Extracting Strong Policies for Robotics Tasks from zero-order trajectory optimizers

Extracting Strong Policies for Robotics Tasks from zero-order trajectory optimizers

Our paper Extracting Strong Policies for Robotics Tasks from zero-order trajectory optimizers got accepted for ICLR2021.

The work is a continuation of our last paper iCEM. We added a policy extraction scheme to learn a policy from the optimal trajectories generated by the zero-order CEM optimizer.

Check out the paper for more details.