Scenario Library Module Design And Microservice Architecture Realization For Automatic Driving Simulation Test

In recent years,with the upgrading of automobile intelligence and electrification,the driver assistance systems have widely used in the automobile market on a large scale.In the scientific and technological revolution represented by artificial intelligence and newgeneration communication technology,as one of the best carriers for the new generation of information technology integration applications,automobiles are accelerating their transformation to intelligence and networking.But self-driving cars must undergo rigorous and adequate testing before they are actually put into commercial use.At present,the test mainly adopts the method of road test,which has a series of problems such as legal and regulatory restrictions,insufficient testing,high degree of danger,long time consumption,and high cost.It also makes the global autonomous driving industry chain still facing challenges,until the emergence of autonomous driving simulation testing,relevant researchers began to shift from road testing to simulation testing.From the current state of research,the simulation test based on the scene library is a feasible route to effectively solve the dilemma of autonomous driving and testing,which has been widely recognized by the autonomous driving industry.Therefore,how to design and construct the scene library effectively has become the hotspot of current research and development.In this thesis,the scene library module for automatic driving simulation test is designed and developed.Starting from the understanding and deconstruction of the driving scene,the scene data clustering is performed by extracting the scene feature elements,and the scene library is simulated and constructed by analyzing the data results,thereby improving the authenticity and reliability of the scene library.Finally,the relevant scene library construction process,simulation construction results and microservice architecture technology are combined to build a scene library system.The main work of this thesis has the following four aspects:（1）The scene library module is designed from the original scene data to the logical scene data,and then to the simulation scene,forming a closed loop of simulation test and scene library construction around the data layer,scene layer and application layer,and based on the driving simulator platform built by the laboratory project,the data of vehicle-to-vehicle intersection driving collision scenarios are obtained as the data samples studied in this thesis.（2）Starting from scene understanding and deconstruction,analyze and extract scene feature elements.The driver’s risk perception parameter is selected to represent the degree of dangerous influence of scene feature elements in the scene.The scene correlation analysis showed that the speed of the host vehicle,the speed of the target vehicle,the distance between the two vehicles,lighting and weather were significantly correlated with the danger of the vehicle-vehicle intersection driving collision scene.（3）The range standardization method is used to standardize the clustering parameters to solve the difference of different parameter units and quantities.In order to ensure the clustering effect of the scene data,the K-means clustering algorithm and the hierarchical clustering algorithm are analyzed and compared,and the better hierarchical clustering analysis result is selected as the logical scene data required for the construction of the scene database,so as to obtain representative results.Sexual three-class vehicle-vehicle intersection driving crash scenario data.Based on MATLAB/Simulink,a joint simulation platform of Car Maker and r Fpro is built,and three types of vehicle-vehicle intersection driving collision simulation scenarios are simulated and constructed with the results obtained by cluster analysis.Taking the first type of vehicle-motor vehicle intersection driving collision scene as an example,the collision control test analysis is carried out,and the application of the simulation test scene in automatic driving is illustrated through the integration of mature adaptive cruise control（Adaptive Cruise Control,ACC）algorithm.The results show that the constructed simulation scene has the practicability and reliability of the ACC algorithm model.（4）A scene library system based on the microservice architecture technology combined with the related scene library construction process and simulation construction results is constructed.Mainly developed the user information management,role management,raw data management,simulation data management function modules,and from the perspective of the administrator account after logging in,the page effect display of the four functional modules of the scene library system and the test were carried out to verify system availability.