AUTONOMOUS GROUND VEHICLE LOW-PROFILE OBSTACLE AVOIDANCE USING 2D LIDAR

Abstract

In future conflicts, the United States Marine Corps will augment the capabilities of the individual Marine with autonomous vehicles able to traverse a wide range of terrain. The objective of this thesis is to investigate the ability of an autonomous ground vehicle to use two-dimensional (2D) down-looking Light Detection and Ranging (LIDAR) to avoid low-profile obstacles. The robot performs three primary tasks: localization, obstacle detection, and control. Optical flow sensors and a Global Navigation Satellite System/Inertial Navigation System accurately localize the robot. The obstacle detection algorithm combines 2D LIDAR sweeps to form a three-dimensional (3D) map and classifies obstacles by terrain gradient. The control algorithm navigates around obstacles to waypoints with turns constrained by a minimum radius. In simulation, the control algorithm navigated several obstacle fields but was not generalizable for any arbitrary obstacle field. In experimental testing, the vehicle successfully navigated around low-profile obstacles including books, stairs, and cones. The test configurations included single obstacles, inside corners, and corridors. The system was proven capable of accurate localization, 3D obstacle detection, and navigation under the nonholonomic minimum-turn-radius constraint. Future research efforts could improve the vehicle by increasing the execution speed of the main control loop, adding 3D LIDARs, and improving the multi-point turn algorithm.