DETAILED ACTION
Claim 7 rejected under 35 USC §112 as indefinite.
Claims 1-3, 7, and 9 are rejected under 35 USC § 103.
Claims 4-6 and 8 are objected to as allowable dependent claims.
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 § 112
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.
Claim 7 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 7
Claim 7 recites “guide the Actor to produce better actions.” The term “better actions” is representative of a subjective term, because defining “better actions” depends solely on the unrestrained, subjective opinion of a particular individual practicing the invention. The specification does not provide any standard for measuring the scope of the term, nor any restrictions on the exercise of subjective judgment. See MPEP 2173.05(b)(IV). Accordingly, claim 7 is indefinite.
Claim Rejections - 35 USC § 103
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 1-2, 7, and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Sha et al., U.S. PG-Publication No. 2023/0026782 A1 (hereinafter SHA), in view of Yeh et al., U.S. PG-Publication No. 2022/0014963 A1 (hereinafter YEH), further in view of Xu et al., U.S. PG-Publication No. 2022/0377137 A1 (hereinafter XU).
Claim 1
SHA discloses a digital twin-based edge-end collaborative scheduling method for heterogeneous tasks and resources. ¶ 0042: The digital-twin management platform “takes a set of parameters into consideration” and provides a digital-twin based solution to “model and simulate a MEC system,” including “the task-offloading process,” to produce “optimized solutions.” ¶¶ 0033-0034: The physical MEC environment includes cloud computing, edge servers, corresponding base stations, and end devices. The digital-twin architecture “models the entire MEC system environment,” including its offloading process. ¶ 0029: Resource-limited end devices wirelessly connect through base stations to edge servers and offload tasks requiring additional “compute, storage, network, etc.” resources. ¶ 0044: Inputs include different workloads, task sizes, required CPU cycles, latency requirements, device transmit powers, bandwidths, network speeds, server dynamics, and edge CPU/VM parameters.
SHA discloses 1) establishing an edge wireless network based on digital twin. ¶ 0029: The MEC system has multiple edge servers with corresponding base stations; each base station acts as an access point through which the user device “wirelessly connect[s]” with an edge server. ¶ 0033: The MEC environment includes “the edge computing environment, the cloud computing environment, and one or more user devices.” SHA constructs a digital twin that “virtually represents” the entire environment or portions of it. ¶ 0034: The digital-twin architecture includes edge-server, cloud, hardware, and end-device digital twin components.
SHA discloses 2) constructing an edge-end collaborative scheduling problem of heterogeneous tasks and resources according to the deadline requirements of the heterogeneous tasks and the constraints of the heterogeneous computation and communication resources. ¶¶ 0030-0031: Offloaded tasks are assigned according to edge-server queue capacity. When one server cannot accept another task, the task is transferred to another server having “availability in its task queue.” ¶¶ 0039-0040: The digital twin receives task-load and real-time performance data, “estimates states of edge servers” and provides indications of which servers are or will be available. ¶ 0044: The scheduling inputs include “latency requirements,” task workload, required CPU cycles, device transmit power, transmission bandwidth and speed, and edge CPU/VM parameters.
SHA discloses 5) performing offline … training … by digital twin. ¶ 0038: In addition to real-time operation “offline (non-real time) optimizations can also be determined,” including “offline cloud optimizations.” ¶¶ 0053-0054: The digital twin uses historical and real-time data, AI algorithms, and repeatable simulation to determine “optimal configurations” and evaluate workloads and resource performance before deployment.
SHA discloses 6) perceiving an environment state online by end devices, and performing … task offloading and computation and communication resource. ¶¶ 0039-0041: The physical MEC system and digital twin exchange “real-time performance data”; the digital twin estimates edge-server states and supplies that state information to the mobile offloading entity. The digital twin shares estimated server-state information with the mobile device so that the device “can make an optimized offloading decision.” ¶¶ 0044-0045: Digital-twin processing considers device transmit power, bandwidth, network speed, edge CPU/VM parameters, task CPU cycles, “communication latency,” “computational latency,” and CPU utilization. ¶ 0057: ML analytics provide “recommendations for offloading optimization,” “suggestions on optimization parameters,” and “resource allocation and management.”
SHA discloses to collaboratively process the heterogeneous tasks and minimize the total task processing delay. ¶ 0034: The digital twin models the entire MEC environment and offloading process “so as to determine ways to reduce latency.” ¶¶ 0048-0050: ML generates “optimal offloading decisions” and assigns edge servers to tasks based on available edge resources and device latency requirements, contemplated objectives include minimizing offloading latency.
SHA does not expressly disclose the method characterized by achieving collaborative scheduling of heterogeneous tasks and heterogenous computation and communication resources based on multi-agent deep reinforcement Learning; 3) converting the scheduling problem into a multi-agent Markov decision process problem; 4) constructing an Actor-Critic neural network model based on the multi-agent deep reinforcement learning to solve the multi-agent Markov decision process problem; 5) performing offline centralized training of the Actor-Critic neural network model by digital twin to obtain an experience pool and neural network parameters; and 6) perceiving an environment state online by end devices, and performing distributed execution of task offloading and computation and communication resource according to the Actor-Critic neural network model under centralized training.
YEH discloses the method characterized by achieving collaborative scheduling of heterogeneous tasks and heterogenous computation and communication resources based on multi-agent deep reinforcement Learning. ¶ 0029: YEH’s embodiments relate to “data processing, service management, resource allocation, compute management, network communication, application partitioning” and “dynamically supporting multiple users, service instances, and applications in a “distributed edge computing environments.” ¶ 0047: The number of active UEs and traffic flows varies over time, and different flows have “different QoS requirements.” YEH provides scalable DRL architectures for this dynamic wireless environment. ¶¶ 0048-0049: YEH expressly identifies a “Collaborative Multi-Agent DRL” architecture in which each UE is configured as an agent and interacts with the environment to collect observations. ¶¶ 0050-0052: The UE agents use a “shared team reward,” the coordinator “centrally train[s] a common model,” and the agents “collaborate with each other” while independently determining actions.
YEH discloses 3) converting the scheduling problem into a multi-agent Markov decision process problem. ¶ 0049: Multiple UE agents interact with an environment, collect observations and runtime statistics, determine the environment state, and receive a shared reward. ¶¶ 0050-0051: Each UE derives and executes an action from its individual observations and actions, receives an updated model, and repeats the process. ¶ 0568: Reinforcement learning is defined as goal-oriented learning through interaction with an environment, and the disclosed RL algorithms expressly include a “Markov decision process” and “deep RL.”
YEH discloses 4) constructing an Actor-Critic neural network model based on the multi-agent deep reinforcement learning to solve the multi-agent Markov decision process problem. ¶ 0055: Actor-Critic learning simultaneously learns a “policy function” and a “value function.” The Actor is the “decision making entity” while the Critic “evaluates the determined action” and provides feedback. ¶ 0056: The Actor outputs an action for the current state, while the Critic evaluates the action and determines adjustments to the “action determination/learning process.” ¶ 0057: The disclosed Actor/Critic learning algorithms expressly include “Deep DPG (DDPG) and “Twin Delayed DDPG (TD3).”
YEH discloses 5) performing … centralized training of the Actor-Critic neural network model … to obtain an experience pool and neural network parameters. ¶ 0049: The coordinator is employed as a “centralized training engine” and gathers the agents’ environment states and rewards. ¶¶ 0050-0051: The coordinator trains and updates the common model using UE observations and actions and deploys the updated model to the participating UE agents. ¶ 0318: The Actor-Critic architecture includes “replay buffer” storing experience structures such as (st, at, rt, st+1). ¶¶ 0140-0144: Training initializes Actor, Critic, and representation-network parameters, stores experience transitions, samples a minibatch from the replay buffer, and updates the neural networks.
YEH discloses 6) perceiving an environment state online by end devices, and performing distributed execution of … computation and communication resource according to the Actor-Critic neural network model under centralized training. ¶¶ 0049-0051: Each UE “collects observations from the environment,” whare are used as state inputs and reward information. ¶¶ 0074-0075: Agent observations include available uplink and downlink radio resources, virtualized resources, power and energy information, QoS flows, network slices, and latency indicators. ¶ 0050: The common model is centrally trained and deployed to UE agents, after which each UE independently predicts an action from its individual observations. YEH expressly calls this “centralized training and distributed inferences.” ¶ 0051: The UEs “update their local model” using the model deployed by the coordinator and perform actions according to the learned strategy and their current observations. ¶¶ 0176-0177: RAN functions perform “uplink and downlink dynamic resource allocation, “radio-bearer management,” “data packet scheduling,” and dynamic allocation of physical resource blocks and modulation/coding schemes to individual UEs. ¶¶ 0246-0248: MEC applications provide delay, throughput, loss, and bandwidth requirements; the bandwidth-management service allocates static or dynamic uplink/downlink bandwidth, including bandwidth size and priority. ¶ 0361: The RIC controls radio-resource allocation, scheduling requests, configured grants, slice-level resource-block quotas, QoS mapping, bearer splitting and carrier aggregation.
YEH discloses collaboratively process the heterogenous tasks and minimize the total task processing delay. ¶ 0029: The disclosed environment dynamically supports multiple users, service instances, and applications through data processing, compute management, resource allocation, network communication, and application partitioning. ¶ 0075: The agent’s state information includes latency performance indicators for network slices and network functions. ¶¶ 0246-0248: MEC applications provide delay requirements that influence traffic-management and bandwidth-allocation operations.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the digital-twin based MEX task offloading and resource optimization framework of SHA to incorporate the multi-agent Actor-Critic architecture, centralized training, and distributed UE execution taught by YEH. One of ordinary skill in the art would be motivated to integrate YEH’s multi-agent Actor-Critic architecture into SHA, with a reasonable expectation of success, in order to accommodate changing number of active UEs and traffic flows, dynamic UE movement, and varying QoS requirements in a scalable manner, while allowing each UE to determine actions from its individual observations using a commonly trained model, as taught by YEH ¶¶ 0047-0051.
SHA-YEH does not expressly teach performing distributed execution of task offloading according to a neural network model.
XU discloses performing distributed execution of task offloading according to a neural network model. ¶ 0010: XU establishes an Industrial Internet of Things edge-computing network divided into multiple cluster domains, each containing an edge server and one or more wireless devices. A computation task may be executed at “the device itself,” an edge server in the device’s cluster domain, or an edge server “in another cluster domain,” with the selected edge server allocating computing resources to the offloaded task. ¶¶ 0013-0014: XU establishes a reinforcement-learning model, trains a deep neural network, and obtains an “optimal offloading location decision” as part of a joint optimization strategy. ¶¶ 0039-0045: The scheduling variables determine whether each task executes locally or is offloaded and identify which one of multiple edge servers executes the task.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the distributed multi-agent scheduling framework of SHA-YEH to incorporate XU’s neural-network based execution of task offloading among end devices, same-domain edge servers, and remote-domain edge servers. One of ordinary skill in the art would be motivated to integrate SU’s distributed task-offloading execution into SHA-YEH, with a reasonable expectation of success, in order to select suitable edge servers, increase computational efficiency, allocate appropriate computing resources to offloaded tasks, and minimize task-processing overhead in a heterogenous edge-computing environment, as taught by XU ¶¶ 0005-0008.
Claim 2
SHA discloses the method, characterized in that the edge wireless network based on digital twin comprises: N base stations configured with edge server and M end devices. ¶ 0042: The digital-twin management platform provides a digital-twin based solution to “model and simulate a MEC system,” including “the task offloading process,” and provide “optimized solutions.” ¶ 0029: The MEC system includes edge servers 104-1 through 104-5 and “corresponding base station” 106-1 through 106-5. A mobile end device wirelessly connects to the corresponding edge servers through those base stations. ¶¶ 0033-0034: The complete MEC environment includes cloud computing, edge servers, corresponding base stations, and “one or more user devices.” The digital twin models the entire MEC environment.
SHA discloses the base stations are configured with the edge servers and used for providing computation resources for a plurality of end devices and supporting scheduling of the end devices within a coverage range. ¶ 0029: Each edge server has a “corresponding base station” connected to it, and each base station provides the wireless access point for connecting a user device to that edge server. ¶¶ 0029-0030: Resource-limited end devices offload tasks requiring additional “compute, storage, network, etc.” resources to one or more edge servers. ¶ 0031: As the mobile device moves among the coverage areas of the different base stations, tasks are scheduled to available edge servers according to device proximity and server queue availability.
SHA discloses the end devices are used for computing the heterogeneous tasks locally, and supporting offloading of the heterogeneous tasks to the edge server through a wireless channel for edge computing. ¶ 0029: The resource limited user device “executes at least part of an application” having one or more computation tasks. The device wirelessly connects through a base station to the corresponding edge server and offloads tasks requiring more resources that the device possesses. ¶¶ 0030-0031: The device offloads tasks to one or more edge servers; a task may also be forwarded to another edge server when the initially connected server lacks available queue capacity.
SHA discloses the digital twin is placed on a cloud server of the network, represented as a virtualization model established by the base stations and the end devices comprised in the network. ¶ 0033: The edge servers and corresponding base stations are coupled to “one or more cloud computing platforms,” and SHA virtually represents the resulting cloud-edge-end MEC environment. ¶¶ 0034-0036: The architecture includes a “cloud digital twin virtual component” that stores data, provides data access, performs advanced analytics, executes trained ML models, and conducts cloud optimization. The architecture contains digital-twin virtual components for edge servers, cloud infrastructure, physical hardware, and end devices, each modeling the functionality of its corresponding physical component. ¶ 0067: The method “generates a virtual representation” comprising one or more digital-twin models of the MEC environment and manages the physical environment through that representation.
SHA discloses the digital twin used for evaluating the operating states of the base stations, the edge server and the end devices, the types of the computation resources, and the amount of the computation and communication resources. ¶¶ 0039-0040: The physical network and digital twin exchange “real-time performance data”; the digital twin “estimates states of edge servers” and determines their present or predicted task-processing availability. ¶¶ 0044-0045: Digital-twin inputs include environmental state, device movement, server dynamics, network conditions, workloads, and device/resource parameters. Outputs include computation, communication, and offloading latency, CPU utilization, energy use, and failed-task rate. ¶ 0029: The resources available to the end device and edge infrastructure include “compute, storage, network, etc.” ¶¶ 0034-0037: Separate virtual components model edge-server, cloud, hardware, and end-device functionality, including CPU/VM operation and trained-model execution. ¶ 0044: Inputs include task CPU-cycle demand, edge CPU/VM parameters, device transmit power, bandwidth, network speed, workload, and server dynamics. ¶ 0045: Outputs include CPU utilization and computation, communication, and offloading latency.
SHA discloses the digital twin supporting the training of a … learning method to carry out the edge-end collaborative scheduling of the network. ¶¶ 0035-0038. Edge and cloud digital-twin components execute trained ML models and perform cloud, local, online, and offline optimization. ¶¶ 0053-0054. The digital twin uses historical and real-time data, AI algorithms, and repeatable simulations to determine optimal configurations and evaluate workloads and resource performance before deployment.
SHA does not expressly disclose the digital twin supporting the training of a deep reinforcement learning method.
YEH discloses supporting the training of a deep reinforcement learning method to carry out the edge-end collaborative scheduling of the network. ¶¶ 0048-0050: YEH provides a “Collaborative Multi-Agent DRL” architecture in which UE agents collect network observations and actions reported by the UEs, deploys the updated model, and the UEs perform scheduling actions based on current observations. ¶ 0051: The coordinator performs centralized training using observations and actions reported by the UEs, deploys the updated model, and the UEs perform scheduling actions based on current observations. ¶¶ 0055-0057: YEH implements deep RL using Actor and Critic neural networks and expressly identifies DDPG and TD3 learning.
Claim 7
YEH discloses the method characterized in that constructing an Actor-Critic neural network model based on the multi-agent deep reinforcement learning comprises an Actor network and a Critic network. ¶ 0055” YEH’s reinforcement-learning architecture includes “an actor network 303, and a critic network 305,” which implement Actor-Critic learning. ¶ 0138: The DRL framework expressly includes “an actor network 903 and a critic network 905” and may employ Actor-Critic or DDPG.
YEH discloses the Actor network adopts strategy-based deep neural networks. ¶ 0055: The Actor learns the “policy function” and operates as the “decision making entity” having tunable policy parameters. ¶ 0057: The Actor’s policy-gradient learning alternatives include DPG, DDPG and “Twin Delayed DDPG (TD3).”
YEH discloses comprising an estimation Actor network for training. ¶¶ 0140-0141: During training, Actor network 903 is initialized with its own parameters and trained with the Critic and representation network. ¶¶ 0145-0146: The current Actor is trained using the policy gradient supplied through the Critic’s evaluation of the Actor’s actions.
YEH discloses a target Actor network for executing the action to generate agent actions. ¶ 0144: Temporal-difference targets are computed using outputs of the “target actor and critic networks,” with the target Actor having its own target weights. ¶ 0057: YEH expressly identifies DDPG and TD3 both of which use current and target Actor networks.
YEH discloses the Critic network adopts value-based deep neural networks, comprising an estimation Critic network. ¶¶ 0140-0141: Critic network 905 is initialized with its own current parameters and trained during the training stage. ¶¶ 0144-0145: The current Critic is trained to minimize a loss determined from sampled experiences and temporal difference targets.
YEH discloses a target Critic network to evaluate the actions of the Actor and guide the Actor to produce better actions. ¶ 0144: YEH expressly computes temporal-difference targets using the outputs of “target actor and critic networks,” with the target Critic having separate target weights. ¶¶ 0055-0056: The Critic “evaluates the determined action” of the Actor and outputs a value indicating the quality of that action relative to other actions. The Actor attempts to improve its policy based on the Critic’s evaluation, and the Critic determines adjustments to the Actor’s action determination and learning process. ¶¶ 0138, 0146: The Actor updates its policy in the direction suggested by the Critic; the Critic’s action gradient updates the Actor when an action has a higher value than a previous action.
Claim 9
SHA discloses a) downloading the offline centralized training results of the digital twin by all the agents. ¶ 0038: The digital twin supports “offline (non-real time) optimizations,” including “offline cloud optimizations.” ¶¶ 0053-0054: The digital twin uses historical and real-time data, AI algorithms, and repeatable simulation to obtain “optimal configurations” and evaluate workloads and resource performance before deployment.
SHA discloses b) perceiving an environment by end devices. all the agents to obtain respective states. ¶¶ 0039-0040: The physical MEC system and digital twin exchange “real-time performance data”; the digital twin estimates edge-server states and provides that state information to the mobile offloading entity.
SHA discloses c) performing task offloading. ¶¶ 0039-0041: The digital twin provides estimated server states to the mobile device so that the device “can make an optimized offloading decision.”
SHA does not expressly disclose a) downloading the offline centralized training results … by all the agents; b) perceiving an environment by all the agents to obtain respective states; computing respective rewards according to the trained neural network parameters, and executing actions online in a distributed mode, wherein after the state sₘ(t) of the agent m is inputted to the target Actor network, the action am(t) is outputted according to the reward rₘ(t), that is, the matching decision result of the computation types, the task offloading ratio, the device transmission power and the computation resource allocation result of the end device m and N edge servers; c) performing task offloading and collaborative computing by all end devices according to the output actions of respective neural networks, that is, the scheduling results of the heterogeneous tasks and resources.
YEH discloses a) downloading the offline centralized training results … by all the agents. ¶ 0050: A coordinator “centrally train[s] a common model” and deploys the trained model to participating UE agents. ¶ 0051: The coordinator deploys the updated AI/ML model to the UEs, and the UEs “update their local model” using the deployed model.
YEH discloses b) perceiving an environment by all the agents to obtain respective states. ¶ 0049: Each UE is configured as an agent and interacts with the environment to collect observations and runtime statistics used to determine the environment state. ¶ 0051: Each UE collects observations used as the state input for the reinforcement-learning agent.
YEH discloses computing respective rewards according to the trained neural network parameters, and executing actions online in a distributed mode. ¶¶ 0049-0051: UE observations and actions are reported to the coordinator, which calculates a centralized reward and uses it to train and update the common model. ¶¶ 0138-0144: The Actor-Critic framework stores, state, action, reward, and next-state experiences and trains the Actor and Critic parameters from sampled experiences. ¶ 0050: After centralized training, each UE independently derives an action from its individual observations. YEH expressly calls the arrangement “centralized training and distributed inferences.” ¶ 0051: The UEs execute actions according to the deployed model and their current observations.
YEH discloses wherein after the state sₘ(t) of the agent m is inputted to the target Actor network, the action am(t) is outputted according to the reward rₘ(t). ¶ 0056: The Actor receives the current state and outputs a value representing the selected action; the Critic evaluates that action and provides feedback for improving the Actor. ¶ 0144: Temporal-difference targets are calculated using outputs of the “target actor and critic networks,” which have separate target weights.
YEH discloses c) performing task offloading and collaborative computing by all end devices according to the output actions of respective neural networks. ¶ 0029: The disclosed distributed edge embodiments encompass “resource allocation, compute management, network communication, application partitioning” and data processing. ¶¶ 0050-0051: The common model is centrally trained and deployed to UE agents, which independently determine and execute actions from their local observations. ¶¶ 0176-0177: RAN functions perform “uplink and downlink dynamic resource allocation,” radio-bearer management, data scheduling, and dynamic allocation of communication resources to individual UEs. ¶¶ 0246-0248: MEC applications provide delay, throughout, loss, and bandwidth requirements, bandwidth services allocate static or dynamic uplink/downlink resources, including bandwidth amount and priority. ¶¶ 0050-0052: UE agents use a common centrally trained model and shared team reward but independently produce and execute actions using their respective observations. The agents “collaborate with each other.”
YEH discloses that is, the scheduling results of the heterogeneous tasks and resources. ¶ 0029: The system supports multiple users, service instances, and applications through data processing, compute management, resource allocation, network communication, and application partitioning.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over SHA, in view of YEH, further in view of XU, further in view of Hu et al., U.S. PG-Publication No. 2021/0266834 A1 (hereinafter HU).
Claim 3
XU discloses the method characterized in that for a single end device, the tasks can be non-offloaded … or completely offloaded to one or more edge servers for computing. ¶ 0010: Each end device generates a computation task. The available execution locations include “the device itself,” an edge server in the device’s cluster domain, and an edge server “in another cluster domain.” ¶ 0039: The first decision variable xi determines whether the computation task Qi executes locally or is offloaded. The third decision variable γi ∈ {1, 2, …, N} identifies the edge server that executes the task. ¶ 0045: XU incorporates the selected offloading location, optimal communication power, and optimal computing-resource allocation into the offloading-location decision model.
XU does not expressly disclose for a single end device, the tasks can be … partially offloaded.
SHA discloses for a single end device, the tasks can be … partially offloaded. ¶ 0029: The user device “executes at least part of an application,” while “one or more tasks” of that application are offloaded to and performed by a nearby device having greater computation, storage, or network resources. ¶ 0030: The user device offloads tasks “to one or more of edge servers 104” for execution.
SHA-YEH-XU does not expressly disclose the transmission rate of the end device during task offloading is:
PNG
media_image1.png
166
452
media_image1.png
Greyscale
wherein Wₘ,ₙ represents the bandwidth between the end device m and the edge server n, σ2n represents the noise at the edge server n, gm,n and gm’,n represent channel power gains from the end device m and the end device m’ to the edge server n, respectively; and pm and pm’ represent the transmit power of the end device m and the end device m', respectively.
Hu discloses the transmission rate of the end device during task offloading is:
PNG
media_image1.png
166
452
media_image1.png
Greyscale
¶ 0018: HU establishes a mobile edge-computing network having an edge server deployed near a base station, multiple cellular user devices, and multiple device-to-device users. A cellular device’s computation task may be executed “at the local of CUE of the edge computing server,” while channel reuse by other devices produces communication interference. ¶ 0016: Defines gamma_m as the received signal from device m divided by noise plus summed interference from other transmitting devices. ¶ 0025: Calculates task-uploading rate r_m using allocated bandwidth B/M multiplied by log2(1+gamma_m). This is the same desired signal over interference plus noise structure; and a direct Shannon transmission-rate calculation using allocated bandwidth and interference-inclusive SINR.
Hu discloses wherein Wₘ,ₙ represents the bandwidth between the end device m and the edge server n. The base station has total bandwidth B, and each device’s uploading rate uses allocated bandwidth B/M.
Hu discloses σ2n represents the noise at the edge server n. ¶¶ 0016-0017: The SINR denominator includes noise power sigma2 at the receiving base station.
Hu discloses gm,n and gm’,n represent channel power gains from the end device m and the end device m’ to the edge server n, respectively. ¶¶ 0017-0019: h_m,B is the desired device-to base-station gain, and h_k,B is an interfering device-to-base-station channel gain.
Hu discloses pm and pm’ represent the transmit power of the end device m and the end device m', respectively. ¶ 0017: p_m is the desired device’s transmission power, and p_k in the interfering device’s transmission power.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the task-offloading and computation/communication-resource scheduling methods of SHA-YEH-XU to incorporate HU’s interference-aware transmission-rate model for wireless task uploading. One or ordinary skill in the art would be motivated to integrate HU’s interference-aware transmission-rate model into SHA-YEH-XU, with a reasonable expectation of success, in order to account for interference caused by channel reuse, effectively utilize available channel resources, reduce task-processing delay and energy consumption, and maintain communication reliability, as taught by HU ¶¶ 0048-0050.
Allowable Subject Matter
Claims 4-6 and 8 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. See Liu et al., U.S. PG-Publication No. 2023/0081937 A1; ¶¶ 0076, 0083, 0091, 0102, 0204-0205: The reference describes multiple terminal devices having computation-intensive, delay-sensitive tasks that may be processed locally or offloaded to an edge node. The terminal devices and edge nodes use multi-agent deep reinforcement learning to determine computation-offloading decisions and allocate edge computing resources while optimizing delay and energy consumption. The disclosed implementation uses offline-trained MADDPG, in which each agent operates an Actor-Critic/DDPG model.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to FRANK D MILLS whose telephone number is (571)270-3194. The examiner can normally be reached M-F 9-5:30 CT.
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, KEVIN YOUNG can be reached at (571)270-3180. 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.
/FRANK D MILLS/Primary Examiner, Art Unit 2194 August 7, 2026
Prosecution Insights