Autonomous Manipulation Combining Task Space Control with Recursive Field Estimation

This paper proposes an autonomous manipulation strategy that enables an articulated manipulator to detect and locate a particular phenomenon occurring in the surrounding environment. More specifically, the sensor-driven task space control of an end-effector is combined with the field estimation and the target tracking in an unknown spatial field of interest. The radial basis function (RBF) network is adopted to model spatial distribution of an environmental phenomenon as a scalar field. Their weight parameters are estimated by a recursive least square (RLS) using the collective measurements from the on-board sensors. Then, the asymptotic source tracking has been achieved by the control law based on the gradient of the estimated field. Experimental results verified the effectiveness of the proposed scheme.