A Kinematic Constrained Batch Informed Trees Algorithm With Varied Density Sampling for Mobile Robot Path Planning

Abstract

we proposed a novel Kinematic Batch Informed Trees algorithm (K-BIT*) to solve problems of the low efficiency, poor geometric smoothness and local optimum when conducting path planning for mobile robots. A variable density sampling strategy is designed which can automatically adjust the searching radius according to the complexity of environment to speed up the search efficiency of feasible paths. Meanwhile, a better escape approach by generating specific way points based on current obstacles information is also developed when detecting entrapment, and which can finally enhance the robot searching capability. With consideration of kinematic constraints, the algorithm ensures generated paths adhere to the robot's physical and dynamic characteristics while maintain smoothness and stability of the motion. Simulation and Experiment results demonstrate that compared to existing algorithms (BIT*, RRT*, and kinematic RRT*), the proposed method can provide superior optimization efficienc (minimum improvement of 23.94%) and higher success rates (nearly 100%) in kinds of complex environments.