Smart Car Path Planning Based On Improved RRT

Intelligent vehicles have important research value because of its huge advantages and broad application prospects in both civilian and military fields.Smart car is a complex system that contains environmental perception,decision-making and controlling.Path planning is the key part of it.This paper aims at the RRT algorithm in the path planning of smart cars.Based on the previous work,further research is done and several improvements are proposed.The main research work and innovation are as follows:(1)Two improved methods,improved RRT based on nearest available point and Goal-biasing RRT based on maximum corner constraint are proposed and implemented.Based on the research and application of the basic RRT algorithm,this paper proposes a node extension method that preferentially expands the nearest available point and converges to the endpoint,then,optimizes the resulting path with pruning function,which is aimed to deal with the shortcomings of large amount of computation and blind random search in the process of RRT node expansion.The Goal-biasing RRT weakens the blindness of the RRT when expanding the nodes to some extent.Based on the vehicle motion model plus the maximum turning angle constraint of the vehicle,the Goal-biasing RRT can make the algorithm converge more quickly and get more accordant with the actual situation of the vehicle,the planning result of motion trajectory.(2)A rapid RRT path planning method based on artificial potential field guidance is proposed and implemented.Referring to the guidance of the repulsive force generated by obstacles and the gravitational force generated by the target point to the autonomous vehicles in artificial potential field method,the uniform distribution probability model in RRT is replaced by the probability distribution model guided by artificial potential field.This is to change the blind random search method in RRT to the search method along the descending direction of potential field.In the meantime,the extended sub-nodes are limited to 8 neighborhood grids.According to the weight of each neighborhood grid,a random extension point is randomly selected according to the roulette selection method.This improved method only needs to calculate the weights of eight neighborhood grids when nod is extended,which has the advantage of less complex computation.Also,the search method along the descending direction of the potential field also accelerates the convergence of the algorithm.With these two redundancy and xy-optimized post-processing methods,path quality is further improved.(3)An RRT path optimization method based on ant colony algorithm is proposed and implemented.RRT randomness will lead to the instability of the planning path,while the ant colony algorithm can almost certainly converge to the global optimal solution.Several RRT planning paths are taken as the primary population in the ant colony algorithm,and the length of the planning path is used as the main assessment criteria to set the number of pheromones.Poor paths eventually converge to an optimal path because in each iteration process the pheromones on the path evaporate quicker than accumulate and are gradually discarded by the ant colony,which completes the optimization of RRT result path.(4)A smart car path planning system platform and its hardware and software subsystems are introduced.The effectiveness of the improved algorithm in this paper is verified using this platform as an experimental platform.