Teleoperation dexterity can be enhanced by incorporating full or partial autonomous execution of subtasks. We focus on tasks with dynamic robot-environment interactions which require velocity matching contacts with and grasping of moving objects. Such tasks present difficulties for full teleoperation due to obstruction of visual feedback to the operator, requirements for high master-slave bandwidth, or the need for excessive amount of sensory feedback to the master. New sensor-based geometric control algorithms are developed and implemented on a SARCOS seven DOF robot arm with a three finger hand. The controller relies on data from a finger embedded Proximity Sensor Network, processed by a nonlinear recursive algorithm for filtering, data fusion and object localization.
J. Damianakis, G. Petryk, M. Buehler