35 U.S.C. § 103)
Applicant’s arguments have been fully considered but are not persuasive for the reasons set forth below. The rejection of the pending claims under 35 U.S.C. § 103 is maintained.
1) “One-way service” and “one-way reservation history with different departure and arrival stations”
Applicant asserts Murakami is “premised on round-trip operations,” allegedly lacking any “one-way service architecture,” and further lacking recognition of one-way imbalances (accumulation/depletion). The record does not support Applicant’s characterization.
(a) Murakami expressly uses different departure and destination stations (ports)
Murakami describes a user operation in which the user inputs a departure port and a destination port and then travels from the departure port to the destination port. This disclosure is inconsistent with Applicant’s position that Murakami is “premised on round-trip operations,” because Murakami explicitly contemplates trips between different stations/ports (origin ≠ destination).
(b) Murakami expressly recognizes inter-station imbalance and the resulting need to relocate vehicles
Murakami describes that, due to “arrival trips from other ports,” ports can experience vehicle shortage or excess vehicles, and that the system “set[s] relocation plan[s] of vehicles” to address such conditions. This directly contradicts Applicant’s assertion that Murakami “lacks any disclosure of the technological challenges unique to one-way services” and that Murakami “cannot and does not provide solutions for them.”
(c) Murakami uses OD-type historical/demand information, not merely “general demand”
Murakami further teaches modeling or extracting demand “between port k and l,” i.e., OD-specific quantities rather than undifferentiated demand.
Accordingly, the “one-way service” and “different departure and arrival stations” contentions do not overcome the rejection.
2) “Differential encoding with bit reduction” (and Applicant’s arguments as to Murakami, Ara, and Rodrigues)
Applicant argues that (i) Murakami and Ara keep arrivals/departures separate, (ii) Rodrigues’s “net flow” is merely visualization/analysis, and (iii) none teach the bit-reduction relationship.
(a) The art teaches using differences/deltas as the encoded value to reduce representation size
Crawford teaches delta/difference encoding and explicitly teaches reducing the number of bits used to represent values (e.g., by shifting/truncating a value so that “many bits can be truncated”). This disclosure directly addresses Applicant’s assertion that the cited art fails to teach embedding a subtraction result and reducing representational bits.
(b) The art teaches using net flow (arrivals − departures) as an operational variable (not mere “visualization”)
Applicant’s attempt to cabin “net flow” as mere visualization is not persuasive because the bike/vehicle sharing literature uses net flow/differences as inputs to allocation/optimization decisions. For example, Mahony discusses computing a station “net flow curve” and then using it within an optimization formulation to allocate resources.
(c) Rodrigues is properly applied for the “net flow = arrivals − departures” teaching
Applicant argues Rodrigues does not teach embedding a differential value or bit reduction. The rejection relies on Rodrigues for the net flow (arrivals minus departures) teaching and uses Crawford for the bit-reduction / reduced-bit representation teaching. As reflected in the record developed in the cited mapping, Rodrigues is cited for the express disclosure of “net flow: the difference between the number of arrivals and departures.”
Thus, even accepting arguendo Applicant’s position that Rodrigues alone does not discuss bits, the combination of Rodrigues (difference value / net flow) with Crawford (delta encoding and bit reduction) renders the “difference-and-bit” limitation obvious for the well-recognized reasons of reducing bandwidth/storage and normalizing features for downstream computation.
3) “Prioritization based on arrival-departure differential” (and Applicant’s argument as to Legros)
Applicant asserts Murakami uses unspecified “vehicle shortage” evaluation values and that Legros’s metric is allegedly λ+μ (sum-based “activity”), which Applicant contends is incompatible with a difference-based priority.
(a) Murakami’s shortage/excess logic inherently turns on predicted net change (arrivals vs. departures)
Murakami teaches determining/forecasting trip counts that distinguish occurrence (departures) and arrival counts. Murakami also explicitly describes shortage/excess conditions arising from vehicles moving between ports and the need to relocate vehicles accordingly. A PHOSITA would recognize that shortage/excess at a station over a time interval is driven by the difference between outgoing and incoming flows (i.e., the net change), and Murakami already supplies the needed forecast quantities.
(b) Legros is not limited to a “sum-based” activity metric; Legros explicitly uses imbalance terms tied to arrivals vs. departures
Applicant’s description of Legros as “sum-based” is incomplete. Legros expressly frames the operational difficulty as the “imbalance of bike arrivals and departures at a station.” Further, Legros includes a prioritization rule that directs prioritizing more “imbalanced” stations.
Most importantly, Legros’s policy formulation explicitly incorporates an imbalance term based on λ and μ and inventory deviation, rather than merely a λ+μ activity sum. In particular, Legros defines λ and μ as the arrival and departure rates and uses a factor based on their relationship as part of policy decisions. The policy table further reflects prioritization using an imbalance-sensitive structure (e.g., with a |λ−μ| term coupled with inventory deviation and an inverse time component).
Accordingly, Legros is properly applied to teach (and motivate) prioritization based on imbalance (difference-driven behavior), not merely high-traffic “sum” activity.
4) “Inverse proportional cost weighting in MCMF” (Karamanis)
Applicant argues Karamanis’s costs are merely distance/flow based and lack any inverse proportionality to station vehicle counts. The record does not support that contention.
Karamanis teaches that wait time is inversely proportional to the square root of the average number of idle vehicles, expressly: ( w \propto 1/\sqrt{\bar{n}} ). Karamanis further incorporates wait time into its generalized cost expression (with travel time terms), i.e., cost includes both travel-time/distance related components and the wait-time component that is inversely dependent on idle vehicles.
Thus, Karamanis teaches a cost structure that is proportional to travel time/distance while also being inversely related to vehicle availability (via the wait-time component). Applicant’s characterization of Karamanis as “uniform cost redistribution without demand prediction” is therefore not persuasive.
5) “Clustering based on unsupervised learning for one-way services” (Mahony)
Applicant argues Mahony is not directed to “one-way vehicle services” and instead clusters only “rush hour temporal patterns.” This argument is not persuasive because the claim requires only that a clustering model based on unsupervised learning be used to divide an area into unit areas based on historical usage data; the clustering technique is not limited to any one vehicle modality.
Mahony teaches clustering stations based on observed usage, identifying “groups of stations with similar behavior during rush-hours,” and further explains using observed trip data to define “stations with similar behavior.” A PHOSITA would have been motivated to apply such unsupervised clustering to the historical reservation/OD datasets already present in the car/vehicle sharing context (e.g., Murakami’s departure/destination port data), and such an application is a predictable use of a known technique to organize a service region for downstream prediction and relocation planning.
6) “Training with labeled correct answer data” (Ara)
Applicant asserts Ara does not disclose the claimed labeled dataset structure. The record indicates otherwise as to supervised learning with targets (correct answers).
Ara describes forming a time-indexed training set over prior periods and training to predict a future interval, i.e., using historical nodes and predicting the “next hour,” which necessarily supplies supervised targets/correct answers for training. Additionally, Ara is expressly concerned with OD-type demand (pickup-destination pair). A PHOSITA would recognize that such OD demand data can be aggregated into per-station departing and arriving counts for each target interval, which is the exact form of “correct answer data” recited.
7) “The References Cannot Be Combined / Disparate principles / no motivation” (including Applicant’s “Murakami is balanced round-trip” premise)
Applicant’s “cannot be combined” argument is not persuasive because it is premised on an incorrect factual characterization of Murakami and an unduly rigid view of obviousness.
(a) The references address the same operational problem space—imbalance and allocation in shared mobility
Murakami recognizes inter-port movements creating shortages/excess and relocation planning to correct it. Legros similarly frames the central operational issue as arrival/departure imbalance and teaches prioritizing imbalanced stations. Karamanis provides a minimum-cost-flow framework with cost components that include inverse dependence on available idle vehicles. Ara supplies supervised predictive modeling of OD demand. Mahony supplies unsupervised clustering of stations based on observed usage. Crawford supplies delta/difference encoding and explicit bit-reduction techniques.
These teachings are complementary (forecast → cluster → prioritize → optimize routes/costs → compress/represent features) and are directed to the predictable and well-understood engineering objective of improving operational performance and computational efficiency in a mobility sharing platform.
(b) KSR / PHOSITA rationale
A PHOSITA tasked with improving a mobility sharing system facing imbalance would have been motivated to incorporate:
prediction (Ara) to forecast departures/arrivals,
imbalance-based prioritization (Legros) and net flow features (Rodrigues/Mahony),
cost-aware flow optimization (Karamanis) for routing/redistribution, and
difference/bit-efficient encoding (Crawford) to reduce storage/bandwidth and improve computation.
This is not an “improper dissection,” but rather a conventional obviousness analysis in which known components solving known subproblems are combined in a predictable way to yield the claimed integrated system.
For at least the reasons above, Applicant’s arguments as to Murakami, Ara, Rodrigues, Legros, Karamanis, and Mahony do not establish error in the rejection. The §103 rejection is therefore maintained, and the application is not in condition for allowance.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLUWABUSAYO ADEBANJO AWORUNSE whose telephone number is (571)272-4311. The examiner can normally be reached M - F (8:30AM - 5PM).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jelani Smith can be reached at (571) 270-3969. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/OLUWABUSAYO ADEBANJO AWORUNSE/Examiner, Art Unit 3662
/JELANI A SMITH/Supervisory Patent Examiner, Art Unit 3662