Research On Path Planning Method Of Driverless Vehicle Based On Improved RRT Algorithm

Driverless vehicles can not only reduce the incidence of traffic accidents,but also improve the efficiency of car travel,so they are widely concerned by scientific research institutions,automobile companies and Internet companies.As one of the core technologies of driverless vehicles,path planning technology has always been a difficult and hot topic in the field of driverless technology.Path planning for driverless vehicles requires consideration of two aspects.On the one hand,driverless vehicles need to plan a feasible path that satisfies the kinematic constraints of the vehicle in real time in response to a rapidly changing traffic environment.On the other hand,the driverless vehicles path planning algorithm should have better versatility and be able to successfully perform path planning in a variety of different traffic driving scenarios.However,the existing solutions still have many shortcomings in solving the above problems.For this reason,the research has been improved.The specific research contents are as follows:(1)The RRT(Rapidly Random-exploring Trees)algorithm is used for driverless vehicle path planning,and the basic RRT algorithm needs to be improved and optimized.The basic RRT algorithm avoids the modeling of the environment space by randomly sampling the path of the environment space,and the planning efficiency is high,but the sampling expansion is blind and the generation path is not optimal due to the random sampling mode.Firstly,the artificial potential field method is used to generate the extended domain that guides the improvement of the RRT algorithm,and the blindness of the improved RRT algorithm sampling expansion is reduced.Then,using the sampling expansion strategy of the biased target,the strategy is extended bidirectionally from the starting node and the target node,when generating a new node,it is necessary to judge whether it can reach the new collision detection strategy of the target node without collision,and reduce the redundant node and path length.And the pruning strategy of unnecessary steering,proposes an improved RRT path planning algorithm to improve the rate of path planning.(2)Aiming at the problem of discontinuous curvature of polygonal path generated by improved RRT algorithm,the trajectory shape was optimized.Based on the 4thorder Bézier curve,the orbital optimization is performed on the segmentation path generated by the improved RRT algorithm.The optimization includes path point segmentation,Bézier curve parameterization and addition of constraint optimization to generate vehicle dynamics,initial state,target state and Feasible path with continuous bounded curvature and other constraints can solve the problems of path zigzag and discontinuity of curvature that still exist after the RRT algorithm generates the path.(3)Aiming at the problem that the track shape optimization path does not satisfy the feasibility requirement of driverless vehicle,the action optimization of the path is carried out.Based on Frenet coordinate system,this paper optimizes the trajectory after the above trajectory optimization.By decoupling the trajectory into horizontal and vertical independent optimization,and then jointly evaluating the two direction cost functions,the trajectory sequence satisfying the executable requirements of the driverless vehicle is obtained.(4)Based on the planning time,path length,sampling points and path curvature index,the simulation results show that the indexes of the algorithm are obviously improved,and the real vehicle test under low speed condition is validated by the improved RRT algorithm.