Bearing only visual homing: Observer based approach

Abstract

Many robotic applications such as pick and place, navigation, docking, etc. need to visit a known place as an intermediate task. Homing is a method to reach a known position and is challenging to address specially in a GPS denied indoor environment. The approach is to match the information collected at home and current locations. The information collected from images is rich in amount, however to extract the relevant information from images is challenging. This paper solves the homing problem using bearing only measurements as this information can be easily extracted from the features in an image. The observable system shows that the bearing only measurements avoid the need to know the locations of landmarks and instantaneous distance of the robot with respect to the home position. Instantaneous state of the robot is estimated using EKF (Extended Kalman Filter). Simulation results with real images from two data-sets show that the estimation of the robot's state is of the order of 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> cm in range, 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> radians in bearing and 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-16</sup> radians in orientation. Comparison with existing visual homing techniques show that the proposed technique has faster convergence of the robot's position to the home position due to good estimates.