CN-122019010-A - Dependency task computing and unloading method based on collaborative vehicle-mounted edge computing network

CN122019010ACN 122019010 ACN122019010 ACN 122019010ACN-122019010-A

Abstract

The invention discloses a dependency task computing and unloading method based on a collaborative vehicle-mounted edge computing network, which belongs to the field of vehicle-mounted edge computing and comprises the steps of constructing a dynamic computing network pool for fusing a fixed edge server and a mobile opportunity vehicle. The method comprises the steps of identifying and combining repeated subtasks in a task directed acyclic graph to eliminate redundant calculation, adopting a two-stage collaborative optimization strategy, firstly solving a calculation unloading decision by utilizing an improved quantum genetic algorithm introducing a load balancing mechanism, then modeling task scheduling as a Markov decision process, extracting topological characteristics by using a graph convolution neural network, and carrying out self-adaptive scheduling based on a near-end strategy optimization algorithm. The method can effectively reduce the average execution delay of the task and the total energy consumption of the system, and improves the overall utilization efficiency and scheduling stability of resources in a dynamic network environment.

Inventors

XUE DUAN
LI JING
LIU LI
XU GUOHUA
HOU HAIXIA
ZHANG YULIN
LI QIZE

Assignees

太原学院

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1.A method for calculating and unloading dependent tasks based on a collaborative vehicle-mounted edge computing network is characterized by comprising the following steps: constructing a dynamic vehicle-mounted edge computing network paradigm consisting of user vehicles, opportunity vehicles and edge servers, wherein the opportunity vehicles are vehicles with idle computing resources and predictable movement tracks; Identifying an on-vehicle computing task with a dependency relationship generated by the user vehicle, and representing the task as a directed acyclic graph, wherein repeated subtasks are combined in the directed acyclic graph; Determining a calculation offloading decision of the subtasks based on the resource state of the dynamic vehicle-mounted edge calculation network paradigm and the directed acyclic graph, wherein the offloading decision comprises local execution, offloading to an edge server execution or offloading to an opportunistic vehicle execution; and scheduling the subtasks according to the calculation unloading decision so as to determine the execution sequence and the starting time of each subtask.
2. The method of claim 1, wherein the identifying user vehicle-generated on-board computing tasks having dependencies comprises: Acquiring a subtask set generated by a user vehicle, wherein each subtask is represented by a data volume, required computing resources and a maximum tolerable delay triplet; Constructing an initial directed acyclic graph according to the execution sequence dependency relationship among the subtasks; the same subtasks in the initial directed acyclic graph are detected and combined to eliminate redundant computation.
3. The method of claim 1, wherein the determining a computational offload decision comprises: modeling a computational offload problem as a multiprocessor scheduling problem; solving the problem with an improved quantum genetic algorithm to obtain an offloading decision that minimizes a system cost function, wherein the system cost function is a weighted sum of task execution delay and system energy consumption; the minimization of the system cost function satisfies the constraint that each subtask is executed on only one processor, the computing resources allocated to the task do not exceed the total amount of processor resources, the subtask must begin to execute after all of its predecessor tasks are completed, and the actual execution time of the subtask does not exceed its maximum tolerable delay.
4. A method according to claim 3, wherein the improved quantum genetic algorithm comprises: Adopting quantum bit coding to represent the unloading decision of the subtasks; Updating the quantum population by using a quantum rotating gate, wherein the magnitude of the rotating angle is dynamically adjusted according to the evolution state of the algorithm; Introducing a load balancing correction coefficient into the fitness function, wherein the correction coefficient is dynamically calculated based on the resource utilization variance of the edge calculation node.
5. The method of claim 1, wherein scheduling the subtasks comprises: modeling task scheduling problems as a markov decision process; And solving the Markov decision process by adopting a reinforcement learning method based on near-end strategy optimization so as to acquire a task scheduling strategy.
6. The method of claim 5, wherein the state of the markov decision process is a directed acyclic graph of a current task, and wherein the encoding of the state is accomplished by extracting topological features and node attribute features of the graph through a graph convolution neural network.
7. The method of claim 6, wherein the graph convolutional neural network aggregates pieces of neighbor information for nodes by stacking multiple layers of graph convolutional layers and globally pools all node features to generate a global state vector for reinforcement learning decisions.
8. The method of claim 5, wherein the act of the markov decision process is to add edges in the directed acyclic graph to reduce task dependencies, the selecting of the act comprising sequentially selecting a start node and an end node; The resource availability factor and the estimated waiting time are introduced into the calculation model of the opportunistic vehicle, and the estimated waiting time is calculated according to the calculated amount of the existing tasks in the vehicle task queue and the resource availability factor.
9. An electronic device comprising a memory, a processor and a computing program stored in the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1-8 when executing the computing program.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the method of any of claims 1-8.

Description

Dependency task computing and unloading method based on collaborative vehicle-mounted edge computing network Technical Field The invention belongs to the field of vehicle-mounted edge calculation, and particularly relates to a dependency task calculation unloading method based on a collaborative vehicle-mounted edge calculation power network. Background With the popularization of intelligent vehicle-mounted applications such as automatic driving, augmented reality navigation and the like, the vehicle-mounted computing task has the characteristics of high complexity, strong subtask dependence and severe real-time requirements. The traditional cloud computing mode is difficult to meet the requirement of a vehicle on real-time response due to large return delay. Vehicle-mounted edge computing (VEC) effectively relieves the limitation of computing power and energy consumption of a vehicle-mounted terminal by offloading computing tasks to a fixed server (such as a road side unit) at the network edge. Currently, mainstream research and practice is mostly based on a three-layer fixed architecture of 'cloud-edge-end', aiming at reducing delay and energy consumption by optimizing task offloading and resource allocation. However, existing VEC systems still face several challenges, firstly, the fixed edge server deployment density and cost are limited, and contradiction exists between coverage and dynamic high-rise vehicle computing requirements, which easily causes resource shortage or uneven utilization. Secondly, existing architectures fail to fully mine and utilize fragmented computing power resources carried by a large number of vehicles in an idle state in a roadway environment. Furthermore, most studies do not take into account the same subtasks (repeatability) that may exist between different vehicle tasks, resulting in redundant computation and resource waste. Finally, in the face of the problem of directed acyclic graph scheduling composed of large-scale dependent tasks, the existing task scheduling method based on heuristic or traditional reinforcement learning is often faced with the problems of difficult convergence and unstable strategy in a high-dimensional state space, and the joint refinement optimization of delay and energy consumption is difficult to realize. Disclosure of Invention In order to solve the technical problems, the invention provides a dependency task computing and unloading method based on a collaborative vehicle-mounted edge computing network, which comprises the following steps: constructing a dynamic vehicle-mounted edge computing network paradigm consisting of user vehicles, opportunity vehicles and edge servers, wherein the opportunity vehicles are vehicles with idle computing resources and predictable movement tracks; Identifying an on-vehicle computing task with a dependency relationship generated by the user vehicle, and representing the task as a directed acyclic graph, wherein repeated subtasks are combined in the directed acyclic graph; Determining a calculation offloading decision of the subtasks based on the resource state of the dynamic vehicle-mounted edge calculation network paradigm and the directed acyclic graph, wherein the offloading decision comprises local execution, offloading to an edge server execution or offloading to an opportunistic vehicle execution; and scheduling the subtasks according to the calculation unloading decision so as to determine the execution sequence and the starting time of each subtask. Optionally, the identifying the vehicle-mounted computing task with the dependency relationship generated by the user vehicle includes: Acquiring a subtask set generated by a user vehicle, wherein each subtask is represented by a data volume, required computing resources and a maximum tolerable delay triplet; Constructing an initial directed acyclic graph according to the execution sequence dependency relationship among the subtasks; the same subtasks in the initial directed acyclic graph are detected and combined to eliminate redundant computation. Optionally, the determining calculates an offloading decision comprises: modeling a computational offload problem as a multiprocessor scheduling problem; The problem is solved using a modified quantum genetic algorithm to obtain an offloading decision that minimizes a system cost function, which is a weighted sum of task execution delay and system energy consumption. Optionally, the improved quantum genetic algorithm comprises: Adopting quantum bit coding to represent the unloading decision of the subtasks; Updating the quantum population by using a quantum rotating gate, wherein the magnitude of the rotating angle is dynamically adjusted according to the evolution state of the algorithm; Introducing a load balancing correction coefficient into the fitness function, wherein the correction coefficient is dynamically calculated based on the resource utilization variance of the edge calculation node. Optionally, th