Development of Obstacle Detection and Avoidance System Using Kinect Sensor

Abstract

This paper presents the development of a collision avoidance system for a Micro Air Vehicle (MAV). The sensor used for obstacle detection is Microsoft Kinect and the processing is done onboard the vehicle using a single board computer (Beagleboard). Microsoft Kinect is chosen as the path finding sensor due to its range, accuracy and level of detail all of which are appropriate for indoor applications. A program written in Python processes the data obtained from the Kinect depth sensor using OpenCV functions and Numpy and finds a window of safe passage closest to the current heading of the vehicle. To take care of the increased take-off weight associated with the use of Kinect a twin coaxial vehicle in tandem configuration is constructed using two off-the-shelf coaxial helicopter models. The goal of this research is to test and validate the algorithm proposed in this paper in gust free indoor environment. An Arduino UNO board is used onboard to drive motors and servos on the vehicle which are controlled based on manual wireless input through a wireless communication system and computations by Beagleboard. The results obtained in this paper demonstrate the capability of Kinect sensor as a complete solution for obstacle detection and avoidance. The performance of Beagleboard as the potential solution for onboard processing is also established.