Remote area internet of vehicles task offloading method and device based on high altitude platform assistance

Abstract

The present application relates to the technical field of Internet of Vehicles communication and calculation, in particular to a remote area Internet of Vehicles task offloading method and device based on high altitude platform assistance, which comprises the following steps: constructing a vehicle connection judgment model based on a high altitude platform and an edge server, obtaining the position of a vehicle, judging whether the vehicle is connected to the edge server or the high altitude platform or whether the vehicle is not connected to the edge server or the high altitude platform; constructing a communication model, a time delay model and an energy consumption model, calculating the offloading transmission rate, the total time delay of the task and the total energy consumption of the task according to the communication model, the time delay model and the energy consumption model and the connection judgment result; constructing an optimization objective function according to the time delay model and the energy consumption model; modeling the task offloading problem as a Markov decision process and training it by using a MAPPO algorithm to output a task offloading strategy, thereby providing a solution to the problem of limited task offloading and high time delay of the remote area Internet of Vehicles.

Description

Method and device for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance Technical Field

[0001] This invention relates to the field of vehicle-to-everything (V2X) communication and computing technology, specifically to a method and apparatus for offloading V2X tasks in remote areas based on a high-altitude platform. Background Technology

[0002] With the rapid development of the Internet of Vehicles (IoV) and intelligent connected vehicle technologies, the demand for efficient computing power and low-latency response during vehicle operation is increasing dramatically. Modern intelligent connected vehicles need to handle large-scale and complex computing tasks in real time, such as high-precision map updates, path planning, environmental perception, and autonomous driving decisions. These tasks place extremely high demands on computing performance and response speed, but the computing resources of onboard terminals are limited. Therefore, if vehicles rely solely on local processing, excessive computing latency may occur, negatively impacting driving safety and user experience.

[0003] In areas with well-developed ground infrastructure, such as cities, edge servers (ES) can provide convenient task offloading support for vehicles, effectively alleviating the computing pressure on in-vehicle terminals and reducing response latency. However, in remote mountainous areas, rural areas, and highways, task offloading is limited due to sparse ES deployment or poor communication conditions, leading to increased processing latency, higher vehicle energy consumption, and consequently, reduced service quality of the entire vehicle network.

[0004] To address this issue, High Altitude Platform (HAP) technology has received widespread attention in recent years. HAPs typically include tethered balloons, airships, and high-altitude drones, offering advantages such as wide coverage, flexible deployment, and strong scalability. They can supplement communication and computing resources in scenarios where ground infrastructure is insufficient. Existing research shows that HAPs can serve as mobile edge nodes, providing task offloading services for vehicles and IoT terminals, and exhibiting significant advantages in energy efficiency, computing power, and real-time performance. Furthermore, HAPs also possess unique value in emergency communication, wide-area coverage, and low-latency support.

[0005] However, most existing research focuses on the two-tier "vehicle-edge server" structure, with insufficient research on HAP-supported task offloading in vehicular networks. A few studies have analyzed the real-time performance and reliability of HAP-assisted offloading using queuing theory models, or constructed multi-layered heterogeneous offloading systems combining drones and HAP under the 6G architecture. While these studies provide useful references, most remain at the preliminary modeling stage, lacking systematic task offloading and resource coordination strategies for the complex dynamic characteristics of vehicular networks in remote areas (such as high-speed vehicle movement, channel fluctuations, and changes in some computing resources).

[0006] Therefore, this application proposes a method and apparatus for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance. In complex and dynamic environments, especially in remote areas, it enables collaborative task offloading between HAP and ES, thereby improving the real-time computing capabilities and overall service quality of V2X. Summary of the Invention

[0007] The purpose of this application is to propose a method and device for unloading vehicle-to-everything (V2X) tasks in remote areas based on a high-altitude platform to address the aforementioned technical problems.

[0008] In a first aspect, the present invention provides a method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance, comprising the following steps:

[0009] A vehicle connection judgment model is built based on the high-altitude platform and the edge server. The vehicle's location is obtained, and the vehicle's location is input into the vehicle connection judgment model to obtain the connection judgment result. The connection judgment result includes the vehicle being connected to the edge server, the vehicle being connected to the high-altitude platform, and the vehicle not being connected to the edge server or the high-altitude platform.

[0010] Construct a communication model, and calculate the offload transmission rate based on the communication model and connection judgment results;

[0011] Construct a latency model and calculate the total latency of the task based on the offload transmission rate, the latency model, and the connection judgment results;

[0012] Construct an energy consumption model and calculate the total energy consumption of the task based on the offload transmission rate, the energy consumption model, and the connection judgment results.

[0013] Construct an optimization objective function based on the time delay model and energy consumption model;

[0014] The task unloading problem is modeled as a Markov decision process and trained using the MAPPO algorithm to output a task unloading strategy.

[0015] Preferably, a vehicle connection determination model is constructed to obtain the vehicle's location. The vehicle's location is then input into the vehicle connection determination model to obtain the connection determination result, which specifically includes:

[0016] Let the set of vehicles be denoted as The set of edge servers is denoted as The position of vehicle k in time slot t is obtained, and the three-dimensional distance between vehicle k and each edge server is calculated using the following formula:

[0017] ;

[0018] in, This represents the three-dimensional distance between vehicle k and the s-th edge server. This indicates the position of vehicle k in time slot t; Represents the three-dimensional coordinates of the s-th edge server, when At that time, the s-th edge server is located on the ground;

[0019] according to Determine whether vehicle k is within the coverage area of ​​any edge server and whether the channel quality meets a preset threshold. If so, connect vehicle k to the edge server; otherwise, calculate the three-dimensional distance between vehicle k and the aerial platform, as shown in the following formula:

[0020] ;

[0021] in, This represents the three-dimensional distance between vehicle k and the aerial platform. Represents the three-dimensional coordinates of the aerial platform;

[0022] according to Determine if vehicle k is within the coverage area of ​​the high-altitude platform. If so, connect vehicle k to the high-altitude platform; otherwise, do not connect vehicle k to the high-altitude platform or edge server.

[0023] Preferably, the processing procedure of the communication model specifically includes:

[0024] When a vehicle connects to the edge server, the communication link gain and communication rate between the vehicle and the edge server are calculated using the following formula:

[0025] ;

[0026] ;

[0027] in, This represents the communication link gain between vehicle k and the s-th edge server. For reference channel gain, This refers to the edge server link path loss index. This represents the communication rate between vehicle k and the s-th edge server. Let be the channel bandwidth of the s-th edge server; This represents the transmission power between vehicle k and the s-th edge server. Let be the noise power at the receiver of the s-th edge server;

[0028] At this point, the communication rate between vehicle k and the s-th edge server will be... As the offloading transmission rate;

[0029] When the vehicle connects to the aerial platform, the communication link gain and communication rate between the vehicle and the aerial platform are calculated using the following formula:

[0030] ;

[0031] ;

[0032] in, This represents the communication link gain between vehicle k and the aerial platform. The link path loss index for high-altitude platforms; This represents the communication rate between vehicle k and the aerial platform. For the channel bandwidth of the high-altitude platform; This represents the transmission power between vehicle k and the aerial platform. This refers to the noise power at the high-altitude platform receiver.

[0033] At this point, the communication rate between vehicle k and the aerial platform will be... As the offloading transmission rate.

[0034] As a preferred option, the processing procedure for the time delay model specifically includes:

[0035] The tasks generated by the vehicle are recorded as follows ,in, This indicates the size of the task input data generated by the vehicle; Indicates the number of CPU cycles required to process a unit of data; This indicates the maximum tolerable completion time for the task;

[0036] The local latency of the vehicle is calculated based on the tasks generated by the vehicle, using the following formula:

[0037] ;

[0038] in, Indicates the vehicle's local latency. Indicates in time slot The proportion of vehicle tasks that are transmitted to an aerial platform or edge server for processing is unloaded. ; This refers to the vehicle's local CPU frequency.

[0039] When a vehicle connects to an edge server, the unloading latency includes the transmission latency of the task input data from the vehicle's unloading task to the edge server and the computation latency of the edge server processing the unloading task. The unloading latency is calculated based on the size of the unloading task input data and the communication rate between vehicle k and the s-th edge server, as shown in the following formula:

[0040] ;

[0041] in, This indicates the offloading latency of the vehicle when it connects to the edge server; The computing frequency allocated to vehicles by the edge server;

[0042] When a vehicle connects to an aerial platform, the unloading delay includes the transmission delay of the task input data from the vehicle to the aerial platform and the calculation delay of the aerial platform in processing the unloading task. The unloading delay is calculated based on the size of the unloading task input data and the communication rate between vehicle k and the s-th aerial platform, as shown in the following formula:

[0043] ;

[0044] in, This indicates the unloading delay when the vehicle connects to the aerial platform; The computing frequency allocated to the vehicle for the aerial platform;

[0045] The total latency of the task is calculated based on the vehicle's local latency and unloading latency, using the following formula:

[0046] ;

[0047] in, Indicates the total latency of the task. This indicates taking the maximum value.

[0048] As a preferred option, the energy consumption model processing procedure specifically includes:

[0049] The vehicle's local computing energy consumption is calculated based on the tasks generated by the vehicle, using the following formula:

[0050] ;

[0051] in, This indicates the vehicle's locally calculated energy consumption. It is the power factor consumed by local calculation; Indicates the local CPU frequency of the vehicle The square of;

[0052] When a vehicle connects to the edge server, it is based on the time slot. The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's tasks, which are transmitted to the edge server for processing. The calculation formula is shown in the following formula:

[0053] ;

[0054] in, This indicates the energy consumption for offloading tasks from the vehicle to the edge server.

[0055] When the vehicle connects to the aerial platform, according to the time slot The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's task transmission to the high-altitude platform. The calculation formula is shown in the following formula:

[0056] ;

[0057] in, This indicates the energy consumption for unloading the vehicle's mission to the high-altitude platform.

[0058] The total energy consumption for the vehicle's locally calculated energy consumption and unloading energy consumption calculation task is calculated using the following formula:

[0059] ;

[0060] in, This indicates the total energy consumption of the task; and It is an indicator function that is used when a vehicle connects to the edge server. ,otherwise When the vehicle is connected to the aerial platform, ,otherwise .

[0061] As a preferred option, the objective function is shown in the following equation:

[0062] ;

[0063] in, This indicates the task's preference for latency and energy consumption. ; Indicates transmission power. .

[0064] As a preferred approach, the task unloading problem is modeled as a Markov decision process, the modeling process specifically including:

[0065] Define the state space As shown in the following formula:

[0066] ;

[0067] Define action space As shown in the following formula:

[0068] , , , ;

[0069] in, Indicates the selected edge server number; The unloading strategy's action space includes three types of decisions: the vehicle chooses to unload the task locally, the vehicle chooses to unload the task to an edge server, or the vehicle chooses to unload the task to an aerial platform; when When the vehicle chooses to execute the task locally; At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to the specified edge server for execution; when At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload and execute on the high-altitude platform;

[0070] Define reward function The goal is to maximize task completion efficiency while minimizing latency and energy consumption. The expression for the reward function is shown below:

[0071] ;

[0072] in, It is a task Weighting of preferences for latency and energy consumption ; This represents the total delay for vehicle k to complete its computational task within time slot t; This represents the total energy consumed by vehicle k in completing its computational tasks within time slot t. It is a penalty factor used to penalize offloading strategies that violate latency or energy consumption constraints; These are fixed penalty items used to indicate additional penalties for violating constraints.

[0073] Secondly, the present invention provides a remote area vehicle-to-everything (V2X) task offloading device based on a high-altitude platform, comprising:

[0074] The connection judgment module is configured to build a vehicle connection judgment model based on the high-altitude platform and the edge server, obtain the vehicle's location, input the vehicle's location into the vehicle connection judgment model, and obtain the connection judgment result. The connection judgment result includes the vehicle being connected to the edge server, the vehicle being connected to the high-altitude platform, and the vehicle not being connected to the edge server or the high-altitude platform.

[0075] The unloaded transmission rate calculation module is configured to build a communication model and calculate the unloaded transmission rate based on the communication model and connection judgment results.

[0076] The unloading latency calculation module is configured to build a latency model and calculate the total latency of the task based on the unloading transmission rate, the latency model, and the connection judgment results.

[0077] The unloaded energy consumption calculation module is configured to build an energy consumption model and calculate the total energy consumption of the task based on the unloaded transmission rate, the energy consumption model, and the connection judgment results.

[0078] The objective function construction module is configured to construct an optimization objective function based on the time delay model and the energy consumption model.

[0079] The strategy training module is configured to model the task unloading problem as a Markov decision process and train it using the MAPPO algorithm to output a task unloading strategy.

[0080] Thirdly, the present invention provides an electronic device including one or more processors; and a memory for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any implementation of the first aspect.

[0081] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the implementations of the first aspect.

[0082] Compared with the prior art, the present invention has the following beneficial effects:

[0083] (1) The task offloading method for vehicle networking in remote areas based on high-altitude platform assistance proposed in this application improves the computing efficiency of vehicle networking system in remote areas with weak communication infrastructure by designing dynamic task allocation between the vehicle, the edge server and the high-altitude platform, effectively reducing task processing latency and system energy consumption, and improving the service capability and coverage performance of vehicle networking system through the assistance of high-altitude platform, solving the problem of incomplete ground network coverage in remote areas, and improving the feasibility and efficiency of task offloading.

[0084] (2) The remote area vehicle network task unloading method proposed in this application combines the latency model and the energy consumption model to design and optimize the objective function. In remote areas with poor network conditions, it can intelligently select the optimal unloading path according to the real-time location of the vehicle, the network status and the computing resources, thereby reducing the vehicle's computing burden. Through the collaborative work of the high-altitude platform and the edge server, the total latency and total energy consumption of the task are significantly reduced.

[0085] (3) The task offloading method for remote area vehicle networking based on high-altitude platform assistance proposed in this application realizes the adaptive adjustment of task offloading strategy through MAPPO algorithm, which can cope with the complex and ever-changing network conditions and task load in remote area vehicle networking environment. It has strong flexibility and scalability and strong adaptability. Attached Figure Description

[0086] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0087] Figure 1 is a flowchart illustrating a remote area vehicle-to-everything (V2X) task offloading method based on a high-altitude platform assisted by an embodiment of this application.

[0088] Figure 2 is a schematic diagram of the heterogeneous edge computing system architecture of the remote area vehicle-to-everything (V2X) task offloading method based on high-altitude platform assistance according to an embodiment of this application.

[0089] Figure 3 is a flowchart illustrating the MAPPO algorithm in the task unloading strategy training according to an embodiment of this application.

[0090] Figure 4 is a schematic diagram of a remote area vehicle-to-everything (V2X) task unloading device based on a high-altitude platform assisted by an embodiment of this application;

[0091] Figure 5 is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present invention.

[0092] Reference numerals: 11. Vehicle; 12. Aerial platform; 13. Edge server. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0094] Figure 1 illustrates a method for offloading vehicle-to-everything (V2X) tasks in remote areas based on a high-altitude platform-assisted approach, according to an embodiment of this application, comprising the following steps:

[0095] S1. Based on the high-altitude platform and the edge server, a vehicle connection judgment model is constructed to obtain the vehicle's location. The vehicle's location is then input into the vehicle connection judgment model to obtain the connection judgment result. The connection judgment result includes whether the vehicle is connected to the edge server, the vehicle is connected to the high-altitude platform, or the vehicle is not connected to either the edge server or the high-altitude platform.

[0096] In a specific embodiment, a vehicle connection determination model is constructed, the vehicle's location is obtained, the vehicle's location is input into the vehicle connection determination model, and a connection determination result is obtained, specifically including:

[0097] Let the set of vehicles be denoted as The set of edge servers is denoted as The position of vehicle k in time slot t is obtained, and the three-dimensional distance between vehicle k and each edge server is calculated using the following formula:

[0098] ;

[0099] in, This represents the three-dimensional distance between vehicle k and the s-th edge server. This indicates the position of vehicle k in time slot t; Represents the three-dimensional coordinates of the s-th edge server, when At that time, the s-th edge server is located on the ground;

[0100] according to Determine whether vehicle k is within the coverage area of ​​any edge server and whether the channel quality meets a preset threshold. If so, connect vehicle k to the edge server; otherwise, calculate the three-dimensional distance between vehicle k and the aerial platform, as shown in the following formula:

[0101] ;

[0102] in, This represents the three-dimensional distance between vehicle k and the aerial platform. Represents the three-dimensional coordinates of the aerial platform;

[0103] according to Determine if vehicle k is within the coverage area of ​​the high-altitude platform. If so, connect vehicle k to the high-altitude platform; otherwise, do not connect vehicle k to the high-altitude platform or edge server.

[0104] Specifically, in remote areas, terrestrial communication infrastructure may have insufficient coverage and poor network quality.

[0105] Vehicle terminal: As the generator of computing tasks, when in remote areas, the vehicle terminal generates computing tasks and decides whether to offload the tasks to local computing, edge servers or high-altitude platforms based on the characteristics of the tasks, the location and communication conditions.

[0106] Edge Servers (ES): When there is good communication signal on the ground, tasks can be offloaded to edge servers to provide low-latency computing services, making them suitable for urban or relatively dense areas. However, in remote areas, edge server deployment is sparse, and tasks can only be offloaded to edge servers when vehicles enter their service coverage area.

[0107] High Altitude Platform (HAP): This includes vehicles such as hot air balloons, drones, or airships. Due to the instability of ground networks in remote areas, HAPs provide additional computing and communication support. HAPs can provide wide-coverage, low-latency communication services in remote areas, filling the gaps in ground network coverage. Vehicles can only offload tasks to HAPs when they are outside the service range of edge servers.

[0108] Referring to Figure 2, the embodiments of this application construct a heterogeneous edge computing system based on the connection between the local vehicle 11, the high-altitude platform 12, and the edge server 13, enabling air-ground collaboration. This system employs the high-altitude platform-assisted task offloading method for vehicle-to-everything (V2X) in remote areas proposed in this application. This method can not only operate in urban environments but also address the challenges of unstable networks and insufficient resource allocation in remote areas. For scenarios where ground communication infrastructure is scarce and coverage is insufficient in remote areas, the vehicle 11 needs to determine the connectable access nodes based on its location information before task offloading, thereby dynamically deciding whether to connect to the edge server 13 or the high-altitude platform 12. Only if the vehicle is within the service range of the edge server 13 can the task utilize the offloading ratio. When the vehicle is within range of no edge server 13, the task is only allowed to be unloaded locally, therefore no task unloading occurs. When the vehicle is within range of both edge server 13 and high-altitude platform 12, or within range of edge server 13 but not within range of high-altitude platform 12, or not within range of edge server 13 but only within range of high-altitude platform 12, the task unloading strategy will select a suitable edge server 13 to unload the task or process the task locally. If the task unloading strategy determines that local unloading is optimal, the unloading ratio will be adjusted accordingly. .

[0109] The trained task offloading strategy is deployed to vehicle 11 terminals, edge servers 13, or high-altitude platform 12 nodes. Based on real-time communication status and computing resource availability, the task offloading path is determined, enabling the provision of computing and communication support to vehicles 11 in remote areas. In scalable multi-high-altitude platform 12 scenarios, where vehicles 11 are outside the coverage area of ​​ground edge servers, a connection to a high-altitude platform 12 node is selected based on an optimal access strategy. This optimal access strategy combines the task offloading strategy with the current communication and computing resource status to select the edge server that best meets the constraints and offers the highest utility. In the high-altitude platform-assisted task offloading method for remote area vehicle-to-everything (V2X) proposed in this application, only one available edge server 13 needs to be considered.

[0110] S2, construct the communication model, and calculate the offload transmission rate based on the communication model and connection judgment results.

[0111] In a specific embodiment, the processing procedure of the communication model includes:

[0112] When a vehicle connects to the edge server, the communication link gain and communication rate between the vehicle and the edge server are calculated using the following formula:

[0113] ;

[0114] ;

[0115] in, This represents the communication link gain between vehicle k and the s-th edge server. For reference channel gain, This refers to the edge server link path loss index. This represents the communication rate between vehicle k and the s-th edge server. Let be the channel bandwidth of the s-th edge server; This represents the transmission power between vehicle k and the s-th edge server. Let be the noise power at the receiver of the s-th edge server;

[0116] At this point, the communication rate between vehicle k and the s-th edge server will be... As the offloading transmission rate;

[0117] When the vehicle connects to the aerial platform, the communication link gain and communication rate between the vehicle and the aerial platform are calculated using the following formula:

[0118] ;

[0119] ;

[0120] in, This represents the communication link gain between vehicle k and the aerial platform. The link path loss index for high-altitude platforms; This represents the communication rate between vehicle k and the aerial platform. For the channel bandwidth of the high-altitude platform; This represents the transmission power between vehicle k and the aerial platform. This refers to the noise power at the high-altitude platform receiver.

[0121] At this point, the communication rate between vehicle k and the aerial platform will be... As the offloading transmission rate.

[0122] Specifically, to evaluate communication rates and determine the latency and energy consumption of task offloading, the vehicle-to-everything (V2X) system uses two communication transmission links: one between the vehicle and the aerial platform, and the other between the vehicle and the edge server. When the vehicle is in a time slot... If the vehicle and at least one edge server are within the coverage area of ​​the edge server in three dimensions and the channel quality threshold is met, then the connection to that edge server is prioritized, and then... Calculate the offloading transfer rate; if the vehicle is outside the coverage radius of the edge server, it will connect to the aerial platform for task offloading. As the offloading transmission rate.

[0123] S3: Construct a latency model and calculate the total latency of the task based on the offload transmission rate, the latency model, and the connection judgment results.

[0124] In a specific embodiment, the processing procedure of the latency model includes:

[0125] The tasks generated by the vehicle are recorded as follows ,in, This indicates the size of the task input data generated by the vehicle; Indicates the number of CPU cycles required to process a unit of data; This indicates the maximum tolerable completion time for the task;

[0126] The local latency of the vehicle is calculated based on the tasks generated by the vehicle, using the following formula:

[0127] ;

[0128] in, Indicates the vehicle's local latency. Indicates in time slot The proportion of vehicle tasks that are transmitted to an aerial platform or edge server for processing is unloaded. ; This refers to the vehicle's local CPU frequency.

[0129] When a vehicle connects to an edge server, the unloading latency includes the transmission latency of the task input data from the vehicle's unloading task to the edge server and the computation latency of the edge server processing the unloading task. The unloading latency is calculated based on the size of the unloading task input data and the communication rate between vehicle k and the s-th edge server, as shown in the following formula:

[0130] ;

[0131] in, This indicates the offloading latency of the vehicle when it connects to the edge server; The computing frequency allocated to vehicles by the edge server;

[0132] When a vehicle connects to an aerial platform, the unloading delay includes the transmission delay of the task input data from the vehicle to the aerial platform and the calculation delay of the aerial platform in processing the unloading task. The unloading delay is calculated based on the size of the unloading task input data and the communication rate between vehicle k and the s-th aerial platform, as shown in the following formula:

[0133] ;

[0134] in, This indicates the unloading delay when the vehicle connects to the aerial platform; The computing frequency allocated to the vehicle for the aerial platform;

[0135] The total latency of the task is calculated based on the vehicle's local latency and unloading latency, using the following formula:

[0136] ;

[0137] in, Indicates the total latency of the task. This indicates taking the maximum value.

[0138] Specifically, due to the weak terrestrial communication and computing infrastructure in remote areas and the significant fluctuations in network conditions, tasks generated by vehicle terminals often rely more heavily on real-time computing and offloading decisions, with latency being a key factor in determining these decisions. Therefore, the embodiments of this application establish a task model. This represents the task generated by the k-th vehicle, where each task is represented by a triple: Used to characterize vehicles in time slots The characteristics of the resulting computational tasks; Used to describe the computing resources required when processing a task; This describes the allocation of vehicle tasks between the local machine and the high-altitude platform, or between the local machine and the edge server. When, it means that all tasks of the vehicle are executed on the local device; when When this occurs, it indicates that the vehicle's tasks have been completely offloaded to the high-altitude platform or edge server; when At this time, the vehicle's tasks employ a partial offloading strategy, meaning that a portion of the computation is processed locally, while the remaining portion is offloaded to the appropriate high-altitude platform or edge server for processing based on the connection determination result. In the embodiments of this application, the offloading ratio is determined accordingly. Determine the portion of the vehicle's task to be offloaded to an aerial platform or edge service. This allows us to determine the portion of the vehicle's task that is not unloaded and is being processed locally. The local latency of the vehicle is calculated as the ratio of the computation time required for the unloaded task portion to the vehicle's CPU frequency. A latency model is then constructed based on the task model and communication model.

[0139] The total latency of a task is determined by the maximum of the local latency and the unloading latency of the task being processed on the vehicle's local equipment. This ensures that the total time to complete the task is equal to the maximum processing time of the two parts: the local latency and the unloading latency.

[0140] The proposed method for offloading vehicle-to-everything (V2X) tasks in remote areas, based on high-altitude platform assistance, improves the computational efficiency of V2X systems in remote areas with weak communication infrastructure by designing dynamic task allocation between the vehicle, the edge server, and the high-altitude platform. This effectively reduces task processing latency, achieves efficient computational task processing, and enhances the service capabilities and coverage performance of the V2X system through the assistance of the high-altitude platform. This solves the problem of incomplete ground network coverage in remote areas and improves the feasibility and efficiency of task offloading.

[0141] S4, construct an energy consumption model, and calculate the total energy consumption of the task based on the offload transmission rate, the energy consumption model, and the connection judgment results.

[0142] In a specific embodiment, the energy consumption model processing procedure specifically includes:

[0143] The vehicle's local computing energy consumption is calculated based on the tasks generated by the vehicle, using the following formula:

[0144] ;

[0145] in, This indicates the vehicle's locally calculated energy consumption. It is the power factor consumed by local calculation; Indicates the local CPU frequency of the vehicle The square of;

[0146] When a vehicle connects to the edge server, it is based on the time slot. The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's tasks, which are transmitted to the edge server for processing. The calculation formula is shown in the following formula:

[0147] ;

[0148] in, This indicates the energy consumption for offloading tasks from the vehicle to the edge server.

[0149] When the vehicle connects to the aerial platform, according to the time slot The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's task transmission to the high-altitude platform. The calculation formula is shown in the following formula:

[0150] ;

[0151] in, This indicates the energy consumption for unloading the vehicle's mission to the high-altitude platform.

[0152] The total energy consumption for the vehicle's locally calculated energy consumption and unloading energy consumption calculation task is calculated using the following formula:

[0153] ;

[0154] in, This indicates the total energy consumption of the task; and It is an indicator function that is used when a vehicle connects to the edge server. ,otherwise When the vehicle is connected to the aerial platform, ,otherwise .

[0155] Specifically, in remote areas, energy efficiency optimization is crucial for system performance. The vehicle's total energy consumption consists of local computing energy consumption and offloading energy consumption. Specifically, the vehicle's local computing energy consumption is determined by the square of the vehicle's local CPU frequency, the amount of task data, and the computational load. It is directly proportional to the vehicle's local CPU frequency, the size of the task input data, and the required computing resources, reflecting the energy consumption of the vehicle when computing tasks locally. Offloading energy consumption refers to the energy consumption for data transmission between the vehicle and the edge server or between the vehicle and the aerial platform. The total energy consumption of the task consists of both the vehicle's local computing energy consumption and offloading energy consumption. Depending on the vehicle's offloading option—connecting to the edge server or connecting to the aerial platform—the total energy consumption of the task will differ. The total energy consumption of the task is either the sum of the vehicle's local computing energy consumption and the energy consumption when the task is offloaded to the edge server, or the sum of the vehicle's local computing energy consumption and the energy consumption when the task is offloaded to the aerial platform. Therefore, the choice of vehicle determines the allocation of energy consumption during task completion.

[0156] S5. Construct an optimization objective function based on the time delay model and energy consumption model.

[0157] In a specific embodiment, the objective function for optimization is shown in the following equation:

[0158] ;

[0159] in, This indicates the task's preference for latency and energy consumption. ; Indicates transmission power. .

[0160] Specifically, the optimization objective of this invention is to minimize the total latency and total energy consumption of the task, which is particularly important in remote areas. The task offloading decision is driven by an optimization objective function, which aims to minimize the total latency and total energy consumption of the task. Specifically, the optimization objective function is a weighted sum of the total latency and total energy consumption of the task, in order to minimize system energy consumption while ensuring timely task completion. The remote area vehicle-to-everything (V2X) task offloading method proposed in this application, based on a high-altitude platform, combines a latency model and an energy consumption model to design the optimization objective function. In remote areas with poor network conditions, it can intelligently select the optimal offloading path based on the vehicle's real-time location, network status, and computing resources, reducing the vehicle's computational burden. Furthermore, through the collaborative work of the high-altitude platform and edge servers, it significantly reduces the total latency and total energy consumption of the task.

[0161] S6 models the task unloading problem as a Markov decision process and trains it using the MAPPO algorithm to output a task unloading strategy.

[0162] In a specific embodiment, the task unloading problem is modeled as a Markov decision process, and the modeling process specifically includes:

[0163] Define the state space As shown in the following formula:

[0164] ;

[0165] Define action space As shown in the following formula:

[0166] , , , ;

[0167] in, Indicates the selected edge server number. The unloading strategy's action space includes three types of decisions: the vehicle chooses to unload the task locally, the vehicle chooses to unload the task to an edge server, or the vehicle chooses to unload the task to an aerial platform; when When the vehicle chooses to execute the task locally; At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to the specified edge server for execution; when At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload and execute on the high-altitude platform;

[0168] Define reward function The goal is to maximize task completion efficiency while minimizing latency and energy consumption. The expression for the reward function is shown below:

[0169] ;

[0170] in, It is a task Weighting of preferences for latency and energy consumption ; This represents the total delay for vehicle k to complete its computational task within time slot t; This represents the total energy consumed by vehicle k in completing its computational tasks within time slot t. It is a penalty factor used to penalize offloading strategies that violate latency or energy consumption constraints; These are fixed penalty items used to indicate additional penalties for violating constraints.

[0171] Specifically, by modeling the task unloading problem as a Markov decision process, the state space... It describes the vehicle's current environmental information and motion space. Defines the vehicle's selectable unloading strategy and reward function. To measure the quality of unloading decisions, the vehicle can adjust its actions based on its current state. Choose the appropriate uninstallation action. This maximizes long-term cumulative rewards while minimizing task latency and energy consumption.

[0172] In the embodiments of this application, the task offloading strategy is trained using a reinforcement learning-based optimization method, specifically using the Multi-Agent Proximal Policy Optimization (MAPPO) algorithm. The task offloading problem is modeled as an MDP, based on the vehicle's real-time state information, including location, task latency requirements, and computing power. The MAPPO algorithm outputs continuous actions through a policy network, obtaining the offloading ratio and power allocation corresponding to each action. The strategy is then updated based on latency and energy consumption rewards from environmental feedback, thereby achieving dynamic optimization of the offloading ratio and transmission power, thus minimizing task latency and energy consumption. The task offloading method for remote area vehicle-to-everything (V2X) networks proposed in this application, based on a high-altitude platform assistance, achieves adaptive adjustment of the task offloading strategy through the MAPPO algorithm. This enables it to cope with complex and variable network conditions and task loads in remote V2X environments, exhibiting strong flexibility and scalability.

[0173] Referring to Figure 3, the training process using the MAPPO algorithm involves experience replay and policy updates. The MAPPO algorithm continuously optimizes the task unloading strategy and makes optimal decisions based on real-time environmental changes. The training process specifically includes: initializing the parameters of the task unloading strategy network; interacting with the environment based on the current task unloading strategy to obtain the system state, unloading actions, and corresponding rewards; storing the generated states, actions, rewards, and next state sequences in the experience replay buffer; sampling data from the buffer; calculating the loss functions of the strategy network and value network based on the sampled data; updating the parameters of the Actor and Critic networks based on the loss functions; and after completing one round of policy updates, determining whether the current task unloading strategy has converged. If so, the optimal task unloading strategy is output; otherwise, the process returns to the steps of interacting with the environment based on the task unloading strategy. The unloading decision involves deploying the trained task unloading strategy to vehicle terminals, edge servers, or high-altitude platform nodes. Based on real-time communication status and computing resources, the unloading path of the task is determined. The task unloading path includes vehicle local computing, edge server computing, and high-altitude platform computing.

[0174] The steps S1-S6 above do not necessarily represent the order of the steps, but are represented by step symbols. The order of the steps can be adjusted.

[0175] Referring further to Figure 4, as an implementation of the methods shown in the above figures, this application provides an embodiment of a remote area vehicle networking task offloading device based on a high-altitude platform. This device embodiment corresponds to the method embodiment shown in Figure 1, and the device can be specifically applied to various electronic devices.

[0176] This application provides a remote area vehicle-to-everything (V2X) task offloading device based on a high-altitude platform, comprising:

[0177] The connection judgment module 1 is configured to build a vehicle connection judgment model based on the high-altitude platform and the edge server, obtain the vehicle's location, input the vehicle's location into the vehicle connection judgment model, and obtain the connection judgment result. The connection judgment result includes the vehicle being connected to the edge server, the vehicle being connected to the high-altitude platform, and the vehicle not being connected to the edge server or the high-altitude platform.

[0178] The unloading transmission rate calculation module 2 is configured to construct a communication model and calculate the unloading transmission rate based on the communication model and connection judgment results.

[0179] The unloading latency calculation module 3 is configured to build a latency model and calculate the total latency of the task based on the unloading transmission rate, the latency model, and the connection judgment result.

[0180] Unload energy consumption calculation module 4, which is configured to build an energy consumption model and calculate the total energy consumption of the task based on the unloaded transmission rate, the energy consumption model and the connection judgment result.

[0181] Optimization objective function construction module 5 is configured to construct optimization objective functions based on the time delay model and energy consumption model;

[0182] The strategy training module 6 is configured to model the task unloading problem as a Markov decision process and train it using the MAPPO algorithm to output a task unloading strategy.

[0183] Figure 5 is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present invention. As shown in Figure 5, the electronic device of this embodiment includes: a processor 501 and a memory 502; wherein the memory 502 is used to store computer execution instructions; the processor 501 is used to execute the computer execution instructions stored in the memory to implement the various steps performed by the electronic device in the above embodiment. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0184] Alternatively, the memory 502 can be either standalone or integrated with the processor 501.

[0185] When the memory 502 is set up independently, the electronic device also includes a bus 503 for connecting the memory 502 and the processor 501.

[0186] This invention also provides a computer storage medium storing computer execution instructions, which, when executed by processor 501, implement the above method.

[0187] This invention also provides a computer program product, including a computer program that, when executed by a processor 501, implements the above-described method.

[0188] In the embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0189] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0190] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0191] The integrated modules implemented as software functional modules described above can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor 501 to execute some steps of the methods of the various embodiments of this application.

[0192] It should be understood that the processor 501 described above can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor, or the processor 501 can be any conventional processor 501. The steps of the method disclosed in this invention can be directly manifested as the hardware processor 501 executing the steps, or as a combination of hardware and software modules within the processor 501 executing the steps.

[0193] The memory 502 may include high-speed RAM memory, and may also include non-volatile memory NVM, such as at least one disk storage device, and may also be a USB flash drive, portable hard drive, read-only memory, disk or optical disc, etc.

[0194] Bus 503 can be an Industry Standard Architecture (ISA), a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 503 can be divided into address bus, data bus, control bus, etc. For ease of illustration, the bus 503 in the accompanying drawings of this application is not limited to only one bus 503 or one type of bus 503.

[0195] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0196] An exemplary storage medium is coupled to processor 501, enabling processor 501 to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of processor 501. Processor 501 and storage medium can reside in application-specific integrated circuits (ASICs). Alternatively, processor 501 and storage medium can exist as discrete components in an electronic device or host device.

[0197] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0198] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for unloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance, characterized in that, A heterogeneous edge computing system based on vehicles, aerial platforms, and edge servers is constructed, with vehicles connected to either edge servers or aerial platforms. The method includes the following steps: 1) Constructing a vehicle connection judgment model based on the aerial platform and edge servers; 2) Obtaining the vehicle's location; 3) Inputting the vehicle's location into the vehicle connection judgment model to obtain a connection judgment result, which includes the vehicle being connected to the edge server, the vehicle being connected to the aerial platform, and the vehicle not being connected to either the edge server or the aerial platform; 4) Constructing a communication model and calculating the offloading transmission rate based on the communication model and the connection judgment result; 5) Constructing a latency model and calculating the total latency of the task based on the offloading transmission rate, the latency model, and the connection judgment result; 6) Constructing an energy consumption model and calculating the total energy consumption of the task based on the offloading transmission rate, the energy consumption model, and the connection judgment result; 7) Constructing an optimization objective function based on the latency model and the energy consumption model; 8) Modeling the task offloading problem as a Markov decision process and training it using the MAPPO algorithm to output a task offloading strategy. The modeling process specifically includes: defining a state space. As shown in the following formula: ;in, This indicates the position of vehicle k in time slot t; This indicates the size of the task input data generated by the vehicle; Indicates the number of CPU cycles required to process a unit of data; This indicates the maximum tolerable completion time for the task; This represents the communication link gain between vehicle k and the aerial platform; Represent the communication link gain between vehicle k and the s-th edge server; define the action space. As shown in the following formula: ， ， , ;in, Indicates in time slot The proportion of vehicle tasks that are transmitted to an aerial platform or edge server for processing is unloaded. ； Indicates the selected edge server number. The set of edge servers is denoted as . ； The unloading strategy's action space includes three types of decisions: the vehicle chooses to unload the task locally, the vehicle chooses to unload the task to an edge server, or the vehicle chooses to unload the task to an aerial platform; when When the vehicle chooses to execute the task locally; At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to the specified edge server for execution; when At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to a high-altitude platform for execution; define the reward function. The goal is to maximize task completion efficiency while minimizing latency and energy consumption. The expression for the reward function is shown below: ;in, It is a task Weighting of preferences for latency and energy consumption ； This represents the total delay for vehicle k to complete its computational task within time slot t; This represents the total energy consumed by vehicle k in completing its computational tasks within time slot t. It is a penalty factor used to penalize offloading strategies that violate latency or energy consumption constraints; These are fixed penalty items used to indicate additional penalties for violating constraints.

2. The method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance as described in claim 1, characterized in that, The process of constructing a vehicle connection determination model, obtaining the vehicle locations, inputting the vehicle locations into the vehicle connection determination model, and obtaining the connection determination result specifically includes: setting the set of vehicles as denoted as . The position of vehicle k in time slot t is obtained, and the three-dimensional distance between vehicle k and each edge server is calculated using the following formula: ;in, This represents the three-dimensional distance between vehicle k and the s-th edge server. This indicates the position of vehicle k in time slot t; Represents the three-dimensional coordinates of the s-th edge server, when At that time, the s-th edge server is located on the ground; according to Determine whether vehicle k is within the coverage area of ​​any edge server and whether the channel quality meets a preset threshold. If so, connect vehicle k to the edge server; otherwise, calculate the three-dimensional distance between vehicle k and the aerial platform, as shown in the following formula: ;in, This represents the three-dimensional distance between vehicle k and the aerial platform. Represents the three-dimensional coordinates of the aerial platform; according to Determine if vehicle k is within the coverage area of ​​the high-altitude platform. If so, connect vehicle k to the high-altitude platform; otherwise, do not connect vehicle k to the high-altitude platform or edge server.

3. The method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance as described in claim 2, characterized in that, The specific processing steps of the communication model include: when the vehicle connects to the edge server, calculating the communication link gain and communication rate between the vehicle and the edge server, as shown in the following formula: ； ;in, This represents the communication link gain between vehicle k and the s-th edge server. For reference channel gain, This refers to the edge server link path loss index. This represents the communication rate between vehicle k and the s-th edge server. Let be the channel bandwidth of the s-th edge server; This represents the transmission power between vehicle k and the s-th edge server. Let be the noise power of the s-th edge server receiver; at this time, the communication rate between vehicle k and the s-th edge server will be... As the offloading transmission rate; when the vehicle is connected to the aerial platform, the communication link gain and communication rate between the vehicle and the aerial platform are calculated using the following formula: ； ;in, This represents the communication link gain between vehicle k and the aerial platform. The link path loss index for high-altitude platforms; This represents the communication rate between vehicle k and the aerial platform. For the channel bandwidth of the high-altitude platform; This represents the transmission power between vehicle k and the aerial platform. Let be the noise power at the high-altitude platform receiver; at this point, the communication rate between vehicle k and the high-altitude platform will be... As the offloading transmission rate.

4. The method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance as described in claim 3, characterized in that, The processing of the time delay model specifically includes: the task generated by the vehicle is denoted as... ,in, This indicates the size of the task input data generated by the vehicle; Indicates the number of CPU cycles required to process a unit of data; This indicates the maximum tolerable completion time of the task; the local latency of the vehicle is calculated based on the tasks generated by the vehicle, using the following formula: ;in, Indicates the vehicle's local latency. Indicates in time slot The proportion of vehicle tasks that are transmitted to an aerial platform or edge server for processing is unloaded. ； The CPU frequency is the local frequency of the vehicle. When the vehicle connects to the edge server, the unloading delay includes the transmission delay of the task input data from the vehicle unloading task to the corresponding task input data on the edge server and the calculation delay of the edge server in processing the unloading task. The unloading delay is calculated based on the size of the task input data to be unloaded and the communication rate between the vehicle k and the s-th edge server, as shown in the following formula: ;in, This indicates the offloading latency of the vehicle when it connects to the edge server; The computing frequency allocated to the vehicle by the edge server; when the vehicle connects to the aerial platform, the unloading delay includes the transmission delay of the task input data corresponding to the vehicle's unloading task to the aerial platform and the computing delay of the aerial platform in processing the unloading task. The unloading delay is calculated based on the size of the unloading task input data and the communication rate between the vehicle k and the s-th aerial platform, as shown in the following formula: ;in, This indicates the unloading delay when the vehicle connects to the aerial platform; The computing frequency allocated to the vehicle for the high-altitude platform; the total task latency is calculated based on the vehicle's local latency and unloading latency, using the following formula: ;in, Indicates the total latency of the task. This indicates taking the maximum value.

5. The method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance as described in claim 4, characterized in that, The energy consumption model processing procedure specifically includes: calculating the vehicle's local computing energy consumption based on the tasks generated by the vehicle, using the following formula: ;in, This indicates the vehicle's locally calculated energy consumption. It is the power factor consumed by local calculation; Indicates the local CPU frequency of the vehicle The square of; when a vehicle connects to the edge server, based on the time slot The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's tasks, which are transmitted to the edge server for processing. The calculation formula is shown in the following formula: ;in, This indicates the energy consumption for offloading tasks from the vehicle to the edge server; when the vehicle connects to the high-altitude platform, it is based on the time slot. The unloading energy consumption is calculated based on the unloading ratio and unloading transmission rate of the vehicle's task transmission to the high-altitude platform. The calculation formula is shown in the following formula: ;in, This represents the energy consumption for unloading the vehicle's task to the aerial platform. The total energy consumption of the task is calculated based on the vehicle's locally calculated energy consumption and the unloading energy consumption, using the following formula: ;in, This indicates the total energy consumption of the task; and It is an indicator function that is used when a vehicle connects to the edge server. ,otherwise When the vehicle is connected to the aerial platform, ,otherwise 。 6. The method for offloading vehicle-to-everything (V2X) tasks in remote areas based on high-altitude platform assistance as described in claim 5, characterized in that, The optimization objective function is shown in the following equation: ;in, This indicates the task's preference for latency and energy consumption. ； Indicates transmission power. 。 7. A remote area vehicle-to-everything (V2X) task unloading device based on a high-altitude platform, characterized in that, A heterogeneous edge computing system based on vehicles, aerial platforms, and edge servers is constructed, with vehicles connected to either the edge server or the aerial platform. The device includes: a connection determination module configured to build a vehicle connection determination model based on the aerial platform and the edge server, obtain the vehicle's location, input the vehicle's location into the vehicle connection determination model, and obtain a connection determination result, which includes whether the vehicle is connected to the edge server, connected to the aerial platform, or not connected to either the edge server or the aerial platform; and an offload transmission rate calculation module configured to build a communication model and calculate the offload transmission rate based on the communication model and the connection determination result. The system comprises the following modules: a rate-based unloading latency calculation module, configured to construct a latency model and calculate the total latency of the task based on the unloading transmission rate, the latency model, and the connection judgment result; an unloading energy consumption calculation module, configured to construct an energy consumption model and calculate the total energy consumption of the task based on the unloading transmission rate, the energy consumption model, and the connection judgment result; an optimization objective function construction module, configured to construct an optimization objective function based on the latency model and the energy consumption model; and a policy training module, configured to model the task unloading problem as a Markov decision process and train it using the MAPPO algorithm to output a task unloading policy. The modeling process specifically includes: defining the state space. As shown in the following formula: ;in, This indicates the position of vehicle k in time slot t; This indicates the size of the task input data generated by the vehicle; Indicates the number of CPU cycles required to process a unit of data; This indicates the maximum tolerable completion time for the task; This represents the communication link gain between vehicle k and the aerial platform; Represent the communication link gain between vehicle k and the s-th edge server; define the action space. As shown in the following formula: ， ， , ;in, Indicates in time slot The proportion of vehicle tasks that are transmitted to an aerial platform or edge server for processing is unloaded. ； Indicates the selected edge server number. The set of edge servers is denoted as . ； The unloading strategy's action space includes three types of decisions: the vehicle chooses to unload the task locally, the vehicle chooses to unload the task to an edge server, or the vehicle chooses to unload the task to an aerial platform; when When the vehicle chooses to execute the task locally; At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to the specified edge server for execution; when At that time, the vehicle selects to perform tasks according to the unloading ratio. Unload to a high-altitude platform for execution; define the reward function. The goal is to maximize task completion efficiency while minimizing latency and energy consumption. The expression for the reward function is shown below: ;in, It is a task Weighting of preferences for latency and energy consumption ； This represents the total delay for vehicle k to complete its computational task within time slot t; This represents the total energy consumed by vehicle k in completing its computational tasks within time slot t. It is a penalty factor used to penalize offloading strategies that violate latency or energy consumption constraints; These are fixed penalty items used to indicate additional penalties for violating constraints.

8. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images