Search

CN-121996423-A - Calculation task distribution method and control module for vehicle-road cooperative vehicle formation

CN121996423ACN 121996423 ACN121996423 ACN 121996423ACN-121996423-A

Abstract

The invention relates to the field of Internet of vehicles, in particular to a calculation task allocation method and a control module for vehicle-road cooperative vehicle formation. The method comprises the steps of dynamically updating calculation tasks of vehicle formation, classifying the calculation tasks and adding tag information to form a global task pool, obtaining residual calculation force values of calculation force nodes such as a head car, a following car, a drive test MEC and a cloud server, and detecting the health degree of the nodes to form a multi-source calculation force pool. And carrying out basic capability scoring on each computing task, and screening candidate nodes with basic capability scoring higher than a preset value. And finally, taking the total execution cost of the computing tasks as the lowest and the overall load balance of the computing power nodes as optimization targets, and carrying out many-to-many matching on each computing task and candidate nodes in the double pools. The method can be used for solving the problems of larger task time delay and unbalanced equipment load caused by unreasonable calculation task distribution in the existing vehicle formation.

Inventors

  • Zu Xian
  • KAN YIQUN
  • XIE FEI
  • WANG ZHONG

Assignees

  • 合肥师范学院

Dates

Publication Date
20260508
Application Date
20260127

Claims (10)

  1. 1. The calculation task allocation method for the vehicle-road cooperative vehicle formation is characterized by comprising the following steps of: Classifying the calculation tasks and adding tag information comprising calculation power demand values, time delay sensitivity, safety risk coefficients and task source identifiers to form a global task pool; The method comprises the steps of acquiring resource state information reported by all power calculation nodes including a head car, a following car, a drive test MEC and a cloud server, and calculating a residual power value; Predicting potential tasks of each computing task and generating computing force dependencies of each computing task; generating the priority of each calculation task by combining the calculation force dependence and the label information; performing basic capability scoring on each computing node, and screening computing nodes with basic capability scoring higher than a preset value to form candidate nodes; The method comprises the steps of taking priority, calculation force required value, time delay threshold and security level as task side matching parameters, taking residual calculation force value and communication cost as calculation force side matching parameters, presetting constraint conditions of task time delay and calculation force node load, taking the minimum total execution cost of calculation tasks and the whole load balance of calculation force nodes as optimization targets, and adopting an improved greedy algorithm to carry out overall many-to-many matching on calculation tasks in a global task pool and candidate nodes in a multi-source calculation force pool.
  2. 2. The computing task allocation method for vehicle-road-oriented cooperative vehicle formation according to claim 1, wherein: the core tasks of the head car comprise global path planning and formation control, and the edge tasks of the following car comprise local obstacle detection assistance, log statistics and formation state synchronization; The system comprises a control system, a management system, a control system and a management system, wherein the control system comprises a control system, a control system and a management system, and is characterized in that computing tasks are divided into four types, namely a perception type, a decision type, a control type and a management type, wherein the perception type computing tasks are divided into two types of obstacle detection and traffic sign recognition; and/or the information of the computing power requirement of each computing task comprises CPU occupancy rate, GPU occupancy rate and memory occupancy rate, and the time delay sensitivity information comprises maximum tolerable time delay and time delay sensitivity weight; and/or the global task pool adopts a 1S sliding window to receive new tasks in real time, remove completed tasks, and synchronously update the computational power dependency and the priority of the tasks.
  3. 3. The method for assigning computational tasks for vehicle-road-oriented cooperative vehicle formation according to claim 1, wherein the evaluation of the performance of the force nodes by means of heartbeat detection and differentiated probe mechanisms comprises: For a core computing node comprising a head car and a road side MEC, sending a heartbeat packet every 200ms, sending a virtual test packet every 500ms, evaluating a plurality of performance indexes comprising response time delay and packet loss rate, and generating an evaluation value of health degree according to the results of the indexes; For the edge computing power node comprising a following server and a cloud server, sending a heartbeat packet every 500ms, randomly sending a pressure test packet, evaluating a plurality of performance indexes comprising corresponding time delay, load stability and communication reliability, and generating a health evaluation value according to the results of the indexes; and/or the multi-source computing power pool adopts a self-adaptive sliding window to filter failure nodes with health degrees smaller than a threshold value in real time.
  4. 4. The computing task allocation method for vehicle-road-oriented cooperative vehicle formation according to claim 1, wherein: The calculation force demand value of the calculation task and the residual calculation force value of the calculation force node are generated according to index data through a multidimensional vector modeling algorithm and are converted into a unified characterization value; And/or, potential tasks of each real-time computing task are generated through a potential task prediction model pre-trained based on the LSTM model; the potential task prediction model is used for generating triggering probability of various tasks according to the input multidimensional characteristics comprising the speed, the acceleration and the road curve, and further taking the task with the triggering probability higher than a preset value as a potential task of the current calculation task; And/or calculating the calculation force dependence degree of the current calculation task according to the predicted potential task of the current calculation task, wherein the calculation formula is as follows: The calculation force dependency of the current calculation task=0.7×the calculation force demand of the current calculation task+0.3×the calculation force demand of the potential task of the current task.
  5. 5. The method for assigning calculation tasks for vehicle-road-oriented cooperative vehicle formation according to claim 4, wherein the calculation formula of the priority of any calculation task is as follows: priority = 0.4 x calculation force dependency + 0.3 x delay sensitivity weight + 0.2 x security risk factor + 0.1 x task source identification corresponding weight; In the above method, in the weights corresponding to the task source identifiers, the value of the head-car task is 0.6, the value of the following-car task is 0.4, the safety risk coefficient is related to the task type, the value of the control class is 0.8, the value of the decision class is 0.6, the value of the perception class is 0.5, and the value of the management class is 0.2.
  6. 6. The method for assigning computational tasks for vehicle-road-oriented collaborative vehicle formation according to claim 5, wherein the basic capability scores of the computing power nodes are calculated as follows: basic capability score = 0.5 x health +0.3 x residual power value +0.2 x score for communication delay Wherein, when the communication time delay of any computing power node is higher, the score of the communication time delay is lower.
  7. 7. The method for computing task allocation for vehicle-road-oriented collaborative vehicle formation according to claim 6, characterized in that the global many-to-many matching process between computing task and candidate nodes comprises the steps of: (1) Sequencing all calculation tasks in the global task pool according to priority as objects of each row, sequencing all candidate nodes in the multi-source calculation pool according to basic capability scores as objects of each column, and further constructing a matching matrix, wherein elements in the matching matrix are the execution cost of the corresponding calculation tasks when the candidate nodes are executed; (2) Sequentially matching the computing tasks with the top 30% of the priority ranks to candidate nodes which can enable the execution cost of the computing tasks to be the lowest according to the priority order; (3) Distributing the residual calculation tasks to each candidate node according to the principle of minimizing the load rate, wherein each calculation task is preferentially matched with the candidate nodes with the load rate less than 60 percent; (4) And migrating part of calculation tasks on the candidate nodes with the load rate higher than 80% to the candidate nodes with the load rate lower than 50% so as to optimize the overall load balance.
  8. 8. The method for computing task allocation for vehicle-road-oriented cooperative vehicle formation according to claim 7, characterized in that; In a preset constraint condition, the time delay of a core task is less than 50ms, and the time delay of other tasks is less than 200ms, wherein the load of each computing force node is less than 85%; And/or, the calculation formula of the execution cost of any calculation task on any candidate node is as follows: execution cost = 0.4 x computation cost + 0.4 x communication cost + 0.2 x migration cost In the above formula, the calculation cost is the ratio of the calculation power requirement value of the calculation task to the residual calculation power value of the calculation power node, the communication cost is comprehensively calculated based on the network distance and the equipment bandwidth, and the migration cost is characterized by the communication cost between two migration objects.
  9. 9. A computer program product comprising a computer program which, when executed by a processor, implements a method for assigning computational tasks for vehicle-oriented co-vehicle formation according to any one of claims 1-8, dynamically assigning individual computational tasks generated by vehicle formation to individual computing power nodes.
  10. 10. A vehicle formation control module comprises a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor, when executing the computer program, realizes the calculation task allocation method for vehicle formation oriented to the vehicle road according to any one of claims 1-8, and dynamically allocates each calculation task generated by vehicle formation to each calculation power node.

Description

Calculation task distribution method and control module for vehicle-road cooperative vehicle formation Technical Field The invention relates to the field of Internet of vehicles, in particular to a calculation task allocation method for vehicle-road-oriented cooperative vehicle formation, and a corresponding computer program product and a vehicle formation control module. Background In recent years, the intelligent network driving technology has been remarkably developed, and the vehicle formation technology is used as an important application form, so that the air resistance can be effectively reduced, the road utilization rate can be improved, the energy consumption can be reduced, and the traffic safety can be enhanced through the tight coordination and following among vehicles. In the formation driving process, the head car is used as a control core of the whole formation, bears key tasks such as environment sensing, path planning, decision control and the like, and particularly has outstanding demands on computing resources. Although the vehicle formation technology has advanced to some extent, the practical application still faces a plurality of challenges that (1) the calculation burden of the head car is excessive, namely, in the existing formation system, the head car needs to process the self-perception, decision task and the formation overall control logic at the same time, so that the calculation resources are tense, and the real-time performance and reliability of the system are affected. (2) Under the traditional architecture, the formation scale is enlarged, so that the calculation load of the head car is exponentially increased, and the operation boundary and flexibility of formation application are limited. (3) The network dependence is too strong, namely, the scheme which completely relies on the cooperation of the vehicle and the road is easy to cause control delay and even failure in the unstable area (such as tunnels and mountain areas) of the network signals, and the formation safety is influenced. (4) The task allocation strategy is extensive, namely the prior art adopts a fixed task allocation mode, and the real-time characteristic of the task, the network state change and the dynamic fluctuation of the computing resource cannot be fully considered, so that the resource utilization rate is low. (5) The heterogeneous system has the difficulty of cooperative computation of the head car, the edge computing node and the cloud server, wherein the heterogeneous system exists in the aspects of hardware architecture, an operating system, a communication protocol and the like, and the complexity of cooperative computation is increased. The application of the vehicle formation is realized, and whether the vehicle formation is realized by a single automatic driving scheme or a network-connected vehicle formation based on the cooperative energization of the vehicle and the road, a large number of vehicle sensors and a certain amount of calculation resource support are required, but under the condition of limited calculation force resources at the vehicle end, the real-time performance of the calculation of related tasks under the formation condition is difficult to meet, and the vehicle formation technology becomes the landing resistance. Disclosure of Invention In order to solve the problems of large task time delay and unbalanced equipment load caused by unreasonable calculation task allocation in the existing vehicle formation, the invention provides a calculation task allocation method for vehicle-road-oriented cooperative vehicle formation, and a corresponding computer program product and a vehicle formation control module. The technical scheme provided by the invention is as follows: a calculation task allocation method for vehicle-road cooperative vehicle formation comprises the following steps: And dynamically updating the calculation tasks of all vehicles in the vehicle formation, including the core tasks of the head vehicles and the edge tasks of the following vehicles. Classifying each computing task, and adding tag information comprising a computing power requirement value, time delay sensitivity, a security risk coefficient and a task source identifier to form a global task pool. And acquiring resource state information reported by all power calculation nodes including a head car, a following car, a drive test MEC and a cloud server, and calculating a residual power calculation value. And selecting the power computing nodes with the node health degree higher than a preset value to form a multi-source power computing pool. And generating the priority of each computing task by combining the label information and the computing force dependence. And carrying out basic capability scoring on each computing node in the multi-source computing pool according to the health degree, the residual computing value and the communication time delay, and screening computing nodes with basic capabilit