Estimation of Moving Obstacle Dynamics with Mobile RGB-D Camera

Abstract

With the increasing use of autonomous robots, the concept of obstacle detection and avoidance is becoming vital. While estimating true position of obstacles is a rudimentary prerequisite for avoiding them, to optimize the path, the robots should use the velocity and acceleration of obstacles to estimate their course and take the best path according to them. This paper presents an algorithm that calculates the position and estimates the dynamics of moving obstacles with an RGB-D camera mounted on a mobile robot. In this algorithm, the depth map is used to detect the obstacles and their position. Furthermore, velocity and acceleration of the obstacles are estimated using fusion of derivative of position and optical flow. To validate this algorithm, a series of tests are conducted and the results confirm the effectiveness of the proposed algorithm in obstacles detection and their dynamics estimation.