Optimal Potential Shaping on SE(3) via Neural Approximators

Potential shaping in the workspace of a fully actuated robotic system has been investigated for quadratic potentials on SE(3).

This assignment aims at generalizing the resulting controller to those resulting from a more general class of potentials on SE(3). The choice of the specific controller will be encoded in the output of an optimization problem maximizing the performance index characterizing the task to be executed.