Learning Interpretable, High-Performing Policies for Continuous ControlÂ