Terrain Obstacle Detection and Analysis using LIDAR

Introduction Autonomy is crucial in planetary robotics where extreme distance and limited resource availability make continuous teleoperation impractical. Environmental obstacles represent a fundamental hurdle, as natural hazards may prevent robots from accomplishing certain tasks by invalidating path plans. In the worst-case scenario, it may even result in physical harm to the robot. Valuable resources can be saved by allowing the robot to resolve most obstacles automatically and requiring manual intervention only when absolutely necessary. Current obstacle avoidance schemes are able to detect most monolithic obstacles easily by looking for signatures like the range gradient. However, many terrain hazards are missed because the negotiability of passable terrain is rarely considered. In this paper we outline the development of an obstacle detection system for rovers used in the PROSPECT project at CMU and a general approach to terrain analysis for robots in the lunar environment. The prototype system makes use of LIDAR technology as the primary mode of sensing and specifies a one-DOF actuation for the Sick LMS 200 sensor. Data gathered from the sensor are used to analyze the terrain for obstacles and generate an internal hazard representation of the world. Predicted obstacles are then assessed and assigned a certainty and threat value which will aid the robot in navigating the environment.