Visually-extended vector polar histogram applied to robot route navigation

摘要

This paper presents a visual extension of the vector polar histogram (VVPH) for safe and fast real-time obstacle avoidance and navigation in route environments. VVPH is based on an enhanced vector polar histogram method (VPH+). VPH+ is a powerfully performing method in obstacle avoidance and navigation. It depends on highly accurate sensors, such as a laser range finder. The accuracy of visual data for route environments is enhanced using the hidden Markov model (HMM) and extended Kalman filter (EKF). HMM yields faster estimation results in route environments, preventing the smoothing effect of the EKF causing unexpected errors in a situation where features disappear suddenly. The performance of the proposed method, VVPH, is verified via computer simulations and experimental setups, in terms of the total travelling time and the accuracy of the visual histogram.