An Extended Kalman Filter-Based Robot Pose Estimation Approach with Vision and Odometry

Visual cameras and encoders are usually equipped on mobile robotic systems. In this paper, we present a robust extended Kalman filter-based pose estimation approach by fusing the information from both the onboard camera and encoders. Different from existing works, the system state is chosen in a new simplified way, including the robot pose and the depth of feature points. Moreover, a new observation model is formulated and the corresponding Jacobian matrix is derived. A robust feature association approach with an outlier removing mechanism is proposed. Experimental results are provided to demonstrate the effectiveness of the proposed approach.