Path Planning in Uncertain Environment With Moving Obstacles Using Warm Start Cross Entropy

This article studies the robot's path planning problem with observation uncertainties in an environment containing moving obstacles. The framework of partially observable Markov decision process is utilized to deal with the planning problem, in which the environment is characterized as a grid map. To design an efficient and solvable optimization problem, we design safer moving rules in the grid map and formulate the new cost functions, including the collision costs incurred by the static and dynamic obstacles. Furthermore, we propose a cross entropy algorithm with a new initialization scheme called the warm start cross entropy to solve the optimization problem with improved searching speed. The simulation and comparison studies validate the advantages of the proposed method in safety and computational speed. In addition, the hardware experiments on a robot platform verify the applicability and feasibility of the proposed path planning method.