Controlled Diversity with Preference : Towards Learning a Diverse Set of Desired Skills
This addresses the issue of learning safe and human-aligned skills in unsupervised reinforcement learning, which is incremental as it builds on existing diversity methods by adding human guidance.
The paper tackles the problem of autonomously learning diverse behaviors in reinforcement learning without extrinsic rewards, which often results in unsafe or misaligned skills, by proposing Controlled Diversity with Preference (CDP) to incorporate human preferences and ensure desirable skill discovery, demonstrating its ability on 2D navigation and Mujoco environments.
Autonomously learning diverse behaviors without an extrinsic reward signal has been a problem of interest in reinforcement learning. However, the nature of learning in such mechanisms is unconstrained, often resulting in the accumulation of several unusable, unsafe or misaligned skills. In order to avoid such issues and ensure the discovery of safe and human-aligned skills, it is necessary to incorporate humans into the unsupervised training process, which remains a largely unexplored research area. In this work, we propose Controlled Diversity with Preference (CDP), a novel, collaborative human-guided mechanism for an agent to learn a set of skills that is diverse as well as desirable. The key principle is to restrict the discovery of skills to those regions that are deemed to be desirable as per a preference model trained using human preference labels on trajectory pairs. We evaluate our approach on 2D navigation and Mujoco environments and demonstrate the ability to discover diverse, yet desirable skills.