Research On Decision And Planning Of Multi-vehicle In Parking Scenarios

Automated valet parking is known as the L4 level autonomous driving technology that solves users’ "last mile freedom",and is the commercial breakthrough of L4 level autonomous driving technology.Automated valet parking can achieve a complete parking process that include finding parking spaces automatically,driving vehicles to the parking spaces,and parking.The research on this system is valuable to engineering application.Although automated valet parking has received widespread attention from the academic and industrial communities,there are still many issues to be resolved.Firstly,the industry mainly concentrates automated valet parking of single vehicles,which cannot handle the problem of multiple intelligent vehicles parking at the same time.Secondly,automated valet parking includes driving and parking,but there is no a reasonable way to split the two modes.Based on the basic researches of our group on automated valet parking,this paper researches the decision and planning algorithm of multi-vehicle in parking scenario.The main research contents of this paper are as follows:Based on the global analysis of motion planning and multi-vehicle cooperative parking,this paper proposes corresponding solutions for three key issues: vehicle priority allocation,parking spot priority allocation and vehicle motion mode design.For the priority assignment of vehicles,the vehicle that enters the parking lot first is set as the highest priority.It mainly considers the limitations of parking lot environment,and an overtaking trajectory is unreasonable in parking scenario.For the priority assignment of parking spot,this paper analyzes three priority allocation scheme,and selects the priority allocation scheme of gap priority finally.At the same time,this paper constructs an optimal parking planning mode division for a single vehicle,which divides the parking process of a single vehicle into queuing mode and maneuvering mode.In this way,trajectory planning can be carried out according to the different characteristics of the two modes separately.Thus,a single vehicle has the best path and speed when moving,which providing a good basis for multi-vehicle cooperation.Based on the research on multi vehicle collaborative parking path planning,this paper proposes a reasonable path planning method.Because planning the trajectory of multiple vehicles at the same time is a NP-hard problem,this paper planning the path and the speed separately.For the path planning of leading vehicle,this paper compares several planning schemes,and finally constructs a path planning scheme based on guided hybrid A *algorithm.In addition,the planned path is optimized by curve planning method.For the following vehicles,this paper designs a reasonable path planning scheme based on the principle of reusing the leading vehicle path as much as possible.A priority speed planning method for multi vehicle collaborative parking is proposed.Due to no other vehicles are parking in front of the leading vehicle,this paper uses S-curve for the leading vehicle speed planning,which does not need to consider dynamic obstacles.Speed planning of the following vehicles is relatively complicated.In this paper,the vehicles with higher priority is set as the dynamic obstacles of the vehicles with lower priority,and the dynamic obstacles are built into the ST diagram of the following vehicles.First,a feasible speed curve is searched in the ST diagram through A~* algorithm.Based on the feasible speed curve of following vehicle and the position of dynamic obstacles,a convex space is constructed,and then the feasible trajectory is optimized using quadratic programming to obtain a speed curve that has continuous speed and acceleration.Based on the decision planning method proposed in this paper for automated valet parking of multi vehicle,four multi vehicle collaborative interaction scenarios are constructed for simulation verification.This paper compiles the proposed speed planning and path planning scheme into matlab code,and analyzes the problem of multi-vehicle collaborative parking.This paper uses matlab code to realize the proposed speed planning and path planning scheme.Due to there is no movement interaction between most vehicles.Therefore,based on two-vehicle cooperation and three-vehicle cooperation this paper selects four typical working conditions for simulation verification,and finally confirms that the motion planning scheme in this paper is effective for multi-vehicle cooperative planning in parking scenario.