Vehicle-mounted computing power network task offloading method and system based on identity dynamic conversion

Abstract

This invention provides a method and system for task offloading in a vehicle-mounted computing network based on dynamic identity conversion, relating to the field of vehicle-mounted computing network technology. First, a vehicle-mounted computing network is constructed; then, an identity conversion mechanism is established, and vehicles in the vehicle-mounted computing network that meet preset identity conversion conditions are dynamically converted according to the mechanism, resulting in vehicles with converted identities; next, a vehicle-mounted computing network priority factor is calculated, and priority compensation is provided to the vehicles with converted identities using this priority factor. This invention, through the identity conversion mechanism and the vehicle-mounted computing network priority factor, improves the determinism of the vehicle-mounted server as a computing node, enhances system stability, reduces the overall task processing time and average task queue length, and effectively reduces the communication time and queuing time of tasks in the queue.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted computing network technology, and in particular to a method and system for offloading tasks in vehicle-mounted computing networks based on dynamic identity transformation. Background Technology

[0002] Vehicle-mounted computing networks were proposed to address the demands of large-scale computing tasks. They treat vehicles as computing nodes, forming a computing pool together with other edge servers and personal mobile devices. However, vehicle-mounted servers in these networks possess a dual identity: both user and computing node. As computing power requesters, they aim to maximize the use of external resources to quickly complete their own tasks; yet, as computing power providers, they need to minimize the time spent processing tasks for others and reserve sufficient processing time for their own potential tasks. This dual identity, along with the mobility and security issues of vehicles, introduces inherent uncertainties to vehicle-mounted computing networks, impacting the reliability of vehicle-mounted servers as computing nodes.

[0003] The uncertainty brought about by the dual identity of in-vehicle servers also extends to task queue management. For example, how to handle tasks queued on the in-vehicle server when an emergency occurs while the vehicle is on the road, and how to handle tasks in the queue after the vehicle undergoes an identity change, are all problems that urgently need to be solved. Furthermore, the queuing time of tasks in the queue is positively correlated with the queue length. Therefore, when task queuing exists, designing a reasonable task prioritization mechanism to balance task queuing time and the timely completion of critical tasks is a key challenge in reducing the total task processing time.

[0004] In complex vehicular computing network environments, the task offloading problem between users and computing nodes is crucial, and sound decision-making can effectively reduce task processing time. Since the decisions of a single computing node affect other nodes, the system requires an efficient and adaptive algorithm to solve for the optimal offloading strategy between users and computing nodes. Summary of the Invention

[0005] To overcome the problems of instability, long task queuing time, and difficulty in optimizing task unloading in vehicle-mounted computing networks, this invention provides a method and system for unloading tasks in vehicle-mounted computing networks based on dynamic identity transformation.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] This invention provides a method for offloading onboard computing power network tasks based on dynamic identity transformation, comprising:

[0008] Constructing an in-vehicle computing network;

[0009] An identity conversion mechanism is established, and the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions are dynamically converted according to the identity conversion mechanism to obtain the vehicles with converted identities.

[0010] Calculate the vehicle computing power network priority factor, and use the vehicle computing power network priority factor to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network.

[0011] For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated respectively.

[0012] Based on the communication time, computation time, and waiting time, an optimization problem for task offloading in vehicle computing power networks is established.

[0013] The vehicle-mounted computing power network task offloading optimization problem is transformed to obtain the final optimized problem. The final optimized problem is solved to obtain the optimal vehicle-mounted computing power network task offloading result.

[0014] Preferably, the vehicle-mounted computing network includes M user nodes and N computing nodes;

[0015] The user node consists of a personal mobile device and an in-vehicle server;

[0016] The computing nodes consist of personal mobile devices, vehicle-mounted servers, and edge servers;

[0017] Define the user node set as The computing power node set is defined as ,

[0018] in, For personal mobile devices, For vehicle-mounted servers, To serve vehicle aggregation, For edge servers, t represents time slots.

[0019] Preferably, the identity of vehicles in the on-board computing network that meet the preset identity conversion conditions is dynamically converted according to the identity conversion mechanism, including:

[0020] The identity of the vehicle-mounted server is distinguished into mission vehicles and service vehicles;

[0021] When a task vehicle finishes processing a new task, and the CPU utilization of the task vehicle is less than a preset first threshold, the task vehicle's identity is changed from a task vehicle to a service vehicle.

[0022] When a service vehicle finishes processing a new task, and its CPU utilization exceeds a preset second threshold, the service vehicle's identity is changed from a service vehicle to a task vehicle.

[0023] Preferably, the step of performing priority compensation on the vehicle after identity transformation using a preset onboard computing power network priority factor includes:

[0024] The priority factors for the vehicle-mounted computing network are as follows:

[0025]

[0026] in, It is a source factor for the task. It is a mission safety score. It is a collection of tasks. It is a set of computing power nodes. It refers to the data size of the task;

[0027]

[0028] in, It is the time when the task arrives at the computing node. This is the maximum acceptable delay for the task;

[0029] The vehicle computing power network priority factor is used to provide priority compensation to the vehicle after identity conversion, thereby obtaining a priority-compensated vehicle computing power network.

[0030] Preferably, for the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated using preset communication models, computing models, and queuing models, respectively.

[0031] Preferably, the communication model includes a first transmission time communication model from a personal mobile device to a computing node and a second transmission time communication model from an in-vehicle server to a computing node;

[0032] The computing model includes a personal mobile device computing time computing time computing time computing time computing model, an in-vehicle server computing time computing time computing model, and an edge server computing time computing time computing model; the queuing model is a waiting time queuing model.

[0033] The communication time includes the first transmission time from the personal mobile device to the computing node and the second transmission time from the vehicle server to the computing node.

[0034] The computation time includes the computation time of personal mobile devices, the computation time of in-vehicle servers, and the computation time of edge servers.

[0035] The calculation expression for the first transmission time communication model is:

[0036]

[0037] in, The size of task data on a personal mobile device. The transmission rate from personal mobile devices to computing nodes;

[0038] The calculation expression for the second transmission time communication model is:

[0039]

[0040] in, The size of the task data for the vehicle-mounted server. This refers to the transmission rate from the vehicle-mounted server to the computing node.

[0041] The calculation expression for the personal mobile device computing time calculation model is as follows:

[0042]

[0043] in, This indicates the size of the task data to be offloaded from the vehicle server to the personal mobile device. This indicates the size of the task data that will be unloaded from the personal mobile device to the personal mobile device. This indicates the number of CPU cycles required for a personal mobile device to complete a task. This indicates the CPU frequency of the computing power node on a personal mobile device.

[0044] The calculation expression for the vehicle-mounted server computing time calculation model is as follows:

[0045]

[0046] in, This indicates the size of the task data to be offloaded from the vehicle server to the vehicle server. This indicates the size of the task data that will be offloaded from the personal mobile device to the in-vehicle server. This indicates the number of CPU cycles required for the onboard server to complete a task. This indicates the CPU frequency of the onboard server computing node;

[0047] The calculation expression for the edge server computing time calculation model is as follows:

[0048]

[0049] in, This indicates the size of the task data to be offloaded from the vehicle server to the edge server. This indicates the size of the task data that is offloaded from a personal mobile device to an edge server. This indicates the number of CPU cycles required for an edge server to complete a task. This indicates the CPU frequency of the edge server computing nodes;

[0050] The expression for the waiting time queuing model is:

[0051]

[0052] in, Total waiting time for a task on a single device.

[0053] Preferably, the expression for the vehicle-mounted computing power network task offloading optimization problem is as follows:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] in, This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot.

[0060] Preferably, the Lyapunov optimization method is used to transform the on-board computing power network task offloading optimization problem, resulting in the following expression for the final optimized problem:

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] in, The number of tasks arriving in the computing node queue. The task processing capacity of the computing node queue. It is the Lyapunov weight. This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot.

[0067] Preferably, solving the final optimization problem to obtain the optimal onboard computing power network task offloading result includes:

[0068] The final optimization problem is transformed into a Markov decision process problem.

[0069] The Markov decision process problem is solved using a multi-agent proximal optimization method to obtain the optimal onboard computing power network task offloading result.

[0070] The present invention also provides a system comprising:

[0071] The vehicle-mounted computing network construction module is used to build vehicle-mounted computing networks.

[0072] The vehicle identity dynamic conversion module is used to establish an identity conversion mechanism, and to dynamically convert the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions according to the identity conversion mechanism, so as to obtain the vehicles after identity conversion.

[0073] The priority compensation module is used to calculate the priority factor of the vehicle computing power network, and use the priority factor of the vehicle computing power network to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network.

[0074] The time calculation module is used to calculate the communication time, calculation time, and waiting time for the priority-compensated vehicle computing network, respectively.

[0075] The optimization problem establishment module is used to establish an optimization problem for unloading on-board computing power network tasks based on the communication time, computation time, and waiting time.

[0076] The network task unloading module is used to transform the vehicle computing power network task unloading optimization problem into a final optimized problem, solve the final optimized problem, and obtain the optimal vehicle computing power network task unloading result.

[0077] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0078] This invention first constructs an onboard computing power network; then, it establishes an identity conversion mechanism to dynamically convert the identities of vehicles in the onboard computing power network that meet preset identity conversion conditions, resulting in vehicles with converted identities; next, it calculates an onboard computing power network priority factor and uses this factor to provide priority compensation to the vehicles with converted identities, resulting in a priority-compensated onboard computing power network; then, it calculates the communication time, computation time, and waiting time for the priority-compensated onboard computing power network; based on these factors, it establishes an onboard computing power network task offloading optimization problem; finally, it transforms the onboard computing power network task offloading optimization problem to obtain a final optimization problem, solves this final optimization problem, and obtains the optimal onboard computing power network task offloading result. This invention, through the identity conversion mechanism and the onboard computing power network priority factor, improves the determinism of the onboard server as a computing power node and effectively reduces the communication time and queuing time of tasks in the queue, thereby reducing the overall task processing time of the system. Attached Figure Description

[0079] Figure 1 This is a flowchart illustrating the onboard computing power network task offloading method based on dynamic identity conversion in Example 1.

[0080] Figure 2 This is a schematic diagram of the identity conversion structure of the vehicle-mounted computing network in Example 2;

[0081] Figure 3 This is a schematic diagram of the vehicle-mounted computing power network task offloading system based on dynamic identity conversion in Example 3. Detailed Implementation

[0082] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0083] To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions;

[0084] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.

[0085] To facilitate understanding of this embodiment, the prior art information of this embodiment is first introduced as follows:

[0086] To handle large-scale tasks, wireless computing networks have been proposed to provide dense and sufficient computing services. All entities within these networks not only provide computing power but also request computing services according to their needs. With technological advancements, vehicles are considered capable of obtaining computing resources not only from edge servers connected to roadside units or base stations but also providing computing services to neighboring vehicles and users within their coverage area. Thanks to the high-speed mobility and flexibility of vehicles, vehicle-assisted wireless computing networks can address the scheduling issues of computing resources in wireless computing networks. However, in the context of the Internet of Things, there is still a significant gap in computing resources. In this context, in-vehicle computing networks have been proposed. These networks not only utilize the computing resources of edge servers and in-vehicle servers but also consider the idle computing resources of personal mobile devices. Thus, a computing pool composed of personal mobile devices, in-vehicle servers, and edge servers can meet the massive computing demands.

[0087] In vehicular computing networks, edge computing can significantly improve the real-time data processing capabilities of vehicles, supporting rapid response and decision-making in autonomous driving functions, while reducing reliance on remote servers and enhancing system reliability and efficiency. Edge computing is a distributed computing model that pushes computing power and data storage to the network edge, aiming to reduce latency, optimize bandwidth usage, and enhance data privacy. Edge computing achieves these advantages by processing data closer to the data source or terminal device, rather than relying on remote data centers or the cloud. Multiple edge servers exist in the edge layer, typically deployed near micro base stations or cellular network base stations. Thanks to this, the communication quality at the edge layer is much better than in cloud computing, enabling faster feedback of computation results to users. However, due to limitations in device processing power, edge computing relying solely on the computing resources of edge servers can easily lead to resource shortages, preventing the completion of assigned computational tasks. Edge computing is particularly suitable for applications with high real-time requirements, such as autonomous driving, IoT devices, and industrial automation.

[0088] Reinforcement learning, a crucial branch of machine learning, learns optimal behavioral policies through the interaction of agents with their environment. The agent takes actions in the environment, adjusting its policy based on reward signals from the environment to maximize cumulative rewards. This process involves balancing state evaluation, action selection, and reward mechanisms; the agent must continuously explore new actions and optimize its policy using known information. Reinforcement learning is dynamic and adaptive, capable of real-time adjustments based on environmental changes, and is widely applied in games, robot control, and recommender systems to achieve intelligent decision-making and behavior optimization in complex environments. Multi-agent reinforcement learning is another important branch of reinforcement learning, studying how multiple agents learn optimal policies through interaction and cooperation in a shared environment. Each agent has independent states, actions, and reward functions, and continuously optimizes its decisions through interaction with the environment and other agents. In multi-agent reinforcement learning, the agents' behavior follows a stochastic game process, with the goal of maximizing cumulative rewards through the learned policy. Compared to single-agent reinforcement learning, multi-agent reinforcement learning needs to handle more complex dynamic environments and cooperative or competitive relationships between agents. For example, an intelligent agent can learn how to dynamically allocate tasks to in-vehicle devices or edge nodes based on current network conditions, computing resources, and task requirements, in order to achieve goals such as minimizing latency and optimizing energy consumption. By continuously interacting with the environment and adjusting its strategy based on reward signals, reinforcement learning can help the system make optimal decisions in complex and ever-changing in-vehicle and edge computing scenarios.

[0089] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0090] Example 1

[0091] This embodiment provides a method for offloading onboard computing power network tasks based on dynamic identity transformation, such as... Figure 1 As shown, it includes:

[0092] Constructing an in-vehicle computing network;

[0093] An identity conversion mechanism is established, and the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions are dynamically converted according to the identity conversion mechanism to obtain the vehicles with converted identities.

[0094] Calculate the vehicle computing power network priority factor, and use the vehicle computing power network priority factor to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network.

[0095] For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated respectively.

[0096] Based on the communication time, computation time, and waiting time, an optimization problem for task offloading in vehicle computing power networks is established.

[0097] The vehicle-mounted computing power network task offloading optimization problem is transformed to obtain the final optimized problem. The final optimized problem is solved to obtain the optimal vehicle-mounted computing power network task offloading result.

[0098] In the specific implementation process, this embodiment first constructs an on-board computing network; secondly, it establishes an identity conversion mechanism, dynamically converting the identities of vehicles in the on-board computing network that meet the preset identity conversion conditions according to the identity conversion mechanism, resulting in vehicles with converted identities; then, it calculates the on-board computing network priority factor, using the on-board computing network priority factor to provide priority compensation to the vehicles with converted identities, resulting in a priority-compensated on-board computing network; next, it calculates the communication time, computation time, and waiting time for the priority-compensated on-board computing network; based on the communication time, computation time, and waiting time, it establishes an on-board computing network task offloading optimization problem; finally, it transforms the on-board computing network task offloading optimization problem to obtain the final optimization problem, solves the final optimization problem, and obtains the optimal on-board computing network task offloading result. This invention combines a dynamic identity switching mechanism and a task priority factor. The former clarifies the service status of the on-board server, solving its uncertainty problem; the latter optimizes the processing order of tasks in the queue. The two work together to significantly reduce the queuing time of tasks, thereby improving the operating efficiency of the entire on-board computing network.

[0099] Example 2

[0100] This embodiment provides a method for offloading onboard computing power network tasks based on dynamic identity transformation, including:

[0101] Constructing an in-vehicle computing network;

[0102] An identity conversion mechanism is established, and the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions are dynamically converted according to the identity conversion mechanism to obtain the vehicles with converted identities.

[0103] Calculate the vehicle computing power network priority factor, and use the vehicle computing power network priority factor to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network.

[0104] For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated respectively.

[0105] Based on the communication time, computation time, and waiting time, an optimization problem for task offloading in vehicle computing power networks is established.

[0106] The vehicle-mounted computing power network task offloading optimization problem is transformed to obtain the final optimized problem. The final optimized problem is solved to obtain the optimal vehicle-mounted computing power network task offloading result.

[0107] It should be noted that, in this embodiment, the vehicle-mounted computing network includes M user nodes and N computing nodes;

[0108] The user node consists of a personal mobile device and an in-vehicle server;

[0109] The computing nodes consist of personal mobile devices, vehicle-mounted servers, and edge servers;

[0110] Define the user node set as The computing power node set is defined as ,

[0111] in, For personal mobile devices, For vehicle-mounted servers, To serve vehicle aggregation, For edge servers, t represents a time slot;

[0112] For personal mobile devices, in-vehicle servers, and edge servers:

[0113] ;

[0114] ;

[0115] .

[0116] Where I, J, K, M, and N represent personal mobile device, vehicle server, edge server, and user node, respectively. collection, computing nodes A set;

[0117] Indicates the CPU frequency of the computing node;

[0118] It is the total available computing power resources of computing nodes;

[0119] This is the actual number of tasks backlogged in the computing node's task queue.

[0120] A collection of user space tags;

[0121] A collection of spatial labels for computing power nodes;

[0122] It is the number of CPU cycles required to complete a task;

[0123] It is a set of tasks published by personal mobile devices and in-vehicle servers in time slot t, each task Primitive properties are represented by quintuples:

[0124]

[0125] in, It is a mission safety score. It refers to the data size of the task. These are the computing resources required to complete the task. This is the maximum acceptable latency for the task. It is an emergency task vector.

[0126] It should be noted that, in this embodiment, the identity dynamic conversion of vehicles in the vehicle computing network that meet the preset identity conversion conditions according to the identity conversion mechanism includes:

[0127] The identity of the vehicle-mounted server is distinguished into mission vehicles and service vehicles;

[0128] When a task vehicle finishes processing a new task, and the CPU utilization of the task vehicle is less than a preset first threshold, the task vehicle's identity is changed from a task vehicle to a service vehicle.

[0129] When a service vehicle finishes processing a new task, if the CPU utilization of the service vehicle is greater than a preset second threshold, the service vehicle's identity will be changed from a service vehicle to a task vehicle.

[0130] If an urgent task suddenly occurs in time slot t and it is decided to offload locally, Service vehicles from 0 to 1 Prioritize handling your own urgent tasks to ensure real-time response. Also, start a timer to record the time you enter an emergency state. If the emergency task processing time exceeds the threshold Service vehicles State vector Setting the value to 0 forcibly downgrades the vehicle's identity to that of a service vehicle and stops providing computing power services. After the identity change, the vehicle clears all non-self tasks from its task queue and provides priority compensation for interrupted external tasks. Considering the vehicle's high-speed mobility, interrupted external tasks are returned to the onboard computing network for task offloading. If Then continue to complete the unfinished tasks, and then sort the tasks in the queue by priority factor.

[0131] To avoid impacting system stability and efficiency due to frequent vehicle identity changes, vehicles should retain their current identity after any non-emergency identity change. Only after this can the next identity change be carried out, among which It is the minimum role duration.

[0132] Because the identity switching mechanism has a significant impact on task queuing time, analyzing the wasted waiting time caused by identity switching can reveal the identity interruption time. The task's identity interruption time can be mainly composed of the communication time with the old computing node and the waiting time within the old computing node. :

[0133]

[0134] in, It is a task The task was returned to the set of computing node devices in the computing network due to the identity transformation mechanism. It is the communication time of the task. The task is on the equipment Waiting time in the task queue. The optimization goal of task unloading is to minimize the total system task processing time, which includes task communication time, task computation time, and task waiting time.

[0135] It should be noted that, in this embodiment, the schematic diagram of the identity conversion system is as follows: Figure 2 As shown,

[0136] The entire scenario is mainly divided into two parts. The first part involves the in-vehicle server adjusting its identity based on its own CPU occupancy rate. When the in-vehicle server is running on the road and its identity is a service vehicle, it accepts tasks assigned by the in-vehicle computing power network until the CPU occupancy rate reaches the threshold. At this time, the in-vehicle server continues to process the tasks in the task queue and converts its identity to a task vehicle, and the in-vehicle computing power network no longer assigns tasks to this in-vehicle server. When the task vehicle continuously processes the tasks in its own task queue until the CPU occupancy rate is lower than the threshold, the in-vehicle server converts its identity to a task vehicle, and the in-vehicle computing power network includes this in-vehicle server back into the computing power node set. The second part is that when the vehicle encounters an emergency and needs to process the tasks it has issued regarding vehicle driving safety, the vehicle enters an emergency state and starts timing. When the emergency state time τ_i(t) exceeds t_i, the identity is forced to be downgraded to a service vehicle and the computing power service is stopped. After the in-vehicle server undergoes a forced identity conversion, the vehicle clears all non-self tasks in its task queue and provides priority compensation for the interrupted external tasks. Considering the high-speed mobility of the vehicle, the task initiator of the interrupted external task re-performs task offloading at the current geographical location of the initiator. If the emergency state time τ_i (t) < t_i, the vehicle continues to complete the unfinished tasks in the task queue and then sorts the tasks in the queue through a priority factor. Regardless of how the identity of the in-vehicle server changes, it only affects the identity of the in-vehicle server as a computing power node, and the in-vehicle server can always enjoy the rights of a user of the in-vehicle computing power network from beginning to end.

[0137] It should be noted that in this embodiment, in the in-vehicle computing power network, the factors affecting task priority include the following aspects:

[0138] (1) The type of task. Tasks related to vehicle safety are obviously much more important than tasks related to in-vehicle entertainment, and these tasks need to be processed with higher priority;

[0139] (2) The computing power resources required to process the task. Different tasks require different amounts of computing resources. If tasks that require more computing resources are executed first, it will undoubtedly take longer computing time, which will cause subsequent queued tasks to wait longer and affect the efficiency of the offloading system;

[0140] (3) The maximum acceptable delay of the task. During the computing offloading process, it should be ensured as much as possible that the task is completed within the delay constraint. In the VCPN, users continuously generate tasks, and these tasks can be handed over to any computing power node for calculation. The computing power network will reasonably allocate tasks to appropriate computing power nodes according to the learned strategy to achieve the fastest task processing.

[0141] In this situation, the conflict between a large number of computing tasks and computing resources can lead to queuing in the task queues of some computing nodes, regardless of the offloading method adopted. Therefore, in order to reduce the overall queuing time of the system, a priority factor for the vehicle-mounted computing network is set to determine the priority of task processing when queuing occurs.

[0142] The step of applying a preset onboard computing power network priority factor to the vehicle after identity transformation includes:

[0143] The priority factors for the vehicle-mounted computing network are as follows:

[0144]

[0145] in, It is a source factor for the task. It is a mission safety score. It is a collection of tasks. It is a set of computing power nodes. The first term of the formula represents the importance of the task to the computing power node. The second part of the first term represents the proportion of computing power resources required by the task at time slot t to the total computing power resources required by all tasks in the current task queue. When processing tasks on the same computing power node, the more computing power resources a task requires, the more positively correlated it is with the processing time required by the task. Prioritizing tasks with smaller workloads can significantly reduce the queuing time of other tasks in the queue. The second term is the priority compensation term for tasks that need to be re-unloaded because the task was unloaded to a service vehicle and the identity conversion mechanism caused the task to return to the computing power network.

[0146]

[0147] in, It is the time when the task arrives at the computing node, used to distinguish whether the source of the task is the computing node itself. This is the maximum acceptable delay for the task;

[0148] In the absence of urgent tasks, the computing power of time slot t nodes Calculate the on-board computing power network priority factor for each task and sort them according to the priority factor. Select the task with the highest priority to execute. If an emergency task suddenly occurs in time slot t and it is decided to offload locally, pause the tasks being processed and prioritize the emergency task. After the emergency task is processed, process the task with the highest priority factor at that time.

[0149] The vehicle computing power network priority factor is used to provide priority compensation to the vehicle after identity conversion, thereby obtaining a priority-compensated vehicle computing power network.

[0150] For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated using preset communication models, computing models, and queuing models, respectively.

[0151] The communication model includes a first transmission time communication model from a personal mobile device to a computing node and a second transmission time communication model from an in-vehicle server to a computing node.

[0152] The computing model includes a personal mobile device computing time computing time computing time computing time computing model, an in-vehicle server computing time computing time computing model, and an edge server computing time computing time computing model; the queuing model is a waiting time queuing model.

[0153] The communication time includes the first transmission time from the personal mobile device to the computing node and the second transmission time from the vehicle server to the computing node.

[0154] The computation time includes the computation time of personal mobile devices, the computation time of in-vehicle servers, and the computation time of edge servers.

[0155] It should be noted that, in this embodiment, according to the task transfer definition, tasks generated by personal mobile devices and in-vehicle servers can be computed locally or transferred to any target computation. In the in-vehicle computing network, a computing node can only process one task at a time, but can receive more than one task; received tasks are queued and processed according to a task priority mechanism. Indicates that time slot t is generated by a personal mobile device. Transmit to The number of tasks, similarly, let The time slot t is represented by the onboard server. Transmit to The number of tasks.

[0156] For personal mobile device transfer tasks:

[0157] Personal mobile devices can communicate with computing nodes via wireless networks. To reduce interference from other user devices, orthogonal frequency division multiple access (OFDMA) technology is used to reuse the same channels. With computing power equipment The transmission rate between them can be given by the following formula:

[0158]

[0159] in, It is the bandwidth allocated by the computing node to the personal mobile device in time slot t. This indicates the transmission power between the computing node and the personal mobile device during the transmission process. It is the channel gain of the wireless propagation channel between personal mobile devices and computing nodes. That is the power of white noise. It is the set of computing nodes covered by personal mobile devices in time slot t. It refers to the interference power from other personal mobile devices during communication. Determined by the distance between the personal mobile device and the computing node, specifically expressed as the following formula:

[0160]

[0161] in, It is the transmission loss index between personal mobile devices and computing nodes;

[0162] The calculation expression for the first transmission time communication model is:

[0163]

[0164] in, The size of task data on a personal mobile device. The transmission rate from personal mobile devices to computing nodes;

[0165] For in-vehicle server transmission tasks, the transmission rate from the in-vehicle server to the computing node is:

[0166]

[0167] in, It is the bandwidth allocated by the computing node to the vehicle server in time slot t. This indicates the transmission power between the computing node and the vehicle-mounted server during the transmission process. It is the channel gain of the wireless propagation channel between the vehicle-mounted server and the computing node. That is the power of white noise. It is the set of computing nodes covered by the vehicle-mounted server in time slot t. It refers to the interference power from other vehicle-mounted servers during communication. The distance between the vehicle-mounted server and the computing node is determined, and is specifically expressed as follows:

[0168]

[0169] in, It is the transmission loss index between the vehicle-mounted server and the computing node;

[0170] The calculation expression for the second transmission time communication model is:

[0171]

[0172] in, The size of the task data for the vehicle-mounted server. This refers to the transmission rate from the vehicle-mounted server to the computing node.

[0173] For the computing model, since the task can be transferred to any computing node for execution, there are three computing models in the vehicle computing network: personal mobile device computing time computing model, vehicle server computing time computing model, and edge server computing time computing model.

[0174] This embodiment discusses indivisible tasks, therefore each task must and can only be assigned to one computing node.

[0175] set up and These represent task allocation vectors from personal mobile devices and vehicle-mounted servers to other computing power nodes, respectively.

[0176] The calculation expression for the personal mobile device computing time calculation model is as follows:

[0177]

[0178] in, This indicates the size of the task data to be offloaded from the vehicle server to the personal mobile device. This indicates the size of the task data that will be unloaded from the personal mobile device to the personal mobile device. This indicates the number of CPU cycles required for a personal mobile device to complete a task. This indicates the CPU frequency of the computing power node on a personal mobile device.

[0179] The calculation expression for the vehicle-mounted server computing time calculation model is as follows:

[0180]

[0181] in, This indicates the size of the task data to be offloaded from the vehicle server to the vehicle server. This indicates the size of the task data that will be offloaded from the personal mobile device to the in-vehicle server. This indicates the number of CPU cycles required for the onboard server to complete a task. This indicates the CPU frequency of the onboard server computing node;

[0182] The calculation expression for the edge server computing time calculation model is as follows:

[0183]

[0184] in, This indicates the size of the task data to be offloaded from the vehicle server to the edge server. This indicates the size of the task data that is offloaded from a personal mobile device to an edge server. This indicates the number of CPU cycles required for an edge server to complete a task. This indicates the CPU frequency of the edge server computing nodes;

[0185] From the time a task is published to its processing, it must queue in the task queue of the computing node. The processing order is determined by the task priority. Therefore, the queuing time from the time a task is processed to its completion is a significant portion. This queuing time can be divided into two parts: one part is the time the task waits in the queue until it is processed, before being returned to the user due to the identity transformation mechanism. The other part is the period of interruption of identity. This refers to the waiting time of a task that is delayed in the queue because it is waiting in the queue but has not been processed due to the identity conversion mechanism.

[0186] In a continuous-time system, assume the instantaneous task queue length of a device is... Because of the micro-gap All tasks now have an increased waiting time. The total waiting time for this device's tasks is

[0187]

[0188] For discrete systems, computing nodes The total waiting time for a task can also be approximated as...

[0189]

[0190] The expression for the waiting time queuing model is:

[0191]

[0192] in, Total waiting time for a task on a single device.

[0193] It should be noted that, in this embodiment, the optimization objective of task offloading is to minimize the total system task processing time, and the task completion time of personal mobile devices, in-vehicle servers, and edge servers is written as...

[0194]

[0195]

[0196]

[0197] The total system task processing time is taken as the optimization target, expressed as:

[0198]

[0199] The expression for the on-board computing power network task offloading optimization problem is as follows:

[0200]

[0201]

[0202]

[0203]

[0204]

[0205] in, This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot.

[0206] It should be noted that, in this embodiment, the Lyapunov optimization method is used to transform the on-board computing power network task offloading optimization problem;

[0207] Each computing node device has a task queue. To store the tasks that have arrived and are waiting to be assigned, and in time slot t, the number of tasks arriving in each computing node's queue is respectively Processing capacity is The evolution of the queue backlog can be written as:

[0208]

[0209] In a system where users publish tasks, the average number of tasks arriving per time slot should not exceed the average number of tasks processed.

[0210]

[0211] Under this constraint, the computing node queue will not grow indefinitely, thus maintaining the stability of the vehicle-mounted computing system.

[0212] Because the identity transition mechanism is a dynamic process, the potential for task backlog must be considered. If task backlog coincides with the occurrence of urgent tasks, a large number of tasks released and returned to the computing network can severely impact system stability. Based on the Lyapunov optimization framework, the long-term resource-constrained optimization problem is transformed into a deterministic upper bound optimization problem for each time slot to ensure the stability of the task queue and further seek the optimal solution.

[0213] To represent the current state and congestion level of the system, a queue backlog vector is defined. For queue backlog vectors The Lyapunov function is as follows:

[0214]

[0215] From the definition of the Lyapunov function, we know that If and only if When it is a zero vector, it satisfies Furthermore, the Lyapunov drift function is defined as follows:

[0216]

[0217] Lemma 1: For Lyapunov functions ,when At that time, there exists a constant. and This allows the upper bound of the Lyapunov drift to be expressed as:

[0218]

[0219] Proof: According to the formula :

[0220] (1) If Then we can get ,but ,

[0221] (2) If Then we can get:

[0222]

[0223] therefore:

[0224]

[0225] all in all:

[0226] ,

[0227] Extending this to the entire vehicle-mounted computing network, we can obtain:

[0228]

[0229]

[0230] By shifting the terms of the above inequality and performing the expectation operation on both sides, we obtain:

[0231]

[0232]

[0233] Since the number of tasks arriving and the amount of service provided by each queue are bounded, and for any... All exist , and express and The maximum value and We can obtain the following through a simple inequality transformation:

[0234]

[0235] So

[0236]

[0237]

[0238] in It is a positive number, and Lemma 1 is proved.

[0239] Therefore, according to Lemma 1, in time Summing them up, we get:

[0240]

[0241] We can obtain:

[0242]

[0243] This indicates that the backlog of task queues on all computing nodes has an upper limit and cannot grow indefinitely, thus ensuring system stability. By limiting the task arrival rate and the processing rate of computing nodes, the long-term constraint on task queue length in the optimization problem is transformed into a deterministic upper bound optimization problem for each time slot:

[0244]

[0245]

[0246]

[0247]

[0248]

[0249] Furthermore, within the Lyapunov framework, a penalty term is introduced, which allows for a balance between maintaining system stability and reducing task processing time, transforming the optimization problem into the following form:

[0250]

[0251]

[0252]

[0253]

[0254]

[0255] The first term, the Lyapunov drift function, is used to maintain system stability, and the second term is the optimization objective. The Lyapunov control parameter V is a weighting coefficient used to measure the importance of these two objectives. Next, the problem is simplified by minimizing the upper bound of the Lyapunov drift plus a penalty, resulting in the third optimization problem:

[0256]

[0257]

[0258]

[0259] Since B is a constant, we can conclude that the third optimization problem is bounded, and by using Lemma 1, we can obtain the final optimization problem:

[0260]

[0261]

[0262]

[0263]

[0264]

[0265] in, The number of tasks arriving in the computing node queue. The task processing capacity of the computing node queue. It is the Lyapunov weight. This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot.

[0266] Solving the final optimization problem yields the optimal onboard computing power network task offloading result, including:

[0267] The final optimization problem is transformed into a Markov decision process problem.

[0268] The Markov decision process problem is solved using a multi-agent proximal optimization method to obtain the optimal onboard computing power network task offloading result.

[0269] It should be noted that, in this embodiment, within the given vehicular computing network, the computing node observation system determines its own offloading decision to minimize task processing time. The total task processing time and the length of the computing node queue in the vehicular computing network are jointly determined by the joint actions of all computing nodes and the current system state. Since the decisions of a single computing node affect the task execution of other computing nodes, the final optimization problem is transformed into a Markov Decision Process (MDP).

[0270] Multi-agent Programming (MDP) is defined as:

[0271]

[0272] Where M is the set of intelligent agents. It is a set of state spaces. It is an intelligent agent The action space represents the state transition probability. It is an intelligent agent The reward function.

[0273] For the multi-agent problem (MDP), consider two types of computing nodes: all personal mobile devices and in-vehicle servers in the vehicular computing network. These can serve as agents in the MDP. Each agent makes its own task transfer decisions through interaction with its local environment. The agent set is represented as...

[0274] Solving the resource adaptive optimization problem. To implement the proposed method, the optimization problem is formulated as an MDP process, where the state space, action space, and reward function are described below:

[0275] (1) State space: Based on the proposed optimization problem of minimizing task processing time, the state space of the vehicle computing network is... It consists of the states of all devices that can become computing nodes, where:

[0276]

[0277] in, This is a time-varying state unique to vehicle-mounted servers, and it is present in the state space of other computing nodes. Both are 1. =1 indicates that the node provides computing power services for the vehicle-mounted computing network. express arrive The set of signal-to-noise ratios between them.

[0278] (2) Action Space: In order to minimize task processing time and ensure the stability of the on-board computing network, it is necessary to adjust the action variables according to the dynamic state space. Therefore, the task unloading vector and It is a major component of the action space

[0279]

[0280] It is worth noting that when selecting computing nodes for task offloading, only state vectors can be selected. Nodes with a value of 1 provide services.

[0281] (3) Reward function: In order to optimize the final problem, M users should cooperate. In order to minimize the processing time in the optimization problem and ensure the queue stability of the vehicle computing network system, the reward function of user reinforcement learning is set to the negative value of the optimization problem P4.

[0282]

[0283] In the constructed multi-agent model, the given state space and action space are high-dimensional, and the state space suffers from dimensionality changes due to the vehicle's identity switching mechanism. To address these issues, this paper combines a fixed state space dimension with an actor neural network to solve the constructed multi-agent problem. By restricting actions in the action space to only selecting nodes currently providing services for offloading, the dimensionality fluctuations in the state space caused by the vehicle server's identity switching and the potential convergence failure of the multi-agent reinforcement learning algorithm can be avoided. In the actor neural network, the parameters of the orator network are defined as... During optimization, the multi-agent reinforcement learning algorithm will search for the optimal network parameters to maximize long-term reward. Long-term reward can be expressed as:

[0284]

[0285] In a multi-agent reinforcement learning architecture, multiple agents operate using the Policy Probability Optimization (PPO) algorithm based on policy gradients and confidence intervals. Since the action space is discrete (i.e., selecting the target node for task offloading), each agent's policy network uses a softmax function to output the probability distribution of each available action. The agent then samples actions based on this probability distribution.

[0286] In multi-agent reinforcement learning, the agent operation process includes two phases: centralized training and distributed execution. For the agent of a drone, assuming... As a network parameter for critics, the critic network is based on... and Calculate the action value function: .

[0287] The optimization goal of the critic network is to make its estimate as close as possible to the actual advantage estimate. To achieve this goal, this gap is measured and optimized by minimizing a loss function:

[0288]

[0289] in The advantage function, representing the advantage of the current decision compared to other feasible decisions, is expressed as:

[0290]

[0291] in, and These represent the discount factor and the single-step time error, respectively. Represented as:

[0292]

[0293] During the distributed execution phase, the actor network will update the parameters, and then, based on the centralized calculations by the critics... The function results and local observation output actions. The loss function for the action space in the actor network is defined as:

[0294]

[0295] in, Represents parameters in the action space The scope of the update; Represents a confidence interval The parameters, and ; Clipping function Will The update range is limited to the confidence interval to ensure network stability.

[0296] The specific process is as follows:

[0297] (1) Initialize actor network parameters and critic network parameters Initialization state And the experience replay pool D and the vehicle state set ;

[0298] (2) Execute the identity transformation mechanism update, update ,renew

[0299] (3) The latest status of each intelligent agent's observation system Execute actions

[0300] (4) Receive a reward and , sample Store in D

[0301] (5) Take a batch of data from D by minimizing Update critic network parameters By minimizing Update actor network parameters

[0302] (6) Minimize the final problem through multiple iterations.

[0303] Example 3

[0304] This embodiment provides a vehicle-mounted computing power network task offloading system based on dynamic identity transformation, used to implement the vehicle-mounted computing power network task offloading method based on dynamic identity transformation described in Embodiment 1 or 2, such as... Figure 3 As shown, it includes:

[0305] The vehicle-mounted computing network construction module is used to build vehicle-mounted computing networks.

[0306] The vehicle identity dynamic conversion module is used to establish an identity conversion mechanism, and to dynamically convert the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions according to the identity conversion mechanism, so as to obtain the vehicles after identity conversion.

[0307] The priority compensation module is used to calculate the priority factor of the vehicle computing power network, and use the priority factor of the vehicle computing power network to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network.

[0308] The time calculation module is used to calculate the communication time, calculation time, and waiting time for the priority-compensated vehicle computing network, respectively.

[0309] The optimization problem establishment module is used to establish an optimization problem for unloading on-board computing power network tasks based on the communication time, computation time, and waiting time.

[0310] The network task unloading module is used to transform the vehicle computing power network task unloading optimization problem into a final optimized problem, solve the final optimized problem, and obtain the optimal vehicle computing power network task unloading result.

[0311] The same or similar labels correspond to the same or similar parts;

[0312] The terms used to describe positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0313] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for offloading onboard computing power network tasks based on dynamic identity conversion, characterized in that, include: Constructing an in-vehicle computing network; An identity conversion mechanism is established. Based on this mechanism, vehicles in the vehicular computing network that meet preset identity conversion conditions are dynamically converted to obtain the converted vehicles. This includes: classifying the vehicular server's identity as a task vehicle or a service vehicle; when a task vehicle completes a new task and its CPU utilization is less than a preset first threshold, the task vehicle's identity is converted to a service vehicle; when a service vehicle completes a new task and its CPU utilization is greater than a preset second threshold, the service vehicle's identity is converted to a task vehicle. An identity interruption time is generated when the identity conversion mechanism is running. : in, It is a task The task was returned to the set of computing node devices in the computing network due to the identity transformation mechanism. It is the communication time of the task. The task is on the equipment Waiting time in the task queue; the optimization goal of task unloading is to minimize the total system task processing time, including task communication time, task computation time, and task waiting time. Calculate the vehicle-mounted computing power network priority factor, and use the vehicle-mounted computing power network priority factor to provide priority compensation to the vehicle after identity transformation, to obtain a priority-compensated vehicle-mounted computing power network, including: The priority factors for the vehicle-mounted computing network are as follows: in, It is a source factor for the task. It is a mission safety score. It is a collection of tasks. It is a set of computing power nodes. It refers to the data size of the task; in, It is the time when the task arrives at the computing node. This is the maximum acceptable delay for the task; For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated respectively. Based on the communication time, computation time, and waiting time, an optimization problem for task offloading in vehicle computing power networks is established. The problem of optimizing the offloading of onboard computing power network tasks is transformed to obtain the final optimization problem, including: The task offloading optimization problem of the vehicle-mounted computing power network is transformed using the Lyapunov optimization method, and the expression of the final optimization problem is as follows: in, The number of tasks arriving in the computing node queue. The task processing capacity of the computing node queue. It is the Lyapunov weight. This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot; The final optimization problem is solved to obtain the optimal on-board computing power network task unloading result.

2. The method for offloading onboard computing power network tasks based on dynamic identity conversion according to claim 1, characterized in that, The vehicle-mounted computing network includes M user nodes and N computing nodes; The user node consists of a personal mobile device and an in-vehicle server; The computing nodes consist of personal mobile devices, vehicle-mounted servers, and edge servers; Define the user node set as The computing node set is defined as , in, For personal mobile devices, For vehicle-mounted servers, To serve vehicle aggregation, For edge servers, t represents time slots.

3. The method for offloading onboard computing power network tasks based on dynamic identity conversion according to claim 2, characterized in that, For the priority-compensated vehicle computing network, the communication time, computing time, and waiting time are calculated using preset communication models, computing models, and queuing models, respectively.

4. The method for offloading onboard computing power network tasks based on dynamic identity conversion according to claim 3, characterized in that, The communication model includes a first transmission time communication model from a personal mobile device to a computing node and a second transmission time communication model from an in-vehicle server to a computing node. The computing model includes a personal mobile device computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing time computing model; the queuing model is a waiting time queuing model. The communication time includes the first transmission time from the personal mobile device to the computing node and the second transmission time from the vehicle server to the computing node. The computation time includes the computation time of personal mobile devices, the computation time of in-vehicle servers, and the computation time of edge servers. The calculation expression for the first transmission time communication model is: in, The size of task data for a personal mobile device. The transmission rate from personal mobile devices to computing nodes; The calculation expression for the second transmission time communication model is: in, The size of the task data for the vehicle-mounted server. This refers to the transmission rate from the vehicle-mounted server to the computing node. The calculation expression for the personal mobile device computing time calculation model is as follows: in, This indicates the size of the task data to be offloaded from the vehicle server to the personal mobile device. This indicates the size of the task data that will be unloaded from the personal mobile device to the personal mobile device. This indicates the number of CPU cycles required for a personal mobile device to complete a task. This indicates the CPU frequency of the computing power node on a personal mobile device. The calculation expression for the vehicle-mounted server computing time calculation model is as follows: in, This indicates the size of the task data to be offloaded from the vehicle server to the vehicle server. This indicates the size of the task data that will be offloaded from the personal mobile device to the in-vehicle server. This indicates the number of CPU cycles required for the onboard server to complete a task. This indicates the CPU frequency of the onboard server computing node; The calculation expression for the edge server computing time calculation model is as follows: in, This indicates the size of the task data to be offloaded from the vehicle server to the edge server. This indicates the size of the task data that is offloaded from a personal mobile device to an edge server. This indicates the number of CPU cycles required for an edge server to complete a task. This indicates the CPU frequency of the edge server computing nodes; The expression for the waiting time queuing model is: in, Total waiting time for a task on a single device.

5. The method for offloading onboard computing power network tasks based on dynamic identity conversion according to claim 4, characterized in that, The expression for the on-board computing power network task offloading optimization problem is as follows: in, This represents the total processing time for tasks within the vehicle-mounted computing network. The set of tasks published by a personal mobile device in time slot t. A collection of tasks issued by the vehicle-mounted server. This represents the actual number of tasks backlogged in the computing node's task queue. For the size of the task data, For time slot t, the set of personal mobile devices that will be assigned to computing node n for computation. The set of onboard servers that determines which task to assign to computing node n for time slot t. The number of CPU cycles required for a computing node to complete a task. It is the total available computing power resources of computing nodes. For the decision variable of unloading, T is the maximum time slot.

6. The method for offloading on-board computing power network tasks based on dynamic identity conversion according to claim 5, characterized in that, Solving the final optimization problem to obtain the optimal onboard computing power network task offloading result includes: The final optimization problem is transformed into a Markov decision process problem. The Markov decision process problem is solved using a multi-agent proximal optimization method to obtain the optimal onboard computing power network task offloading result.

7. A vehicle-mounted computing power network task offloading system based on dynamic identity conversion, used to implement the vehicle-mounted computing power network task offloading method based on dynamic identity conversion as described in claims 1-6, characterized in that, include: The vehicle-mounted computing network construction module is used to build vehicle-mounted computing networks. The vehicle identity dynamic conversion module is used to establish an identity conversion mechanism and, according to the identity conversion mechanism, dynamically convert the identities of vehicles in the vehicle computing network that meet the preset identity conversion conditions to obtain the vehicles after identity conversion. The priority compensation module is used to calculate the priority factor of the vehicle computing power network, and use the priority factor of the vehicle computing power network to provide priority compensation to the vehicle after identity conversion, so as to obtain the priority-compensated vehicle computing power network. The time calculation module is used to calculate the communication time, calculation time, and waiting time for the priority-compensated vehicle computing network, respectively. The optimization problem establishment module is used to establish an optimization problem for unloading on-board computing power network tasks based on the communication time, computation time, and waiting time. The optimization problem transformation module is used to transform the on-board computing power network task offloading optimization problem into the final optimized problem. The optimization problem-solving module is used to solve the final optimization problem and obtain the optimal on-board computing power network task unloading result.

Images