Research On Path Planning Of Self-driving Vehicle Based On Improved RRT Algorithm

In recent years,self-driving technology has been developing rapidly.As an important part of the self-driving system,the results of path planning algorithm have great impacts on the follow-up vehicle tracking accuracy,thus the stability and comfort of the vehicle were affected.The RRT algorithm was taken as the basic algorithm of local path planning in this thesis,and its defects in vehicle kinematics constraints and search efficiency were optimized and improved,so that it can output a reasonable driving path.Combining with the studies of path planning algorithm,several common path planning algorithms were analyzed in this thesis.Due to the probabilistic completeness and high adaptability to many mobile platforms of RRT algorithm,it was studied in particular.Firstly,based on vehicle kinematics modeling,aiming at the kinematic steering constraints,a new node expansion strategy for RRT algorithm was proposed in this thesis,which was taking vehicle minimum turning radius and minimum radius of passing circle into consideration to calculate whether new node₉0))is in minimum radius of passing circle and further ascertain that the path conforms the vehicle steering constraints and also eliminate nodes that do not meet constrains to ensure the final path conforming vehicle steering constraints.Secondly,aiming at the low search efficiency of RRT algorithm in complex scenes,combined the k-d tree data structure with RRT algorithm,the nearest point search algorithm was improved in this thesis,and used the nearest neighbor search of k-d tree in two-dimensional space to search the random point₍（69）（9）’s nearest point to narrow the search scope and improve the search efficiency.Finally,aiming at the problem that the path turned too many times and it was not easy to track,the path was smoothed and optimized through the fitting of cubic B-spline curve to make the output path of the path planning system more reasonable.In this thesis,the Simulink model was built by using the autopilot toolbox and vehicle controller in Matlab software,and the comparative simulation experiments of the improved RRT algorithm were carried out in many scenarios,and then the application ability of the improved algorithm in the actual scene was tested by using the intelligent car.The experimental results show that,compared with the standard RRT algorithm,the improved algorithm has the advantages of small heading angle,smooth path and high search efficiency,and the task of path planning in simulated road scenes and actual scenes can be better completed,and the driving needs of driverless vehicles are met.