A neural field approach for robot motion control

We introduce a biologically inspired approach for robot motion control. It is based on a so-called neural field which can be described by a nonlinear competitive dynamical system. Movement directions are assigned to the field's artificial neurons by employing codebook vectors. Due to the field's intrinsic dynamical properties the problem of reaching a goal under constraints can be solved efficiently. Our approach is validated by applications to local navigation and manipulator control.