Research On Path Planning Method Based On Improved RRT Algorithm

Avoidance obstacle path planning has been intensively investigated due to its wide application,such as autonomous mobile robots,unmanned aerial vehicle(UAV),and communication route.Among the various algorithms proposed for the path planning problem,the rapid-exploration random tree(RRT)algorithm has shown its potential and been widely used,considering its advantages including high planning efficiency,strong adaptability to dynamic environment,high dimensional availability and complete probability.A mathematical model for obstacle avoidance path planning based on configuration space is utilized in this thesis.Then an improved adaptive goal biasing-RRT(AGB-RRT)algorithm based on the mathematical model mentioned above is designed to address the problems of RRT algorithm,i.e.,the redundancy of node extension caused by random search and the low efficiency due to the lack of target in search direction.The AGB-RRT algorithm is also extended to the asymptotic optimal fast random search tree(RRT*)algorithm.The specific research content as follows:(1)An improved strategy based on bias of sampling point towards object point is proposed.All the sampling points is biased towards the object point first before selecting their parent-nodes.Therefore,all the extension directions of the rapid-exploration random tree will tend towards the goal point,addressing the non-directionality of the RRT algorithm and improving the efficiency of RRT algorithm.(2)An improved strategy for step adjustment of adaptive weight is introduced.The step weight can be adjusted dynamically according to the depth of extension node and the distance between the extended node and the target node.Subsequently,the step size during expansion can adaptively change,resulting in the decrease of the number of sampling points and the improvement of the efficiency of the RRF algorithm.Two-dimensional and three-dimensional simulation experiments of RRT algorithm?AGB-RRT algorithm and AGB-RRT* algorithm are carried out in the simulation environment based on R tree in this thesis.The simulation results show that the AGBRRT algorithm and AGB-RRT* algorithm are improved compared with RRT and RRT*algorithm in sampling points and convergence speed.The effectiveness of the improved strategy is also proved.