Sensor fusing using a convex combination of two Kalman filters - Experimental results

In this work the mobile robot localization problem in an unknown environment is faced and a new version of the Extended Kalman filter is proposed to estimate the robot position and orientation. This new filter uses a convex combination of two filters estimating the same state variables. The first filter is based on measurements provided by robot on board distance sensors while the second one uses out of board distance sensors measurements. The resulting “Mixed” Kalman filter is designed to emphasize the qualities and overcome the defects of each used sensor. The proposed fusing technique has been tested in a real experimental framework using the robot Khepera III. The algorithm has been contrasted with other Extended Kalman filters, based on the on board and on the out of board sensors measurements, yielding to encouraging results.