Value constrained model-free continuous control

Steven Bohez, Abbas Abdolmaleki, Michael Neunert +3

The naive application of Reinforcement Learning algorithms to continuous control problems -- such as locomotion and manipulation -- often results in policies which rely on high-amp…