Develop And Implement Of Autonomous Driving Simulation Svstem Based On Real Traffic Scenarios

In autonomous driving system testing,compared with real road tests,simulation testing helps to reduce economic costs,quickly verify and optimize autonomous driving algorithms,and promote the development of autonomous driving technology.Considering the significance and importance of automatic driving simulation testing,and considering the realistic reduction of simulation scenarios,this research developed and implemented automatic driving simulation system based on real traffic scenarios.And studied the vehicle global routing planning algorithm and simulation test on structured and unstructured road scenarios respectively.This research first developed the communication architecture of autonomous driving simulation,so that the data message packets published or subscribed by the simulator can be received or sent by the simulation system,so as to ensure the timeliness of the simulation system.Then,the interface of behavior decision-making,path planning and chassis control modules in the simulation system are developed,and a complete automatic driving simulation system is realized in combination with the automatic driving simulator.Finally,the end-to-end closed-loop simulation test shows the availability of the autonomous driving simulation system developed in this study.At the same time,considering that the global routing planning algorithm for autonomous vehicles is the primary guarantee for the smooth driving of autonomous vehicles,this research developed global routing planning algorithm and simulation tested.Under the structured road,analyze the high-precision map information required by the autonomous vehicle,and use the information query interface and the lane connection relationship to develop the A*global routing algorithm.The algorithm first obtains the vehicle from the starting lane to the target lane.Then,according to the vehicle’s driving intention and the lane center line sampling,the final global routing planning result is obtained.Under the unstructured road,for the problem that there is no obvious road structure and road network characteristics,the lane connection relationship cannot be established,this research developed an improved A*algorithm based on dynamic grid(Dynamic Lattice A*Route Planning,DLRP-A*),the algorithm uses the idea of combining local dynamic and global dynamic grid division,iterative loop planning to obtain the global routing planning result of the vehicle.The simulation and test results show that under the structured road,the global routing planning of the vehicle can be realized by relying on the lane connection relationship.The simulation test in the simulation system proves the availability of the planning algorithm.Under unstructured roads,the DLRP-A*algorithm reduces the number of grid divisions by an average of 36.13%compared with the global grid A*algorithm,which solves the problem of too many planning grids.In the large-scale map scenario,while ensuring the planning accuracy,the planning speed of the DLRP-A*algorithm after multiple iterations increases,and the planning time is lower than that of the non-global dynamic grid division algorithm.