Research On Automated Valet Parking Path Planning And Control Of Intelligent Vehicles Based On Simultaneous Localization And Mapping

Due to the urgent demand and relatively simple driving environment,autonomous valet parking is considered to be the most commercial prospect of high-level autonomous driving application,and has attracted the attention of major vehicle manufacturers and autonomous driving technology enterprises in China and abroad in recent years.However,the current intelligent vehicles have problems such as large indoor positioning error and insufficient path planning ability in complex scenarios.It is necessary to carry out researches on key technologies such as positioning and path planning for autonomous valet parking to improve its stability and reliability.Therefore,the research will focus on autonomous valet parking,simultaneous localization and mapping,path planning and path tracking control to improve the positioning accuracy and path planning efficiency of intelligent vehicles in indoor parking lots.The specific research contents are as follows:First of all,based on the underground parking lot parking environment,the lidar simultaneous location and mapping method was proposed,respectively,to carry out the classification of radar point cloud,such as ground point cloud filtering.Design extraction and matching strategy based on point cloud line characteristics and plane characteristics,put forward the point cloud loopback detection and position optimization based on graph optimization,and through the implementation of iteration,precision map was built.The probabilistic method is used to construct three-dimensional grid map for vehicle navigation and location.Finally,the validity of the proposed method is verified by real parking lot location test.Secondly,A hybrid A* path planning method considering vehicle kinematics constraints was proposed.The cost function was designed to improve vehicle driving comfort by considering steering wheel angle,gear shift and reversing,and the Reeds-Shepp curve was introduced to make the reversing path be considered in the global path planning to correct vehicle heading angle.At the same time,a two-section tangent arc parking entry path planning method was proposed to complete the vertical and parallel parking scenario planning.Finally,the effectiveness and real-time of global path planning and local parking entry path planning were verified by simulation of typical parking lots.Then,the minimum capture method is used to fit and smooth the discrete waypoints planned.Model predictive tracking controllers based on particle and kinematic models are designed respectively.Quadratic optimization is used to realize the real-time solution of model predictive controller.The results show that the proposed method can achieve stable and highprecision path tracking.Finally,a test platform for autonomous parking of intelligent vehicles,including lidar and high-performance calculator,was built to carry out autonomous parking tests of typical underground parking lots.The experimental results show that the autonomous parking system proposed in this paper can achieve accurate positioning by using simultaneous location and mapping method in the absence of GNSS signal,and accomplish autonomous valet parking tasks through path planning and tracking.