DETAILED ACTION
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 .
This action is responsive to the claims dated 3/7/2024.
Claims 1-19 are presented for examination.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been received for the foreign priority application no. CN 202310493262.5 filed 4/28/2023.
Drawings
The drawings are objected to as failing to comply with 37 CFR 1.84(p)(5) because they include the following reference character(s) not mentioned in the description: reference character "S110" (FIG. 1), which designates the step of obtaining a parking space vacancy rate of each parking lot in the area to be predicted at the moment to be predicted by inputting the fusion feature into the result prediction network. The specification designates the preceding steps of FIG. 1 as S 100 (paragraph [0068]), S 102 (paragraph [0069]), S104 (paragraph [0076]), S 106 (paragraph [0078]) and S 108 (paragraph [0083]), but nowhere recites S110. Corrected drawing sheets in compliance with 37 CFR 1.121(d), or amendment to the specification to add the reference character(s) in the description in compliance with 37 CFR 1.121(b), are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either "Replacement Sheet" or "New Sheet" pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Specification
The disclosure is objected to because of the following informalities:
(a) [0078] begins "S106: a spatial relationship diagram", omitting the introductory phrase "At step" used at [0068], [0069], [0076] and [0083];
(b) [0024], [0049] and [0134] each recite that the parking space vacancy rate prediction model "further includes an environmental feature extraction;", apparently omitting the word "network" that appears in the corresponding recitations elsewhere in the disclosure and in claims 5 and 18;
(c) [0085] recites "a graph convolutional network (GCN)" while the remainder of the disclosure and the claims recite "graph convolution network", see the requirement of 37 CFR 1.71(a) for full, clear, concise, and exact terms; and
(d) [0144] recites "a Silicone Labs C8051F320", which appears to be a misspelling of "Silicon Labs".
Appropriate correction is required.
Claim Objections
Claim 14 is objected to because of the following informalities:
In claim 14, the recitation "wherein the electronic further deploys a pre-trained parking space vacancy rate prediction model" should read "wherein the electronic device further deploys a pre-trained parking space vacancy rate prediction model".
Appropriate correction is required.
Claim Rejections - 35 U.S.C. 112(b)
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 6, 13 and 19 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 6 recites "acquiring parking space vacancy rates of each of parking lots in a designated area for at least two historical moments; taking a parking space vacancy rate of each of the parking lots at a latest historical moment among the at least two historical moments as a label, and taking a parking space vacancy rate of each of the parking lots at other historical moments as a sample", and claim 19 recites the corresponding limitation "acquire parking space vacancy rates of each of parking lots in a designated area for at least two historical moments; take a parking space vacancy rate of each of the parking lots at a latest historical moment among the at least two historical moments as a label, and take a parking space vacancy rate of each of the parking lots at other historical moments as a sample". Claims 6 and 19 recite the limitation "the parking lots" in the label limitation, the sample limitation, the sample-feature limitation and the training limitation. There is insufficient antecedent basis for this limitation in the claims. See MPEP 2173.05(e).
Two different groups of parking lots are recited by the time these limitations are reached. Claim 1, from which claim 6 depends, and claim 14, from which claim 19 depends, recite parking lots in the area to be predicted and thereafter refer to that group as "the parking lots." Claims 6 and 19 then introduce a second and different group, namely the parking lots in "a designated area," which the specification describes as the lots whose historical data are used to train the model rather than the lots whose future vacancy rates are predicted. Every later recitation of "the parking lots" and "each of the parking lots" in claims 6 and 19 is therefore uncertain as to which of the two groups was intended. MPEP 2173.05(e) states that if two different levers are recited earlier in the claim, the recitation of "said lever" in the same or a subsequent claim would be unclear where it is uncertain which of the two levers was intended. The uncertainty here is compounded because claims 6 and 19 expressly qualify one recitation, "constructing a spatial relationship diagram among each of the parking lots in the designated area," while leaving the remaining recitations unqualified, which suggests that the unqualified recitations may have been intended to refer to the parking lots of the area to be predicted.
For purposes of examination, each recitation of "the parking lots" and "each of the parking lots" appearing in claims 6 and 19 after the introduction of the designated area is interpreted under the broadest reasonable interpretation as referring to the parking lots in the designated area, that is, to the training lots. That interpretation is grounded in the specification at [0113], which describes the computing device acquiring parking space vacancy rates of each of the parking lots in a designated area for at least two historical moments and taking the rate at the latest historical moment as a label and the rates at the other historical moments as a sample, and at [0031].
Claim 6 recites "obtaining a sample feature by inputting the sample into the feature extraction network, wherein the sample feature is used to characterize a time-dependent relationship in the historical vacancy rate of each of the parking lots", and claim 19 recites the corresponding limitation "obtain a sample feature by inputting the sample into the feature extraction network, wherein the sample feature is used to characterize a time-dependent relationship in the historical vacancy rate of each of the parking lots". Claims 6 and 19 recite the limitation "the historical vacancy rate" in the sample-feature limitation. There is insufficient antecedent basis for this limitation in the claims. See MPEP 2173.05(e).
Neither claim 6 nor claim 19 recites any historical vacancy rate, in the singular or the plural, for the parking lots in the designated area. What those claims recite for the designated area is "parking space vacancy rates ... for at least two historical moments," a "label" and a "sample." The only earlier recitation of the term is in claim 1 and in claim 14, where "historical vacancy rates" is expressly defined as the parking space vacancy rates of each of the parking lots in the area to be predicted at a plurality of moments before the moment to be predicted. Those rates are not the training data operated upon by claims 6 and 19, and that earlier term is plural whereas claims 6 and 19 recite the singular "the historical vacancy rate." A person having ordinary skill in the art therefore cannot determine with reasonable certainty whether the recited sample feature must characterize a time-dependent relationship in the training data of the designated area or in the data of the area to be predicted recited in the parent claim, and the scope of the sample-feature limitation is accordingly uncertain. See Nautilus, 572 U.S. at 901; In re Packard, 751 F.3d at 1314.
For purposes of examination, "the historical vacancy rate of each of the parking lots" in claims 6 and 19 is interpreted under the broadest reasonable interpretation as the parking space vacancy rates of each of the parking lots in the designated area at the historical moments other than the latest historical moment, that is, as the recited sample. That interpretation is grounded in the specification at [0113] and [0031].
Claim 13 recites "a non-transitory computer-readable storage medium, wherein the non-transitory computer storage medium stores a computer program which, when executed by a processor, the processor is configure to: determine an area to be predicted and a moment to be predicted". The recited limitation is indefinite. The relative pronoun "which," whose referent is the recited computer program, is never supplied with a predicate. The recitation instead breaks off after the conditional clause "when executed by a processor" and resumes with a different grammatical subject, "the processor," to which the verb phrase "is configure to" is attached. Because the claim never states what the computer program itself does, the claim is open to at least two materially different constructions and supplies no basis for choosing between them.
Under the first construction the claim is an article of manufacture only: the recited acts are the acts that the stored program causes a processor to carry out when the program is executed, and the processor is recited only as the environment in which the program is executed. Under the second construction the phrase "the processor is configure[d] to" is an affirmative limitation directed to a processor, so that the claimed medium must be combined with a processor that has in fact been configured to perform the recited acts. The two constructions do not have the same scope and are not infringed by the same act: under the first, infringement is complete when the medium bearing the program is made or sold, whereas under the second, infringement further requires that a processor be configured with and operated under the program. A person having ordinary skill in the art is therefore not informed with reasonable certainty of the scope of claim 13. See Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. 898, 901, 910 (2014); In re Packard, 751 F.3d 1307, 1314 (Fed. Cir. 2014); MPEP 2173.02.
The second construction additionally recites subject matter of more than one statutory class in a single claim. Claim 13 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for improperly mixing method steps and apparatus elements. It is unclear whether infringement of the claim occurs when the apparatus is made or when the method is performed. See IPXL Holdings, L.L.C. v. Amazon.com, Inc., 430 F.3d 1377, 1384 (Fed. Cir. 2005); MPEP 2173.05(p).
For purposes of examination, claim 13 is interpreted under the broadest reasonable interpretation as a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to carry out the recited acts, the processor not being a positively recited element of the claimed medium. That interpretation is grounded in the specification at [0140] which describes a computer-readable storage medium storing a computer program that can be used to execute the disclosed method of predicting the parking space vacancy rate, and it corresponds to the method of claim 1. The specification informs the interpretation of the claim but its limitations are not read into the claim. See MPEP 2111.
Claim Rejections - 35 U.S.C. 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.
Claims 1, 3, 6, 13-14, 16, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Feng et al. (hereinafter Feng) “Predicting vacant parking space availability zone-wisely: a graph based spatio-temporal prediction approach” 2022 in view of Liu et al. (hereinafter Liu), US 2021/0232588 A1.
Liu was disclosed in an IDS dated 3/7/2024.
Regarding independent claim 1, Feng teaches a method of predicting a parking space vacancy rate (Feng: page 1, Abstract, "the prediction model is applied to predict the number VPSs of 8 public parking lots in Santa Monica"; the number of VPSs (a parking space vacancy rate) is the count of unoccupied spaces reported for each parking lot, and Feng further scales that quantity to a normalized fractional value, page 2, Section II.A, "Use the Min-Max normalization method to scale all the data to the range [0,1]", so the quantity Feng predicts is a measure of the vacancy of each parking lot), and the parking space vacancy rate prediction model comprising a feature extraction network, a graph fusion network and a result prediction network, the method comprising (Feng: page 1, Abstract, "a graph data-based model ST-GBGRU (Spatial-Temporal Graph Based Gated Recurrent Unit)…the temporal correlation of historical VPS data is extracted by GRU, on the other hand, the spatial correlation of historical VPS data is extracted by GCN inside GRU" and Section I. Introduction; the ST-GBGRU model (the parking space vacancy rate prediction model) is composed of the Gated Recurrent Unit GRU network (a feature extraction network) and the Graph Convolutional Network GCN network (a graph fusion network), and Feng carries the model output forward to a prediction stage, page 4, Section II.C, "in single-step direct prediction, m historical observation data are directly used to predict", which is the result prediction network): determining an area to be predicted and a moment to be predicted (Feng: page 2, Section II.A, "The dataset come from the real parking data of 8 parking lots (St1-St8) in Santa Monica, California, USA"; the set of eight parking lots in Santa Monica (an area to be predicted) is fixed before prediction, and Feng selects the instant for which the output is generated, page 1, Abstract, "the number of VPSs can be predicted both in short-term (i.e., within 30 min) and in long-term (i.e., over 30min)"; the short-term or long-term horizon so chosen is the future instant (a moment to be predicted) at which the prediction is issued); acquiring parking space vacancy rates of each of parking lots in the area to be predicted at a plurality of moments before the moment to be predicted as historical vacancy rates of each of the parking lots (Feng: page 2, Section II.A, "The data collection frequency of each parking lot is 5 minutes, and each parking lot has 9108 recorded data"; each parking lot supplies a series of the number of VPSs (parking space vacancy rates) sampled at successive earlier instants, and that series is the model input, page 4, Section II.C, "in single-step direct prediction, m historical observation data are directly used to predict"; the m earlier observations (historical vacancy rates) all precede the instant being predicted);
obtaining a first feature by inputting the historical vacancy rates of each of the parking lots into the feature extraction network (Feng: page 3, Section II.B.3, "The role of the GCN network is to extract the spatial characteristics of VPS, and the role of the GRU network is to extract the temporal characteristics of VPS"; the historical VPS sequence (the historical vacancy rates) is supplied to the GRU network (the feature extraction network), which returns the temporal characteristics of VPS (a first feature)),
wherein the first feature is used to characterize a time-dependent relationship in the historical vacancy rates of each of the parking lots (Feng: page 1, Abstract, "the temporal correlation of historical VPS data is extracted by GRU"; the temporal characteristics of VPS (the first feature) express the correlation of each lot's values across successive sampling instants, which is a relationship that depends on time);
constructing a spatial relationship diagram among each of the parking lots in the area to be predicted (Feng: page 2, Section II.A, "The graph of the parking lot is built according to the geographical distance between parking lots. We describe the spatial dependence between parking lots by constructing an adjacency matrix"; the adjacency matrix (a spatial relationship diagram) is assembled over the parking lots of the study area and its entries record the spatial dependence among those lots);
obtaining a fusion feature by inputting the spatial relationship diagram and the first feature into the graph fusion network (Feng: page 3, Section II.B.3, "The model input consists of two parts, a graph of parking lot and a time series of VPS"; the graph of parking lot (the spatial relationship diagram) enters the model alongside the series data, and the GCN network (the graph fusion network) is applied to the temporal representation as well as to the raw input, page 3, Section II.B.3, "We merge GCN and GRU models and use GCN to extract spatial features from input and hidden layers"; the hidden layer carries the temporal characteristics of VPS (the first feature), so the product of that graph convolution is a representation combining the graph with the temporal representation (a fusion feature), page 3, Section II.B.3, "The model integrates GCN and GRU, which can fully extract the spatial features of the hidden layer output and achieve good prediction performance"); and obtaining a parking space vacancy rate of each of the parking lots in the area to be predicted at the moment to be predicted by inputting the fusion feature into the result prediction network (Feng: page 4, Section II.C, "in single-step direct prediction, m historical observation data are directly used to predict"; the merged spatial and temporal representation (the fusion feature) is carried into the prediction stage (the result prediction network), which issues a value for every lot of the study area, page 1, Abstract, "the prediction model is applied to predict the number VPSs of 8 public parking lots in Santa Monica"; the number of VPSs (a parking space vacancy rate) is returned for each of the eight lots at the selected horizon).
Feng does not expressly teach being performed by a computing device, a pre-trained parking space vacancy rate prediction model being deployed in the computing device.
However, Liu teaches being performed by a computing device, a pre-trained parking space vacancy rate prediction model being deployed in the computing device (Liu: [0177], "As shown in FIG. 5, the electronic device comprises: one or more processors 501, a memory 502, and interfaces connected to components"; the electronic device (a computing device) supplies the processor and memory on which the prediction method runs, [0177], "The processor can process instructions for execution within the electronic device, including instructions stored in the memory"; the instructions so executed are those of the prediction method, and the model they run is trained in advance, [0098], "the local space correlation and global space correlation of the parking lots may be modeled using a graph attention neural network model and a hierarchical graph neural network model to obtain a final space correlation representation of the parking lots, and the final space correlation representation is input to a gated recurrent neural network model to predict future free parking space information of the parking lots"; the parameters of those network models are fixed before use, [0045], "when performing model training, selecting N parking lots with real-time sensors as sample parking lots, building annotation data based on historical free parking space information of the sample parking lots, performing training optimization based on the annotation data"; a model whose parameters were fixed by that prior training and which then resides on the electronic device is a pre-trained model deployed in the computing device as the embodiment).
Because Feng and Liu are analogous art and within the same field of endeavor, specifically forecasting the availability of parking spaces at a plurality of parking lots by applying neural network models to a graph of those parking lots together with their historical availability series, they address the same problem solving area of supplying drivers with a reliable forecast of where parking will be available, accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention, to deploy the trained prediction model of Feng on the processor and memory of the electronic device taught by Liu, with a reasonable expectation of success, such that the forecast is computed by a machine that can serve it to drivers, to teach being performed by a computing device, a pre-trained parking space vacancy rate prediction model being deployed in the computing device. This modification would have been motivated by the desire to obtain a current forecast for lots that carry no sensors of their own, because instrumenting every lot is impractical (Liu: [0096], "since real-time sensors are costly, they are mounted in only very few parking lots to monitor in real time the current free parking space information which usually refers to the number of free parking spaces. Hence, it is very necessary to predict free parking space information of parking lots"; running the trained model on an electronic device is what makes the prediction available where a sensor is absent). See KSR International Co. v. Teleflex Inc., 550 U.S. 398 (2007); the combination is no more than the use of a known technique to improve a similar device in the same way, MPEP 2143.01.
Regarding dependent claim 3, Feng, in view of Liu, teach the method of claim 1, wherein constructing the spatial relationship diagram among each of the parking lots in the area to be predicted comprises: constructing the spatial relationship diagram by taking each of the parking lots in the area to be predicted as a node (Feng: page 3, Section II.B.1, "GCN captures the spatial dependencies of a graph through messages passing between graph nodes"; the graph is assembled over the parking lots of the study area so that each lot occupies one of the graph nodes (a node)), and a distance between each of the parking lots in the area to be predicted as a weight of an edge (Feng: page 2, Section II.A, "According to the formula (2), the adjacency matrix is calculated using geographic distance"; the entry that couples any two lots in the adjacency matrix is computed from the geographic distance (a weight of an edge) separating them, and Feng derives that distance from the lots' coordinates, page 2, Section II.A, "Using the latitude and longitude of each parking lot, calculate the distance between all parking lots").
Regarding dependent claim 6, Feng, in view of Liu, teach the method of claim 1, wherein the parking space vacancy rate prediction model is trained by: acquiring parking space vacancy rates of each of parking lots in a designated area for at least two historical moments (Feng: page 2, Section II.A, "The dataset come from the real parking data of 8 parking lots (St1-St8) in Santa Monica, California, USA" and "The data collection frequency of each parking lot is 5 minutes, and each parking lot has 9108 recorded data"; the recorded numbers of VPSs (parking space vacancy rates) of the eight lots of Santa Monica (a designated area) are collected at successive earlier instants (at least two historical moments)), taking a parking space vacancy rate of each of the parking lots in the designated area (interpreted per the 35 U.S.C. 112(b) rejection above) at a latest historical moment among the at least two historical moments as a label, and taking a parking space vacancy rate of each of the parking lots in the designated area (interpreted per the 35 U.S.C. 112(b) rejection above) at other historical moments as a sample (Feng: page 2, Section II.A, "Use the sliding window method to intercept the dataset, and reshape the data into samples of a specified length for training, where the training data is in the form of [X1, X2, X3, ..., Xn], and X represents the input, n represents the length of the input sequence"; each window of earlier observations is a training input (a sample), and the observation that follows the window is the value the model is trained to reproduce, page 4, Section II.C, "m historical observation data are directly used to predict" the number of VPSs at the later instant, and page 4, Section II.D.1, "are the actual number of VPSs and predicted numbers of VPSs of i-th parking lot at time t"; the actual number of VPSs at that later instant (a label) is what the predicted number is scored against), obtaining a sample feature by inputting the sample into the feature extraction network, wherein the sample feature is used to characterize a time-dependent relationship in[[ the historical vacancy rate of each of the parking lots]] the sample (interpreted per the 35 U.S.C. 112(b) rejection above) (Feng: page 1, Abstract, "the temporal correlation of historical VPS data is extracted by GRU"; during training the GRU network (the feature extraction network) consumes each window (the sample) and returns its temporal characteristics (a sample feature)), constructing a spatial relationship diagram among each of the parking lots in the designated area as a sample spatial relationship diagram (Feng: page 2, Section II.A, "The graph of the parking lot is built according to the geographical distance between parking lots"; the adjacency matrix over the eight lots (a sample spatial relationship diagram) is built as a preprocessing step of the training data), obtaining a sample fusion feature by inputting the sample spatial relationship diagram and the sample feature into the graph fusion network (Feng: page 3, Section II.B.3, "use GCN to extract spatial features from input and hidden layers"; the GCN network (the graph fusion network) is applied to the graph and to the hidden-layer representation of the window during training exactly as in prediction, returning the graph-convolved representation (a sample fusion feature)), obtaining a prediction result of the parking space vacancy rate of each of the parking lots in the designated area at the latest historical moment by inputting the sample fusion feature into the result prediction network (Feng: page 4, Section II.D.1, "MAE measures the average absolute error between the actual values and predicted value"; the predicted number of VPSs (a prediction result) is issued for every lot at the later instant of each window and compared with the actual value), and training the parking space vacancy rate prediction model with an objective of minimizing a difference between the prediction result of the parking space vacancy rate of each of the parking lots in the designated area (interpreted per the 35 U.S.C. 112(b) rejection above) and the label (Feng: page 6, Section II.D.2, "The ratio of training set and test set is 4:1" and "The above models are trianed using the Adam optimizer" (sic); the Adam optimizer adjusts the model parameters over the training set against an objective, and Liu: page 3, [0045], "when performing model training, selecting N parking lots with real-time sensors as sample parking lots, building annotation data based on historical free parking space information of the sample parking lots, performing training optimization based on the annotation data, and minimizing an objective function"; the objective function scores the predicted free parking space information against the annotation data (the label) and the optimization drives that difference down. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to train the model of Feng by minimizing an objective function against annotation data built from the later observations as Liu teaches, with a reasonable expectation of success, because the Adam optimizer that Feng applies requires such an objective and the objective of Liu is directed to the same quantity, the free parking space information of the lots; the motivation to combine Feng and Liu is otherwise the same as that set forth above for claim 1).
Regarding independent claim 13, it is a non-transitory computer-readable storage medium claim that is substantially the same as the method of claim 1. Thus, claim 13 is rejected for the same reason as claim 1. In addition, Liu teaches a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores a computer program which, when executed by a processor,[[the processor is configure to]] causes the processor to (as interpreted by the 35 U.S.C. 112(b) rejection above) perform the recited operations (Liu: [0080], "A non-transitory computer-readable storage medium storing computer instructions therein, wherein the computer instructions are used to cause the computer to" perform the disclosed method; the non-transitory computer-readable storage medium (a non-transitory computer-readable storage medium) holds the computer instructions (a computer program) that drive at least one processor (a processor), [0178], "The memory 502 is a non-transitory computer readable storage medium provided by the present disclosure. Wherein, the memory stores instructions executable by at least one processor, so that the at least one processor executes the method provided by the present disclosure").
Regarding independent claim 14, it is an electronic device claim that is substantially the same as the method of claim 1. Thus, claim 14 is rejected for the same reason as claim 1. In addition, Liu teaches an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the electronic device (interpreted per the Claim Objections above) further deploys a pre-trained parking space vacancy rate prediction model (Liu: [0177], "As shown in FIG. 5, the electronic device comprises: one or more processors 501, a memory 502, and interfaces connected to components"; the electronic device (an electronic device) holds the memory 502 (a memory) and one or more processors 501 (a processor), and the memory carries the executable instructions, [0177], "The processor can process instructions for execution within the electronic device, including instructions stored in the memory"; those stored instructions are the computer program (a computer program), and it is that device on which the trained network models run, [0098], "the final space correlation representation is input to a gated recurrent neural network model to predict future free parking space information of the parking lots").
Regarding dependent claims 16 and 19, these are electronic device claims that are substantially the same as the method of claims 3 and 6, respectively. Thus, claims 16 and 19 are rejected for the same reasons as claims 3 and 6.
Claims 2, 4, 15 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Feng in view of Liu, as applied in the rejections of claims 1 and 14 above, and further in view of Li et al. (hereinafter Li) “Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting” 2021.
Regarding dependent claim 2, Feng, in view of Liu, teach all the elements of claim 1.
Feng and Liu do not expressly teach wherein before inputting the spatial relationship diagram and the first feature into the graph fusion network, the method further comprises: obtaining a second feature by inputting the historical vacancy rates of each of the parking lots into the feature extraction network, wherein the second feature is used to characterize similarities of the historical vacancy rates of each of the parking lots; and wherein inputting the spatial relationship diagram and the first feature into the graph fusion network further comprises: inputting the spatial relationship diagram, the first feature and the second feature into the graph fusion network.
However, Li teaches wherein before inputting the spatial relationship diagram and the first feature into the graph fusion network, the method further comprises: obtaining a second feature by inputting historical into the feature extraction network (Li: page 4190, Algorithm 1, "Temporal Graph Generation Input: N time series from V(|V| = N)", "disti,j = TDL(Vi, Vj) (Alg. 2)" and "return Weighted Matrix W of Temporal Graph G"; the N time series, one per node, are the quantity fed to the temporal-graph generation stage, and the stage returns the weighted matrix W (a second feature). That stage is a feature extraction network within the meaning of the claim: the instant specification states that the feature extraction network may include a first feature extraction network and a second feature extraction network, and that the second feature extraction network may be a dynamic time warping (DTW) (specification, [0093]-[0094]), and the stage of Li is that dynamic time warping computation, page 4191, "Dynamic Time Warping is a typical algorithm to measure similarity of time series", page 4192, Section Spatial-Temporal Fusion Graph Construction, "Temporal Graph ATG generated by Alg. 1"; second feature W being assembled into the fusion graph that the graph module then consumes), wherein the second feature is used to characterize similarities of the historical (Li: page 4190, Introduction, "we propose a novel data-driven method for graph construction: the temporal graph learned based on similarities between time series"; each entry of W records whether the series of node i and node j are among each other's most similar under the dynamic time warping distance, page 4191, Section Similarity of Temporal Sequences, "is the final distance between X and Y with the best alignment which can represent the similarity between two time series"; the series are historical, so W (the second feature) characterizes their similarities); and wherein inputting the spatial relationship diagram and the first feature into the graph fusion network further comprises: inputting the spatial relationship diagram, the first feature and the second feature into the graph fusion network (Li: page 4192, Section Spatial-Temporal Fusion Graph Construction, "It consists of three kinds N × N matrix: Spatial Graph ASG which is given by dataset, Temporal Graph ATG generated by Alg. 1, and Temporal Connectivity graph ATC whose element is nonzero iif previous and next time steps is the same node."; the spatial graph (the spatial relationship diagram) and the temporal graph (the second feature) are supplied together, as one fusion graph, to the graph module).
Because Feng, in view of Liu, and Li are analogous art and within the same field of endeavor, specifically forecasting a quantity at each node of a graph of locations from the historical series observed at those locations by applying a graph convolution network together with a recurrent network (Feng builds its model on the GCN+GRU traffic-flow model of its reference [37], page 3, Section II.B.3, "Reference [37] has proposed a model of stacking GCN+GRU"), they address the same problem solving area of capturing correlations between locations that a geographic graph alone does not reflect, accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention, to add the temporal-graph generation stage of Li to the model of Feng as deployed on the electronic device of Liu, computing the dynamic time warping similarity matrix over the historical VPS series of the parking lots and supplying it to the GCN of Feng together with the distance-based adjacency matrix and the GRU hidden state, with a reasonable expectation of success, such that lots whose vacancy series move alike are connected in the graph even when they are far apart, to teach wherein before inputting the spatial relationship diagram and the first feature into the graph fusion network, the method further comprises: obtaining a second feature by inputting the historical vacancy rates of each of the parking lots into the feature extraction network, wherein the second feature is used to characterize similarities of the historical vacancy rates of each of the parking lots; and wherein inputting the spatial relationship diagram and the first feature into the graph fusion network further comprises: inputting the spatial relationship diagram, the first feature and the second feature into the graph fusion network. This modification would have been motivated by the desire to capture the correlation between distant lots that the distance-thresholded adjacency matrix of Feng (page 2, Section II.A, "the adjacency matrix is calculated using geographic distance") cannot express: Liu recognizes that in this field "two working regions which are far away from each other might simultaneously be in a state in high demand for parking lots upon working time" (Liu: [0097]), and Li supplies the known technique for exactly that problem, page 4189, Abstract, "a data-driven method of generating 'temporal graph' is proposed to compensate several existing correlations that spatial graph may not reflect", page 4189, Introduction, "those distant nodes may have certain correlations, i.e., they would share similar 'temporal pattern'". The expectation of success is reasonable because the temporal graph is an N × N matrix of the same form as the adjacency matrix the GCN of Feng already consumes, and Li reports that the model so built "consistently outperforms baselines" (page 4190). See KSR International Co. v. Teleflex Inc., 550 U.S. 398 (2007); the combination is no more than the use of a known technique to improve a similar device in the same way, MPEP 2143.01.
Regarding dependent claim 4, Feng, in view of Liu and further in view of Li, teach the method of claim 2, wherein the graph fusion network comprises: an attention network (Liu: [0098], "the local space correlation and global space correlation of the parking lots may be modeled using a graph attention neural network model"; the graph attention neural network model (an attention network) is the stage that operates on the parking lot association graph), and a graph convolution network (Feng: page 3, Section II.B.1, "GCN captures the spatial dependencies of a graph through messages passing between graph nodes."; the GCN network (a graph convolution network) performs the graph-side computation of the model), and inputting the spatial relationship diagram, the first feature and the second feature into the graph fusion network comprises: obtaining an output result weighted by attention by inputting the spatial relationship diagram, the first feature and the second feature into the graph fusion network (Liu: [0056], "Attention represents a graph attention mechanism"; the edge weights applied to the parking lot association graph are produced by that attention mechanism, and the graph-side stage then combines the lots under those weights, [0054], "aggregates the environment context features of the neighboring parking lots according to the weights of edges between the neighboring parking lots and the parking lot i to obtain a representation vector of the parking lot i, and regards the representation vector as the local space correlation information of the parking lot i at the current time"; because the aggregation is carried out under attention-derived edge weights, the representation vector so produced is an output result weighted by attention; in the combination, the per-lot features that the graph attention mechanism of Liu weights are the hidden-layer output of the GRU of Feng (the first feature) and the row of the temporal graph W of Li for that lot (the second feature), taken over the adjacency matrix of Feng (the spatial relationship diagram); the attention mechanism of Liu derives its edge weights from whatever per-lot feature is supplied to it, [0056], "represents the environment context feature of the parking lot i at the current time", and substituting the temporal and similarity features of the combination for the context feature of Liu is the substitution of one known per-lot input for another, yielding the predictable result of attention weights responsive to those features, MPEP 2143(I)(B)), and inputting the output result and the historical vacancy rates of each of the parking lots into the graph convolution network (Feng: page 3, Section II.B.3, "We merge GCN and GRU models and use GCN to extract spatial features from input and hidden layers."; the GCN network (the graph convolution network) receives both the hidden-layer representation and the raw input series, that is, the historical VPS sequence (the historical vacancy rates); in the combination, the representation vector produced by the attention stage of Liu takes the place of the hidden-layer output that the GCN of Feng already convolves together with the input series: Feng teaches that the hidden-layer representation must itself be spatially extracted by the GCN, page 3, Section II.B.3, "This method ignores the spatial feature extraction of the output of the hidden layer, however the output of the hidden layer has an influence on the accuracy of the result", and Liu carries its attention-stage output onward as the per-lot feature that its downstream stages consume (Liu: [0129]-[0130], the local space correlation information is concatenated with the global space correlation information and the concatenation result is supplied to the gated recurrent stage), so that the GCN of Feng receives the attention-weighted output result in place of the hidden layer, and the historical VPS series (the historical vacancy rates) as its input).
Regarding dependent claims 15 and 17, these are electronic device claims that are substantially the same as the method of claims 2 and 5, respectively. Thus, claims 15 and 17 are rejected for the same reasons as claims 2 and 5.
Claims 5 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Feng in view of Liu, as applied in the rejections of claims 1 and 14, respectively, above, and further in view of Zeng et al. (hereinafter Zeng) “Parking Occupancy Prediction Method Based on Multi Factors and Stacked GRU-LSTM” 2022.
Regarding dependent claim 5, Feng, in view of Liu, teach the method according to claim 1, wherein the parking space vacancy rate prediction model further comprises inputting the first feature and the fusion feature into the result prediction network (Feng: page 3, Section II.B.3, "We merge GCN and GRU models and use GCN to extract spatial features from input and hidden layers", page 5, Fig. 5 (Overview of ST-GBGRU model); in which the prediction stage is fed by the output of the last feature-extraction cell, so the prediction stage receives the first feature and the fusion feature as well wherein the prediction stage of Feng receives the hidden-layer output of the final GRU cell, which carries the temporal characteristics of VPS (the first feature) forward together with the graph-convolved representation (the fusion feature)).
Feng and Liu do not expressly teach an environmental feature extraction network; wherein before inputting the fusion feature into the result prediction network, the method further comprises: acquiring environmental information of each of the parking lots in the area to be predicted before the moment to be predicted, wherein the environmental information at least comprises weather information and holiday information; and obtaining an environmental feature by inputting the environmental information into the environmental feature extraction network; and wherein inputting the fusion feature into the result prediction network further comprises: inputting the first feature, the fusion feature, the environmental feature and the historical vacancy rates of each of the parking lots into the result prediction network.
However, Zeng teaches an environmental feature extraction network (Zeng: page 47365, Section III.E, "First, a GRU layer that trains data with simple network structure, fewer parameters and easier convergence is adopted."; the GRU layer (an environmental feature extraction network) is the stage that consumes the exogenous inputs), wherein before inputting into the result prediction network, the method further comprises: acquiring environmental information of each of the parking lots in the area to be predicted before the moment to be predicted, wherein the environmental information at least comprises weather information and holiday information (Zeng: page 47361, Abstract, "uses multi factors, including occupancy, weather conditions and holiday, as input to predict parking availability"; the multi factors gathered for each parking lot (acquiring environmental information of each of the parking lots in the area) over the earlier period include weather conditions (weather information) and holiday (holiday information), and Zeng reduces both to model-ready values, page 47367, Section IV.A, "the weather and holiday data quantification tables are shown in Tables 2 and 3"), and obtaining an environmental feature by inputting the environmental information into the environmental feature extraction network (Zeng: page 47365, Section III.E, "The combined prediction model takes multi factors as input, such as occupancy, weather conditions, vacations, etc."; the multi factors (the environmental information) are fed to the GRU layer (the environmental feature extraction network), which returns their encoded representation (an environmental feature)), and wherein inputting the fusion feature into the result prediction network further comprises: inputting the environmental feature and the historical vacancy rates of each of the parking lots into the result prediction network (Zeng: page 47365, Section III.E, "Then the model superimposes two LSTM layers with more parameters and higher prediction accuracy"; the encoded exogenous representation is carried forward together with the occupancy series into the stacked output layers that issue the forecast, so the prediction stage receives the environmental representation alongside the parking series rather than the parking series alone; the occupancy series that Zeng supplies to the combined model, page 47365, Section III.E, "The combined prediction model takes multi factors as input, such as occupancy, weather conditions, vacations, etc.", is the historical vacancy rates of the parking lots, so in the combination the prediction stage receives the historical vacancy rates of each of the parking lots together with the environmental feature).
Because Feng, in view of Liu, and Zeng are analogous art and within the same field of endeavor, specifically forecasting the availability of parking spaces at parking lots from their historical availability series using neural network encoders, they address the same problem solving area of improving the accuracy of a parking forecast, accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention, to add the exogenous encoder of Zeng to the model of Feng as deployed on the electronic device of Liu and to carry its output into the prediction stage, with a reasonable expectation of success, such that weather and holiday conditions inform the forecast, to teach an environmental feature extraction network; wherein before inputting the fusion feature into the result prediction network, the method further comprises: acquiring environmental information of each of the parking lots in the area to be predicted before the moment to be predicted, wherein the environmental information at least comprises weather information and holiday information; and obtaining an environmental feature by inputting the environmental information into the environmental feature extraction network; and wherein inputting the fusion feature into the result prediction network further comprises: inputting the first feature, the fusion feature, the environmental feature and the historical vacancy rates of each of the parking lots into the result prediction network. This modification would have been motivated by the desire to account for the irregular swings in demand that a purely periodic model misses (Zeng: page 47367, Section IV.A, "weather, holidays and other factors are randomly changing external factors, affecting people's travel rate and travel mode. At the same time, these factors may affect the aperiodic parking demand, which will have an immediate impact on the occupancy changes, and will cause sequence trend fluctuations"; supplying those factors to the model is what lets it track demand that the historical series alone does not explain). See KSR International Co. v. Teleflex Inc., 550 U.S. 398 (2007); the combination is no more than the combination of prior art elements according to known methods to yield predictable results, MPEP 2143.01.
Regarding dependent claim 18, this is an electronic device claim that is substantially the same as the method of claim 5. Thus, claim 18 is rejected for the same reason as claim 5.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Monasseian et al., US 20220076282 A1 (Mar. 10, 2022) (ABSTRACT A personal electronic parking system and method adapted to identify, track, predict, alert, manage and collect payment, and enforce on-street and off-street parking, the system including a central cloud network adapted to generate and manage user data and parking data, a user interface adapted to show users parking information relevant to the user, a unique machine-readable code, wherein the code provides identification of a specific vehicle used by the user, one or more sensors, one or more meter devices adapted to connect with the central cloud network and the user interface, and a parking payment and enforcement portal).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KUANG FU CHEN whose telephone number is (571)272-1393. The examiner can normally be reached M-F 9:00-5:30pm ET.
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, Jennifer Welch can be reached on (571) 272-7212. 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.
/KC CHEN/Primary Patent Examiner, Art Unit 2143
Prosecution Insights