Neural net based robot visuo-motor coordination via 3D perception

The authors present a new neural net approach to robot visuo-motor coordination. The approach, neurobotics, establishes neural-net based visual error servoing as a fundamental mechanism of implementing visuo-motor coordination. Visual error servoing is achieved by projecting the robot task space on the 3-D perception net representing the robot internal 3-D space, and generating an incremental change of the 3-D space arm configuration in the 3-D perception net based on potential field-based reactive path planning. The 3-D space arm configuration planned in the 3-D perception net is then translated into the corresponding joint angles through the arm kinematic net. The arm kinematic net is formed based on a radial-based competitive and cooperative network. Simulation results are shown