Study On ACC Performance Test And Analysis Based On Virtual Scene

Nowadays,every country is working on self-driving car technology.Some high-end models are equipped with the autonomous driving function represented by ADAS.However,in the research and development stage,limited by the disadvantages of high risk,high cost and low efficiency of public road test,it will greatly increase the difficulty of research and development.To solve this problem,many institutions at home and abroad have set up their own closed test sites.However,the cost of setting up a closed test site is high,and the test scene is relatively single.Therefore,the traditional automobile test methods have been unable to meet the test requirements of self-driving cars.With the advantages of high efficiency,low cost,rich test scenarios and low test risks,virtual scene test has gradually become a research focus of self-driving vehicle test technology.Based on the extraction and analysis of natural driving data,this paper proposes a virtual test scenario construction method for adaptive cruise system testing.Through the simulation experiment,the scene parameters that affect the performance of ACC system are analyzed and some design opinions of ACC system are put forward.The main work is as follows:1.According to the natural driving data,ACC boundary conditions were set and characteristic parameters for ACC test scenarios were proposed.The data sources of test scenarios can be roughly divided into three types: real data,simulated data and expert experience.The test scenario based on the natural driving data can guarantee the authenticity and richness of the scenario.The original natural driving data segment collected by various research institutions contains all kinds of complex environmental information,so it is necessary to conduct data mining from the natural driving data segment according to the test requirements.Aiming at the application scenarios of ACC system,500 tailgate scenarios were extracted from the natural driving data segment by setting boundary conditions.Then,based on the characteristics of ACC system,12 scene parameters of 5 categories were selected as the key parameters to construct ACC virtual scene,in which the acceleration of the front car was the time series data.By establishing the front car model,the acceleration of the front car was discretized and represented by 6 variables after discretization.2.Based on data mining,build a representative test scenario with statistical knowledge.Firstly,the contribution rate of each component was calculated by principal component analysis,and 10 principal components with a total contribution rate of 80% were selected to reduce the data dimension.Then,by combining the systematic clustering method and the hierarchical clustering method in the clustering method,the samples were initially clustered by the class average method,and the number of categories was determined to be 6 by combining the inconsistency coefficient.Next,the k-means clustering method was used to classify 500 tailgating scenarios,and 6 representative test scenarios were obtained.Through the significance analysis,the significance parameters of each kind of scene are obtained.The scenario is constructed with the significance parameter as a constant and the non-significance parameter as a variable.3.For quantitative variables in scene parameters,each parameter has a value interval.In order to illustrate the effect of different values on the test results.Five evenly distributed points are selected from the value interval corresponding to each certain number of parameters.Based on the simulation test software Prescan,the virtual scene was built.The parameters of road,traffic participant,traffic indication,sensor and weather are modeled and set respectively.The radar detection range is set to 200 m.For the control parameters of the vehicle,the control model built in Simulink is used to control the vehicle and the target vehicle.The first kind of scene is the city traffic light intersection.The second type of scene is the suburban road in front of the car and from the car at a faster speed.The third scene is the wet road.The fourth scene is the deceleration of the vehicle in front of the expressway.The fifth scenario is the lane change insertion of adjacent lanes.The sixth scene is the front car lane change cut out.4.In order to conduct follow-up tests,an ACC system based on model predictive control theory was established.The system adopts hierarchical design,and the variable headway strategy is adopted in the spacing strategy.In the upper design,the corresponding prediction model and objective optimization function are established based on the MPC theory and the dynamic relationship of the workshop.In the lower design,according to the input and output requirements,the engine inverse model and brake inverse model are established.In order to make a more intuitive evaluation of ACC system,three test indexes of tracking,comfort and safety were designed.Finally,the ACC system model was built based on Matlab/Simulink,and the scenarios set up in Prescan were combined to form a joint test platform.5.Orthogonal experimental design and result analysis of ACC performance test.In order to reduce the number of experiments and reduce the experiment cost,the simulation test scheme was designed by orthogonal experiment design method,and 47 experiments were carried out by orthogonal table with 10 factors and 5 levels.The simulation results show that the scenario construction method for ACC system virtual test given in this paper can effectively test ACC system and provide corresponding Suggestions for ACC system design based on the test results to improve the design.The simulation results based on virtual scene show that the relative distance,h0 and h1 have great influence on the performance index in various scenarios.In this paper,h0 represents the driver style of the car in front and h1 represents the effect of the acceleration at the current moment on the acceleration at the next moment.Therefore,when developing ACC system,it should focus on the spacing control strategy.In the spacing control strategy,more influencing factors should be considered to make the output expectation more flexible with the workshop spacing.In addition,for different application scenarios,more detailed driver behavior characteristics should be studied.For different application scenarios,a variety of system working modes should be designed.ACC system with single function cannot satisfy the complex traffic environment,so it is necessary to select the appropriate working mode through the corresponding switching logic according to the actual situation.