Collision avoidance for a low-cost robot using SVM-based monocular vision

Collision-free navigation is an important problem in autonomous robots. In most of the applications, camera vision techniques using stereo-vision and laser scanners have been used. These techniques are not commercially viable for miniature robots due to size and computational limitations. Optical flow based models using monocular vision have shown promise in biomimetic systems to estimate depth information from a scene. In this paper, we propose an obstacle avoidance algorithm that learns optical flow patterns through an SVM classifier. Experimental results and simulation results are presented to validate our approach. The system can be used for indoors and outdoors without modifying the algorithm.