End-side cloud collaborative computing unloading method based on resource clustering

By building a vehicle resource cluster and an edge server resource pool, combining the graph neural network and service intention scoring mechanism, the dynamic allocation of computing tasks among multi-level resources is solved, and the problem that traditional computing offloading methods cannot fully utilize distributed resources is improved. Computational resource utilization and system performance.

CN119071303APending Publication Date: 2024-12-03YUNNAN COMM INVESTMENT & CONSTR GRP CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411056298.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Traditional computing unloading methods cannot fully utilize distributed computing resources in complex and changeable computing environments, resulting in waste of computing resources, increased task delays, and degradation of system performance, which cannot meet the needs of large-scale computing tasks.

Method used

By building vehicle resource clusters, edge server resource pools and cloud service computing pools, the end-edge cloud collaborative computing unloading method based on resource clustering is adopted to dynamically allocate computing tasks, use graph neural network to form vehicle resource clusters, and set up service vehicle intention functions and Communication maintains time to realize optimized scheduling of tasks among multi-level resources.

Benefits of technology

It improves the utilization rate of computing resources, reduces task processing delays, enhances the robustness and flexibility of the system, adapts to complex and changeable computing environments, and is suitable for fields such as intelligent transportation systems and industrial Internet of Things.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119071303A_ABST
    Figure CN119071303A_ABST
Patent Text Reader

Abstract

The invention discloses an end-side cloud collaborative computing unloading method based on resource clustering. The method comprises the following steps: constructing an end-side cloud collaborative computing framework; dynamically clustering to form a vehicle resource cluster; and unloading the task. Through collaborative scheduling and dynamic allocation of multi-level resources, the utilization rate of computing resources is increased, the delay of task processing is reduced, the robustness and flexibility of the system are enhanced, and the method adapts to a complex and changeable computing environment. The method has wide application prospects in the fields of intelligent traffic systems, Internet of Vehicles, industrial Internet of Things and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a computing offloading method for end-edge-cloud collaboration based on resource clustering. Background Art

[0002] With the rapid development of Internet of Things (IoT) technology and intelligent vehicle technology, more and more computing tasks need to be processed collaboratively between vehicle terminals, edge servers and cloud servers to meet the requirements of real-time, high efficiency and low latency. Traditional computing offloading methods mainly rely on the allocation of single computing resources, such as offloading computing tasks directly to cloud servers or edge servers. However, this method often fails to fully utilize distributed computing resources in the face of complex and changing computing environments, resulting in problems such as waste of computing resources, increased task delays, and decreased system performance.

[0003] In recent years, edge computing, as an emerging computing paradigm, has effectively reduced network latency and improved the real-time performance of data processing by sinking computing power to edge nodes closer to the data source. However, the computing power and resources of a single edge node are limited and cannot meet the needs of large-scale computing tasks. Therefore, how to achieve efficient computing task offloading between the end (vehicle terminal), edge (edge ​​server) and cloud (cloud server) has become a key issue that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a computing offloading method for end-edge-cloud collaboration based on resource clustering to address the above-mentioned problems that currently exist. By constructing a vehicle resource cluster, an edge server resource pool and a cloud service computing pool, the dynamic allocation of computing tasks among multi-level resources is achieved.

[0005] The technical solution of the present invention is as follows:

[0006] A computing offloading method for end-edge-cloud collaboration based on resource clustering includes the following steps:

[0007] Build a device-edge-cloud collaborative computing framework;

[0008] Dynamic clustering to form vehicle resource clusters;

[0009] Task offloading: The vehicle generates a task offloading request and determines whether the currently required program is cached on the vehicle side. If the cache exists, further check whether the current vehicle has sufficient power. If the power is sufficient, the computing task is offloaded on the vehicle side; otherwise, the computing task is partially offloaded to the vehicle resource cluster.

[0010] In the vehicle resource cluster, if there is a program cache, the sharing willingness of the vehicle resource cluster is determined; if the service vehicle is willing to provide shared resources, task offloading is performed in the vehicle resource cluster; if there is no program cache in the vehicle resource cluster, the computing task is offloaded to the edge server resource pool;

[0011] The edge server calculates the current service load; if the load is too high, the computing task is assigned to the adjacent edge server computing pool for calculation; if the current edge server computing pool still cannot complete the computing offload task, the task is offloaded to the cloud server for calculation execution.

[0012] Furthermore, the dynamic clustering to form a vehicle resource cluster specifically includes the following steps:

[0013] Construct dynamic vehicle graph structure;

[0014] Set up the service vehicle willingness function;

[0015] Set the inter-vehicle communication holding time: set the signal propagation time; set the data transmission time.

[0016] Furthermore, the construction of the dynamic vehicle graph structure specifically includes the following steps:

[0017] The connections between the moving vehicles are represented by a dynamic graph. By using a graph neural network, the original vehicle features are used as the node features of the graph.

[0018] Vehicle Node v i The vehicle feature at time t is x i (t) = {lo i ,la i ,s i ,a i ,c i}, where s i is the speed of the vehicle, lo i is the longitude of the vehicle, la i is the dimension of the vehicle, a i is the vehicle acceleration, c i is the computing power of the vehicle; the subscript i represents the vehicle node v i ,The edges represent the communication relationships between different vehicles;

[0019] The SAGEConv of the GNN model generates node embedding by aggregating the information of local neighbors. The aggregation of SAGEConv is represented as:

[0020]

[0021] in, For node v i In the embedding at the kth layer, Ni For node v i Neighborhood set, W k is the learnable weight parameter of the fully connected layer k, F ij is the weight of the aggregated message, σ(·) is the activation function ReLU;

[0022] Basic GNN clustering algorithm: input graph G(V,E), edge weight F ij , vehicle node characteristics Search Depth Neighbor node set N i , number of iterations T, loss function J G ; Output cluster assignment; For each time step T, perform the following steps: For each layer search depth K, perform the following steps: For each vehicle node V, perform the following steps: Calculate the loss J G (z i ), in the output embedding z i Run k-means on to get the final clustering results;

[0023] Among them, J G (z i ) is the objective function:

[0024] J G (z i )=-∑ i,j∈V y ij logy′ ij +(1-y ij )log(1-y′ ij );

[0025]

[0026] Furthermore, the setting of the service vehicle willingness function specifically includes the following steps:

[0027] The service willingness score W is expressed as:

[0028]

[0029] Among them, R c is the remaining computing resources, R b is the remaining battery power, L t is the current task load, D is the distance to the vehicle requesting service, is the expected stay time, S h are historical service records and reputation, and α, β, γ, δ, ε, θ are weight parameters.

[0030] Furthermore, the setting of the inter-vehicle communication holding time specifically includes the following steps:

[0031] Set any two vehicles to maintain a distance L in V2V communication, and the communication time T includes the signal propagation time and data transmission time;

[0032] Communication holding time between vehicle i and vehicle j:

[0033] in, is the signal propagation time, which depends on the distance L between vehicles and the signal transmission speed is the data transmission time, which depends on the data size S and the channel bandwidth B.

[0034] Furthermore, the signal propagation time:

[0035] Where L is the distance between vehicles and r is the signal transmission speed;

[0036] The data transmission time:

[0037] Among them, S is the size of the transmitted data, and B is the channel bandwidth;

[0038] Total communication time:

[0039] Furthermore, the task offloading specifically includes the following steps:

[0040] CH vehicle status information broadcast: CH vehicles periodically broadcast status information through the control channel so that other vehicles can obtain the status of CH vehicles in time and make corresponding decisions;

[0041] Communication channel quality evaluation and update: After receiving the broadcast signal of the CH vehicle, other vehicles will estimate the current communication channel quality. The vehicle updates the channel status information by measuring the signal strength and signal-to-noise ratio parameters;

[0042] User request and CSI transmission: The user vehicle sends its CSI and task offloading request to the CH vehicle through the control channel. At the same time, the server vehicle also sends its CSI and current resource usage to the CH vehicle, so that the CH vehicle can fully understand the resource distribution and communication status of the entire network;

[0043] Task offloading decision of CH vehicle: After receiving information from all user vehicles and server vehicles, the CH vehicle will conduct a comprehensive analysis based on the preset offloading algorithm and make the optimal task offloading decision;

[0044] Decision broadcast: After the CH vehicle makes an unloading decision, it broadcasts the decision result to other vehicles through the control channel;

[0045] Task input data transmission: The user vehicle sends the task input data to the designated Server vehicle through the data channel according to the unloading decision of the CH vehicle;

[0046] Task execution and result return: After receiving the task input data, the server vehicle immediately starts to execute the task. After the task is completed, the server vehicle returns the calculation results to the user vehicle through the data channel.

[0047] Furthermore, the status information includes remaining computing resources, battery power, current location, and current task load of the CH vehicle.

[0048] Furthermore, the request includes the task type of the UV, computing resource requirements, and urgency information of the task.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] A computing offloading method based on resource clustering and edge-cloud collaboration not only improves the utilization of computing resources and reduces the delay of task processing through the collaborative scheduling and dynamic allocation of multi-level resources, but also enhances the robustness and flexibility of the system and adapts to complex and changing computing environments. This method has broad application prospects in the fields of intelligent transportation systems, Internet of Vehicles, and Industrial Internet of Things. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a system framework diagram of a computing offloading method based on resource clustering and end-edge-cloud collaboration.

[0052] Figure 2 The present invention is a schematic diagram of an end-edge-cloud collaborative computing architecture for an end-edge-cloud collaborative computing offloading method based on resource clustering.

[0053] Figure 3 A schematic diagram of vehicle resource clusters formed by dynamic clustering for a computing offloading method based on resource clustering and end-edge-cloud collaboration.

[0054] Figure 4 This is a computational offloading process diagram of a computational offloading method based on resource clustering and end-edge-cloud collaboration.

[0055] Figure 5 The present invention is a specific flow chart of a computing offloading method for end-edge-cloud collaboration based on resource clustering. DETAILED DESCRIPTION

[0056] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0057] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0058] See also Figure 1-5 ,A computation offloading method based on resource clustering and edge-cloud collaboration, e.g. Figure 1 and Figure 5 As shown, the following steps are included:

[0059] Build a device-edge-cloud collaborative computing framework, such as Figure 2 As shown;

[0060] Dynamic clustering to form vehicle resource clusters;

[0061] Task offloading: The vehicle generates a task offloading request and determines whether the currently required program is cached on the vehicle side. If the cache exists, further check whether the current vehicle has sufficient power. If the power is sufficient, the computing task is offloaded on the vehicle side; otherwise, the computing task is partially offloaded to the vehicle resource cluster.

[0062] In the vehicle resource cluster, if there is a program cache, the sharing willingness of the vehicle resource cluster is determined; if the service vehicle is willing to provide shared resources, task offloading is performed in the vehicle resource cluster; if there is no program cache in the vehicle resource cluster, the computing task is offloaded to the edge server resource pool;

[0063] The edge server calculates the current service load; if the load is too high, the computing task is assigned to the adjacent edge server computing pool for calculation; if the current edge server computing pool still cannot complete the computing offload task, the task is offloaded to the cloud server for calculation execution.

[0064] The dynamic clustering to form a vehicle resource cluster specifically includes the following steps:

[0065] Construct dynamic vehicle graph structure;

[0066] Set up the service vehicle willingness function;

[0067] Set the inter-vehicle communication holding time: set the signal propagation time; set the data transmission time.

[0068] The construction of the dynamic vehicle graph structure specifically includes the following steps:

[0069] The connections between the moving vehicles are represented by a dynamic graph. By using a graph neural network, the original vehicle features are used as the node features of the graph.

[0070] Vehicle Node v i The vehicle feature at time t is x i (t) = {lo i ,la i ,s i ,a i ,c i}, where s i is the speed of the vehicle, lo i is the longitude of the vehicle, la i is the dimension of the vehicle, a i is the vehicle acceleration, c i is the computing power of the vehicle; the subscript i represents the vehicle node v i ,The edges represent the communication relationships between different vehicles;

[0071] The GNN model consists of four layers, including an input layer, two SAGE convolutional layers and an output layer. The dimension of the input layer is the vehicle features, and the output dimension is predefined.

[0072] The SAGEConv of the GNN model generates node embedding by aggregating the information of local neighbors. The aggregation of SAGEConv is represented as:

[0073]

[0074] in, For node v i In the embedding at the kth layer, N i For node v i Neighborhood set, W k is the learnable weight parameter of the fully connected layer k, F ij is the weight of the aggregated message, σ(·) is the activation function ReLU;

[0075] The number of layers of GNN, i.e., the search depth, is the number of neighbors of the target node in the number of hops from which the target node aggregates information. In a k-layer GNN, a node can collect information from its k-hop neighbors. A clustering algorithm based on GNN is proposed.

[0076] The basic GNN clustering algorithm process is shown in Algorithm 1:

[0077] Algorithm 1: GNN-based clustering algorithm

[0078] 1. Input: graph G(V,E), edge weight F ij , vehicle node characteristics Search Depth Neighbor node set N i , number of iterations T, loss function J G

[0079] 2. Output: Cluster Assignment

[0080] 3. For each time step T, perform the following steps

[0081] 4. For each layer search depth K, perform the following steps

[0082] 5. For each vehicle node V, perform the following steps

[0083] 6.

[0084] 7.

[0085] 8. Calculate the loss J G (z i )

[0086] 9. Embed z in the output i Run k-means on the cluster to get the final clustering result

[0087] Nodes gradually acquire more and more information from further away in the graph, and this process is iterative.

[0088] Among them, J G (z i ) is the objective function:

[0089] J G (z i )=-∑ i,j∈V y ij logy′ ij +(1-y ij )log(1-y′ ij );

[0090]

[0091] Use 2 SAGEConv layers (i.e. search depth K=2). There are several aggregator architectures to choose from, such as Mean Aggregator, Long Short-Term Memory (LSTM) Aggregator, and Pooling Aggregator. Choose Mean Aggregator, which is a simple but significantly better way to compute embeddings. The output is a low-dimensional vector embedding of the node.

[0092] Service willingness is an important factor in determining whether a vehicle participates in computing resource sharing. The level of service willingness directly affects the allocation of edge computing tasks and the overall performance of the system. In order to more accurately reflect the service willingness of each vehicle, an optimized service willingness score function is proposed, which comprehensively considers multiple influencing factors. The service willingness model combines multiple factors, including remaining computing resources, remaining battery power, current task load, distance to the requesting service vehicle, expected stay time, and historical service records and reputation. By combining these factors into a service willingness score, resource scheduling and task allocation can be more efficient, thereby optimizing the overall performance of the system.

[0093] The setting of the service vehicle willingness function specifically includes the following steps:

[0094] The service willingness score W is expressed as:

[0095]

[0096] Among them, R c is the remaining computing resources, R b is the remaining battery power, L t is the current task load, D is the distance to the vehicle requesting service, is the expected stay time, S h is the historical service record and reputation, and α, β, γ, δ, ε, θ are weight parameters. By dynamically evaluating the service willingness between vehicles, the model can effectively encourage the sharing of computing resources between vehicle devices, thereby improving the efficiency and performance of the entire edge computing network.

[0097] The setting of the inter-vehicle communication holding time specifically includes the following steps:

[0098] Set any two vehicles to maintain a distance L in V2V communication, and the communication time T includes the signal propagation time and data transmission time;

[0099] Communication holding time between vehicle i and vehicle j:

[0100] in, is the signal propagation time, which depends on the distance L between vehicles and the signal transmission speed is the data transmission time, which depends on the data size S and the channel bandwidth B.

[0101] The signal propagation time:

[0102] Where L is the distance between vehicles and r is the signal transmission speed;

[0103] The data transmission time:

[0104] Among them, S is the size of the transmitted data, and B is the channel bandwidth;

[0105] Total communication time:

[0106] like Figure 4 As shown, the task offloading specifically includes the following steps:

[0107] Cluster Head (CH) vehicle status information broadcast: CH vehicles periodically broadcast status information through the control channel so that other vehicles can obtain the status of CH vehicles in time and make corresponding decisions;

[0108] Communication channel quality assessment and update: After receiving the broadcast signal from the CH vehicle, other vehicles will estimate the current communication channel quality. The vehicle updates the channel state information (CSI) by measuring the signal strength and signal-to-noise ratio parameters. The CSI update is dynamic and can reflect the changes in communication quality in the current environment, providing a basis for subsequent task offloading and resource allocation.

[0109] User request and CSI transmission: The user vehicle sends its CSI and task offloading request to the CH vehicle through the control channel. At the same time, the server vehicle also sends its CSI and current resource usage to the CH vehicle, so that the CH vehicle can fully understand the resource distribution and communication status of the entire network;

[0110] Task offloading decision of CH vehicle: After receiving information from all user vehicles and server vehicles, the CH vehicle will conduct a comprehensive analysis based on the preset offloading algorithm and make the optimal task offloading decision. The process involves consideration of factors such as the priority of each task, the resource status of each vehicle, and the quality of the communication channel. The goal is to maximize the overall system performance and minimize task delays and resource consumption.

[0111] Decision broadcast: After the CH vehicle makes an unloading decision, it broadcasts the decision result to other vehicles through the control channel; the decision result includes detailed information such as the execution plan of each task, the corresponding server vehicle, and the data transmission arrangement, so that each vehicle can perform the corresponding operation according to the instructions of the CH vehicle.

[0112] Task input data transmission: The user vehicle sends the task input data to the designated Server vehicle through the data channel based on the unloading decision of the CH vehicle; the data channel is usually a high-bandwidth, low-latency communication path to ensure that the task data can be transmitted to the target Server vehicle quickly and reliably.

[0113] Task execution and result return: After receiving the task input data, the server vehicle immediately starts to execute the task. After the task is completed, the server vehicle returns the calculation result to the user vehicle through the data channel. The process needs to ensure the reliability and timeliness of data transmission to meet the needs of real-time applications.

[0114] The status information includes the remaining computing resources, battery power, current location, current task load, etc. of the CH vehicle.

[0115] The request includes the task type of UV, computing resource requirements, and task urgency information.

[0116] This method has the following characteristics:

[0117] Resource clustering mechanism

[0118] Dynamic resource allocation: By building a vehicle resource cluster, an edge server resource pool, and a cloud server computing pool, dynamic allocation of computing tasks among multiple levels of resources is achieved. Tasks can be flexibly scheduled based on the vehicle's computing resources, battery power, task load, and other states to improve resource utilization.

[0119] Reduce task migration costs: By utilizing the resource clustering mechanism, when vehicle movement causes changes in network topology, the frequency and cost of task migration can be minimized to ensure the stability and continuity of computing tasks.

[0120] Service willingness scoring mechanism

[0121] Comprehensive multi-factor consideration: Dynamically calculate the service willingness score by comprehensively considering the vehicle's remaining computing resources, battery power, task load, communication channel quality, expected stay time, historical service records, and penalty items. Use nonlinear functions and weight parameters to improve the accuracy and adaptability of the score.

[0122] Incentivize resource sharing: Design a reasonable service willingness scoring mechanism to encourage the sharing of computing resources between vehicle equipment, reduce the negative impact of selfish behavior on the overall performance of the system, and improve the success rate and efficiency of computing tasks.

[0123] Multi-level collaborative computing

[0124] Device-edge-cloud collaboration: A multi-level collaborative computing offloading method is proposed to dynamically schedule task offloading between the vehicle end, edge server, and cloud server. Tasks are allocated between different levels according to actual computing needs and resource status to optimize system performance.

[0125] Distributed computing and collaborative optimization: Combined with distributed computing technology, through the end-edge-cloud collaborative optimization task offloading strategy, efficient scheduling and utilization of computing resources can be achieved, the delay of task processing can be reduced, and the overall service quality can be improved. The above-mentioned embodiments only express the specific implementation methods of this application, and the description is relatively specific and detailed, but it cannot be understood as a limitation on the scope of protection of this application. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the technical solution of this application, several variations and improvements can be made, which all belong to the scope of protection of this application.

Claims

1. A computing offloading method for device-edge-cloud collaboration based on resource clustering, characterized in that: The following steps are involved: Build a device-edge-cloud collaborative computing framework; Dynamic clustering to form vehicle resource clusters; Task offloading: The vehicle generates a task offloading request and determines whether the currently required program is cached on the vehicle side. If the cache exists, it further checks whether the current vehicle power is sufficient; If the power is sufficient, the computing task is offloaded on the vehicle side; Otherwise, choose to offload part of the computing task to the vehicle resource cluster; In the vehicle resource cluster, if there is a program cache, the sharing willingness of the vehicle resource cluster is determined; If the service vehicle is willing to provide shared resources, task offloading is performed in the vehicle resource cluster; If there is no program cache in the vehicle resource cluster, the computing task is offloaded to the edge server resource pool; The edge server calculates the current service load; if the load is too high, the computing task is assigned to the adjacent edge server computing pool for calculation; if the current edge server computing pool still cannot complete the computing offload task, the task is offloaded to the cloud server for computing execution.

2. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 1 is characterized in that: The dynamic clustering to form a vehicle resource cluster specifically includes the following steps: Construct dynamic vehicle graph structure; Set up the service vehicle willingness function; Set the inter-vehicle communication hold time: set the signal propagation time; set the data transmission time.

3. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 2 is characterized in that: The construction of the dynamic vehicle graph structure specifically includes the following steps: The connections between moving vehicles are represented by a dynamic graph. By using a graph neural network, the original vehicle features are used as node features of the graph. Vehicle node v i The vehicle feature at time t is x i (t) = {lo i ,la i ,s i ,a i ,c i }, where s i is the speed of the vehicle, lo i is the longitude of the vehicle, la i is the dimension of the vehicle, a i is the vehicle's acceleration, c i is the computing power of the vehicle; subscript i represents the vehicle node v i ,edges represent the communication relationships between different vehicles; The SAGEConv of the GNN model generates node embedding by aggregating the information of local neighbors. The aggregation of SAGEConv is represented as: in, For node v i In the embedding at the kth layer, N i For node v i Neighborhood set, W k is the learnable weight parameter of the fully connected layer k, F ij is the weight of the aggregated message, σ(·) is the activation function ReLU; Basic GNN clustering algorithm: input graph G(V,E), edge weight F ij , vehicle node characteristics Search Depth Neighbor node set N i , number of iterations T, loss function J G ; Output cluster assignment; For each time step T, perform the following steps: For each layer search depth K, perform the following steps: For each vehicle node V, perform the following steps: Calculate the loss J G (z i ), in the output embedding z i Run k-means on the dataset to obtain the final clustering results; Among them, J G (z i ) is the objective function: J G (z i )=-∑ i,j∈v y ij logy′ ij +(1-y ij )log(1-y′ ij ); 4. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 2 is characterized in that: The setting of the service vehicle willingness function specifically includes the following steps: The service willingness score W is expressed as: Among them, R c is the remaining computing resources, R b is the remaining battery power, L t is the current task load, D is the distance to the service requesting vehicle, is the expected stay time, S h are historical service records and reputation, and α, β, γ, δ, ε, θ are weight parameters.

5. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 2 is characterized in that: The method of setting the inter-vehicle communication holding time specifically includes the following steps: Assume that any two vehicles maintain a distance L in V2V communication, and the communication time T includes the signal propagation time and data transmission time; Communication holding time between vehicle i and vehicle j: in, is the signal propagation time, which depends on the distance L between vehicles and the signal transmission speed is the data transmission time, which depends on the data size S and the channel bandwidth B.

6. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 5 is characterized in that: The signal propagation time: Where L is the distance between vehicles and r is the signal transmission speed; The data transmission time: Where S is the size of the data being transmitted, and B is the channel bandwidth; Total communication time:

7. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 1 is characterized in that: The task offloading specifically includes the following steps: CH vehicle status information broadcast: CH vehicles periodically broadcast status information through the control channel so that other vehicles can obtain the status of CH vehicles in a timely manner and make corresponding decisions; Communication channel quality assessment and update: After receiving the broadcast signal from the CH vehicle, other vehicles will estimate the current communication channel quality. The vehicle updates the channel status information by measuring the signal strength and signal-to-noise ratio parameters. User request and CSI transmission: The user vehicle sends its CSI and task offloading request to the CH vehicle through the control channel. At the same time, the server vehicle also sends its CSI and current resource usage to the CH vehicle, so that the CH vehicle can fully understand the resource distribution and communication status of the entire network; Task offloading decision of CH vehicles: After receiving information from all user vehicles and server vehicles, CH vehicles will conduct a comprehensive analysis based on the preset offloading algorithm and make the optimal task offloading decision; Decision broadcast: After the CH vehicle makes an unloading decision, it broadcasts the decision result to other vehicles through the control channel; Task input data transmission: The user vehicle sends the task input data to the designated server vehicle through the data channel based on the unloading decision of the CH vehicle; Task execution and result return: After receiving the task input data, the server vehicle immediately starts executing the task. After the task is completed, the server vehicle returns the calculation results to the user vehicle through the data channel.

8. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 7 is characterized in that: The status information includes the remaining computing resources, battery power, current location, and current task load of the CH vehicle.

9. The method for computing offloading based on resource clustering and edge-cloud collaboration according to claim 7, characterized in that: The request includes UV's task type, computing resource requirements, and task urgency information.

Citation Information

Cited By

  • Internet of vehicles secure unloading and caching optimization method based on graph neural network

    CN121099376A