Research On Acceleration Technology Of Virtual Test Of Autopilot System

The increasing demand for human travel has prompted the rapid transformation of automobiles in the direction of the "new four modernizations",and autonomous driving systems have also flourished.As the requirements of autonomous driving systems continue to escalate,the complexity and number of test scenarios have also increased dramatically,making traditional stand-alone tests unable to meet the demands of massive autonomous driving test scenarios.In response to this problem,based on the overall design,this paper develops an automated distributed simulation test platform through the study of a balanced scheduling system and fault-tolerant mechanism for simulation tasks,and verifies its functions and acceleration performance.The specific research content is as follows:Based on the analysis of the defects of the distributed simulation test platform of the existing automatic driving system,the paper conducts demand analysis,and designs an overall plan for the goal of achieving full performance mining,stable operation,and improvement of automatic testing.Based on this,compare the advantages and disadvantages of the existing commercial virtual simulation test software.And independent design of key software to improve platform performance.Further,consider the requirements of platform coordination capabilities and other requirements to select a distributed messaging system for communication.Simulation test scenarios have different internal complexity and are affected by the performance of external nodes.The test time is difficult to estimate,and the scenarios are independent,which leads to uneven distribution of manual allocation methods,and node performance resources are idle and wasted.In response to this problem,conventional scheduling strategies and performance indicators are analyzed on the basis of the overall design of the scheduling system,and then dynamic scenarios are constructed,and the fastest response method is selected as the scheduling strategy through experiments.Further,build a dispatch message center and design a slave deployment mode to ensure the implementation of the strategy.By designing a dynamically adjusted scheduling system,the load balance of each node of the platform is realized.Because the test platform runs for a long time and involves many elements,if the platform does not have adaptive processing functions,it is difficult to guarantee test reliability.Based on the characteristics of the platform test work,this paper analyzes and obtains typical faults such as software faults and host faults,and based on the extent of the fault’s impact,divides the faults into faults affecting test operation and faults affecting test evaluation.Under the platform architecture,fault handling is carried out by establishing a fault-tolerant mechanism with adaptive functions to improve the reliability of the test.In order to achieve rapid verification of the above systems and mechanisms,based on the integration of the platform,a cross-node closed-loop control method and a cross-software self-starting call method are designed to jointly build a distributed drive call mode to improve platform automation performance.Based on the testing and verification of the platform’s automation level and load balancing ability,various faults are simulated to perform reliability experiments on the fault tolerance mechanism.And compare and analyze the simulation test acceleration of stand-alone and platform.