Obstacle detection and avoidance on a mobile robotic platform using active depth sensing

Abstract

The ability to recognize and navigate surrounding environments free from collision with obstacles has been at the forefront of mobile robotic applications since its inception. At the price of nearly one tenth of a laser range finder, the Xbox Kinect uses an infrared projector and camera to capture images of its environment in three dimensions. The objective of this thesis was to investigate if the Xbox Kinect can be utilized to detect thin or narrow obstacles that are often invisible to the P3-DX mobile robotic platform. We present an algorithm to process and analyze point cloud data from the Xbox Kinect sensor and transform it into a two-dimensional map of the surrounding environment for further use with the P3-DX. Obstacle avoidance scenarios were then performed using two separate algorithms: a narrow corridor following algorithm and a potential fields algorithm. The results demonstrate that in a structured testing environment, the Xbox Kinect can be used to detect and avoid narrow obstacles that are not immediately recognized by the onboard sonar array of the P3-DX.