Research On Path Planning And Tracking Control Of Intelligent Vehicle With Lidar

With the rapid development of artificial intelligence,intelligent vehicles have gradually permeated various fields of daily life.However,most of the current research about intelligent vehicles mainly focuses on structured roads such as highways rather than park roads such as campus,industrial parks and other areas.Park road is characterized of simple working conditions,low driving speed,static obstacles and easy landing applications.Therefore,research on park road is conducted in this paper,in which lidar point cloud registration,path planning and tracking control methods for intelligent vehicle are proposed,and moreover,verified through simulation and vehicle test platform.The main research contents of this paper are as follows:(1)For the point cloud data of vehicle lidar,an improved FPFH-ICP point cloud registration method is proposed.First,Voxel Grid filter and Statistical Outlier Removal filter are employed for preprocessing.Second,point cloud features are extracted by Fast Point Feature Histograms(FPFH),and initial registration is carried out based on Sample Consensus Initial Alignment(SAC-IA).Then,K-D tree is established and the Euclidean distance threshold and normal vector threshold are added to the traditional Iterative Closest Point(ICP)registration to accomplish accurate registration.Finally,experiment is carried out,of which the results exhibit that the root mean square error and registration time of the improved FPFH-ICP registration are reduced by 7.56% and 41.22%,respectively,compared with ICP registration,and by 30.28% and 18.95%,respectively,compared with Point Feature Histograms(PFH)registration.Therefore,regarding registration efficiency and accuracy,the improved FPFH-ICP registration is proven to be superior to other methods,guaranteeing positioning performance for path planning.(2)A multi-objective path planning method for intelligent vehicle based on Sparrow Search Algorithm(SSA)is proposed.Firstly,the transformation between Frenet coordinate system and Cartesian coordinate system is deduced,and popular path planning models are analyzed,yielded that the fifth-order polynomial path planning model has better performance.Secondly,in view of the lack of global information in park scenes,a method of long-short term path planning is proposed.Combining the fifth-order polynomial and the improved Artificial Potential Field(APF)method,the potential field model for in lane,between lanes and the obstacle are established to obtain the candidate cluster of obstacle avoidance trajectory.Thirdly,collision detection is carried out on the candidate trajectories through the vehicle rectangular model.Then,taking safety,comfort and lane change efficiency as evaluation indexes,the objective function is constructed,and the SSA is adopted to obtain the comprehensive optimal obstacle avoidance trajectory.Finally,different obstacle avoidance scenarios are set and simulated,and the results manifest that the SSA can better ensure the real-time performance of obstacle avoidance trajectory without reducing the solution accuracy.(3)A trajectory tracking control method of intelligent vehicle based on step-wise prediction is proposed.Considering the low driving speed of intelligent vehicle in park road,the vehicle kinematics model is established,and the turning angle of front wheel and driving speed are set as control variables.Secondly,in order to improve the real-time performance of tracking control,the vehicle kinematics model is linearized to construct the linear time-varying model of intelligent vehicle.Then the method of step-wise prediction is proposed.Multiple prediction models are constructed in the predictive time domain for tracking control with appropriate objective functions and constraints.Finally,comparisons of the proposed method with Linear Quadratic Regulator(LQR)control and Model Predictive Control(MPC)under different working conditions is evaluated,revealing that the method based on step-wise prediction under different driving speeds has better tracking accuracy while meets the real-time requirements of intelligent vehicles.(4)The joint simulations and real-world experiments are carried out for the proposed path planning and tracking control method.Firstly,a simulation platform based on Prescan/ Simulink is built,and different park scenes are established to verify the simulation effect of the proposed path planning and tracking control method.Then an experiment platform is built to verify the proposed point cloud registration,path planning and tracking control method in campus.The simulation and vehicle experiment results show that the proposed point cloud registration algorithm has advantages in registration accuracy and efficiency,and the paths planning and tracking control algorithm provide better safety and comfort in park scenes under the premise of ensuring real-time performance.