Research On The Evaluation Method Of Autonomous Driving Motion Planning And Control System Based On Vehicle-in-the-Loop

The functions of the autonomous driving motion planning and control system include local path planning,speed planning,trajectory tracking and vehicle motion control.The system drives the autonomous vehicle actuator based on the driving environment information from the autonomous driving sensing system,and then changes the motion state of the autonomous vehicle.Its performance directly determines the safety,intelligence and reliability of the autonomous vehicles during driving.Therefore,in order to ensure the safe and reliable application of autonomous vehicles on a large scale,it is necessary to conduct efficient and comprehensive testing and evaluation research on the autonomous driving motion planning and control system.The existing test and evaluation methods are mainly deficient in two aspects:on the one hand,the traditional test methods are difficult to balance cost,accuracy and efficiency,and there are problems such as long test cycles,high test costs and low test accuracy;on the other hand,the traditional evaluation methods have few types and single indexes,and the results are difficult to reflect the interference of random factors during the operation of the autonomous driving motion planning and control system.To address these problems,the thesis relies on the National Key R&D Program of China "Digital twin-based traffic-in-the-loop testing technology for autonomous driving"(2021YFB2501204),and carries out research on the evaluation method of autonomous driving motion planning and control system based on the vehicle-in-loop testing and cloud model.The vehicle-in-the-loop testing method effectively integrates the advantages of virtual simulation test safety,high efficiency,low cost and high authenticity of road test data;the evaluation method based on cloud-model is widely adaptable,and the evaluation results can reflect the stability of the operation process of the evaluated object and the uncertainty of the evaluation results.The main work of the paper is as follows:(1)Based on the investigation of existing domestic and foreign autonomous driving motion planning and control technologies and their evaluation methods,the research idea of integrating the vehicle-in-the-loop test method with the evaluation method based on cloud model was established;the vehicle-in-the-loop test platform for autonomous driving was improved to ensure the real-time acquisition of multi-source data from the tested autonomous vehicle and the platform during the test;for the needs of multi-source data collection and comprehensive evaluation and analysis in the test,the data collection and evaluation software was developed to realize the real-time control and visualization of key data in the evaluation process.(2)For the dynamic and discrete nature of the autonomous driving motion planning and control system,the optimal reference trajectory dispersion measurement method is proposed:firstly,a multi-degree-of-freedom dynamics model of the autonomous vehicle is established;secondly,the trajectory planning problem is decoupled into path planning problem and velocity planning problem by considering various factors such as path curvature,velocity smoothness and acceleration smoothness;finally,the planned trajectory is tracked by the Model Predictive Control(MPC)algorithm to generate a basis for the subsequent evaluation of the autonomous driving motion planning and control performance by the cloud model.Finally,the planned trajectory is tracked by MPC algorithm to generate the optimal evaluation reference trajectory,which provides the basis for the subsequent evaluation of the autonomous driving motion planning and control performance by cloud model.(3)The evaluation method based on cloud model is carried out for the operation characteristics of the autonomous driving motion planning and control system and the multisource data collection characteristics of the vehicle-in-the-loop test platform.First,the direct evaluation indexes of motion planning and control performance are extracted and multiple composite evaluation indexes of motion planning and control performance are designed based on the vehicle-in-the-loop test data;then,a comprehensive evaluation method based on the cloud model is proposed,which reduces the influence of subjective and objective differences through game-theoretic-combination assignment,and the evaluation results can reflect the stability and uncertainty in the operation of the automatic driving motion planning and control system uncertainty.(4)Based on the combined virtual and real automated vehicle-in-the-loop test platform,a joint test environment of Car Sim,Pre Scan and Matlab was established,and a two-way fourlane intersection static obstacle scenario and a dynamic traffic participant interaction scenario were designed to test and evaluate the automated driving motion planning and control performance.Three groups of control experiments were set up,and the single performance and comprehensive performance of each group of experimental test vehicles were evaluated and analyzed based on the evaluation method proposed in this paper.The experimental results show that the evaluation method proposed in this paper can effectively test and evaluate the operation process of the autonomous driving motion planning and control system,and the evaluation results can reflect the stability and uncertainty in the operation process of the motion planning and control system,and make up for the shortcomings of the existing evaluation methods.The research results of this paper can provide well support for the development of autonomous driving motion planning and control technology.