Fusion of discrete and continuous epipolar geometry with wheel and IMU odometry for localization of mobile robots

This work presents a novel sensor fusion implementation to improve the accuracy of mobile robot localization by combining multiple visual odometry approaches with wheel and Inertial Measurement Unit (IMU) odometry. Discrete and continuous homography matrices are used to recover position, orientation, and velocity from image sequences of tracked feature points. An IMU and wheel encoders also measure the linear and angular velocity of the robot. The camera's limited field of view is addressed by chaining vision-based motion estimates. As feature points leave the field of view, new sets are acquired. The discrete motion estimate is then reinitialized and chained to the previous state estimate. A Kalman filter fuses the wheel encoder measurements with those from visual and inertial measurement systems. Time varying matrices in the Kalman filter compensate for known changes in sensor accuracy, including periods when visual features cannot be reliably tracked. Experiments are performed to validate the approach.