An Adaptive Algorithm for Position and Linear Velocity Estimation of Mobile Robots Using Inertial Sensors and Omnidirectional Vision

This paper presents a novel and effective method for estimating the position and linear velocity of a mobile robot in real-time in unknown environments by fusing visual information from an omnidirectional vision system with the measurements of an inertial sensor system. Based on a new projection model of the omnidirectional vision system, we propose a novel adaptive algorithm, similar to the Slotine-Li algorithm in model-based adaptive control, to estimate the position and linear velocity of the robot on-line using continuously tracked natural feature points in image sequence, the robot's acceleration, angular velocity and orientation measured by inertial sensor system. It is proven that our algorithm leads to global exponential convergence of the estimation errors of the robot's position and linear velocity to zero. Experiments demonstrate the performance of the proposed algorithm.