Path Planning and Obstacle Avoidance in CG Space of a 10 DOF Rover using RRT

Abstract

There has been a recent demand for algorithms to plan the motion on 3D terrain for applications in space exploration, rescue and relief, unmanned vehicles, defense applications etc. Conventional path planning algorithms in 2D cannot be used in 3D as the work space cannot be divided into obstacles and free space. A few algorithms have been proposed for articulated rovers in 3D and they all require optimization to find the wheel and ground terrain contact and hence cannot be used in real time. In this paper, a new method is proposed to find the path in the CG space of a 10 DOF rover without the need for optimization. This CG space planning method can operate in real time. The CG space is the collection of possible CG points of the rover on a given terrain that is similar to the C-space in robot motion planning. The terrain geometry used for generating the CG position of rover is obtained using a Microsoft Kinect V2. A multivariable optimization process is used to extract the CG locus data of the rover as a discrete point cloud to generate CG space. Then using RRT* algorithm, the feasible path to reach a goal location from an initial point avoiding obstacles has been found out. During the motion planning for 10 DOF rover, RRT* algorithm directly samples a node from the CG locus data. It searches globally for an optimal path via two optimizing features in the extend function. Simulations on different types of terrains with different obstacle shapes show the usefulness of the method.