Progressive Reinforcement Learning with Distillation for Multi-Skilled Motion Control
This work addresses the challenge of multi-skilled motion control for robotics and simulation, but it is incremental as it builds on existing policy distillation and transfer learning techniques.
The paper tackles the problem of integrating multiple specialized skills in continuous control by extending policy distillation to continuous actions and introducing input injection for augmenting policies, resulting in a method that incrementally adds new skills and outperforms three baselines in simulated bipedal locomotion across terrains.
Deep reinforcement learning has demonstrated increasing capabilities for continuous control problems, including agents that can move with skill and agility through their environment. An open problem in this setting is that of developing good strategies for integrating or merging policies for multiple skills, where each individual skill is a specialist in a specific skill and its associated state distribution. We extend policy distillation methods to the continuous action setting and leverage this technique to combine expert policies, as evaluated in the domain of simulated bipedal locomotion across different classes of terrain. We also introduce an input injection method for augmenting an existing policy network to exploit new input features. Lastly, our method uses transfer learning to assist in the efficient acquisition of new skills. The combination of these methods allows a policy to be incrementally augmented with new skills. We compare our progressive learning and integration via distillation (PLAID) method against three alternative baselines.