Micro-Interaction Analysis Of Human-Vehicle And Vehicle Motion Planning Research Under Typical Traffic Scenarios

Pedestrians are a vulnerable group among road participants,and pedestrian safety is an important topic in the study of autonomous driving.Aiming at the safety problem of interaction between autonomous vehicles turning at intersections and pedestrians on zebra crossings,a vehicle lateral and longitudinal decoupling obstacle avoidance motion planning algorithm for pedestrians with different crossing styles is proposed.For pedestrians,a pedestrian crossing model considering crossing styles is designed,and the pedestrians driven by the model are used as vehicle obstacle avoidance objects;For vehicles,a lateral and longitudinal decoupling obstacle avoidance motion planning algorithm is designed,and the effectiveness of vehicle lateral and longitudinal decoupling obstacle avoidance motion planning for pedestrians is verified in a joint simulation platform.The main research contents of this article are as follows:(1)A pedestrian crossing model considering crossing styles is proposed,in which unmanned aerial vehicles(UAVs)are used to capture human-vehicle interaction videos and demarcate them into datasets.Then,an integrated learning algorithm based on decision trees is used to analyze the datasets to obtain the crossing styles of pedestrian samples;The two wheel differential motion model and the dynamic window trajectory search algorithm are used as the control carrier and the path search algorithm for the pedestrian crossing model,respectively,and the crossing style is used as the parameter setting basis for the pedestrian crossing model.(2)Based on the Frenet coordinate system,a vehicle lateral and longitudinal decoupling obstacle avoidance path planning algorithm is designed.The pedestrian under the control of the pedestrian crossing model is regarded as the vehicle obstacle avoidance object.The dynamic programming and quadratic programming algorithms are used to calculate the lateral obstacle avoidance path and longitudinal speed planning trajectory for the vehicle lateral(s-l)and longitudinal(s-t)spaces.(3)The vehicle obstacle avoidance path tracker is constructed based on the vehicle dynamics model,the lateral obstacle avoidance path tracker is designed using the Model Predictive Control algorithm,and the longitudinal velocity trajectory tracker is established using the Velocity-Position Dual PID(Proportion Integral Derivative)control algorithm.(4)Based on the interaction scenario between turning vehicles at intersections and pedestrians crossing the street,Pre Scan-Simulink-Car Sim joint simulation environment is established to set pedestrian obstacles with different crossing styles and sudden changes in movement status for vehicles,to verify the effectiveness of the vehicle lateral and longitudinal decoupling obstacle avoidance motion planning model.In the conventional pedestrian crossing simulation model,most pedestrians are static and known preset obstacles without interaction with vehicles.The innovation of this topic is to establish an independent motion model for pedestrians,and use the future trajectory of pedestrians outputted from the motion model to generate the obstacle st region in the vehicle speed planning s-t space,so that pedestrian motion is fed back to the vehicle avoidance,while pedestrians can also avoid vehicles,Make human-vehicle interaction more realistic and test the robustness of vehicle motion planning algorithms.Simulation experiments show that the proposed lateral and longitudinal decoupled obstacle avoidance motion planning model can control the steering vehicle to complete pedestrian obstacle avoidance under the premise of ensuring pedestrian vehicle safety and dynamic constraints by analyzing the human-vehicle interaction motion data and vehicle dynamics performance.