Visual servoing for fast mobile robot: adaptive estimation of kinematic parameters

This paper addresses the problem of estimating the kinematic parameters of a relative motion between a mobile robot and an unknown obstacle. The authors use the information issued from a camera coupled with a laser stripe. In the first part, an adaptive extended Kalman filter (AEKF) is proposed to perform this estimation, the performances of this method are displayed for a complex motion example realized in simulation. In the second part, two parallel implementation methods of this filter are investigated, the first is data oriented and the second one is based on the properties of the filter. In the third part, the authors describe the hardware implementation which has been realized on transputer boards and give the calibration results and performances of this system. Finally, a real application on a laboratory mobile robot is given.<>