Research On Automation Simulation Test And Evaluation Technology For Intelligent Driving Systems

The position of intelligent vehicles has been gradually risen to the level of national strategy because they can effectively reduce traffic accidents,prevent environmental pollution and alleviate energy crisis.However,due to the algorithms and structures of the intelligent driving systems are complex,the internal and external factors are numerous and the application environment is dynamic and uncontrollable,the existing test and evaluation technology cannot meet the requirements of high safety,reliability and timeliness in terms of efficiency and effectiveness.The lack of reasonable and efficient evaluation methods has become a bottleneck restricting the development and application of intelligent vehicles.Therefore,based on the commercial software and hardware platform,this paper designs a test scenario generation algorithm considering both test coverage and test efficiency,and develops an automatic simulation test evaluation platform,which realizes the comprehensive optimization of test coverage,cost and effectiveness.The specific research work is as follows.In order to design efficient test cases,so as to find system defects quickly and improve the research and development efficiency of intelligent driving systems,it is quite important to evaluate the “effectiveness” of test cases quantitatively.However,due to the complexity of intelligent driving system itself and its application environment,it is hard to directly evaluate the testing effect of the test case without executing it first,which makes it difficult to take the effectiveness of the test cases into account when designing them.To solve this problem,based on the construction of tree analysis model of the influencing factors of intelligent driving system,the relative objectification of engineer's subjective experience is realized with the help of the scaling method and Delphi method,then the calculation model of relative importance indices of influencing factors is established with eigenvector method and possibility-satisfaction method,and the complexity index of test case is further proposed.In the end,the indirect evaluation of the effectiveness of test case is realized,which provides a basis for the generation and optimization methods of test cases showing below.The existing test case generation methods of intelligent driving system mostly focus on the number and coverage of test cases,but rarely consider its effectiveness,which results in low overall test efficiency.In this paper,based on the combinatorial testing algorithm,the complexity index is introduced to improve the generation process of test cases,that is,based on the complexity improvement coefficient,two generation strategies of “quantity first” and “effectiveness first” are selected on the premise of ensuring the test coverage,and then an improved combinatorial test case generation algorithm based on complexity is proposed.The experimental results in the lane departure warning system show that the improved algorithm can improve the average complexity of the test cases by more than 67% compared with PICT and other traditional combinatorial testing algorithms.The design of complexity improvement coefficient is the key part of the improved combinatorial testing algorithm,but the traditional optimization method is not suitable because of the black box attribute of the algorithm.In order to solve this problem,based on the establishment of the quantitative boundary estimation equation of the generated test cases to achieve the normalization of the optimization index,the statistical distribution characteristics of the objective optimization function are inferred by using Bayesian theory,so as to achieve the fast and accurate positioning of the optimal algorithm parameters.Furthermore,in order to make the “static” and “discrete” test cases suite more “dynamic” and “continuous”,the traditional hierarchical clustering method is applied to aggregate similar test cases into the same test scenario,and the sorting algorithm based on similarity and complexity is introduced to further enhance the similarity between adjacent test cases in the scenarios,and make the test cases with higher complexity be executed first as much as possible,so as to enhance the “dynamic” and“continuous” of the test scenarios,and improve the efficiency of system fault detection in the meantime.Finally,in order to realize the rapid verification of the above-mentioned test scenario automatic generation and optimization methods,aiming at the weakness of the existing simulation testing tools mostly build three-dimensional test scenarios manually,this paper develops an automated simulation test platform to further improve test efficiency.Through the design of automatic test and evaluation execution program compiled in MATLAB,the data and resource sharing among the components of the platform are realized,and the full automation of the links of test and evaluation including the threedimensional modeling of the test scenario,the joint calling of simulation software,scenario initialization and case switching,as well as the result analysis and report generation is completed.The application results show that the platform can reduce the average execution time of the test and evaluation of the experimental scenario to less than5% compared with completing the above works manually,and the effectiveness of the automation simulation test and evaluation method of intelligent driving systems proposed in this paper is also verified in the platform.