Decoupling Task and Behavior: A Two-Stage Reward Curriculum in Reinforcement Learning for Robotics
This work provides a more robust and effective method for designing reward functions in robotic control, which is a common bottleneck for practitioners in the field.
The paper addresses the challenge of designing effective reward functions in robotic reinforcement learning by proposing a two-stage reward curriculum. This method decouples task-specific objectives from behavioral terms, first training on a simplified task-only reward for exploration, then introducing the full reward with auxiliary behavioral terms. The approach substantially outperforms baselines trained directly on the full reward and shows higher robustness to reward weightings.
Deep Reinforcement Learning is a promising tool for robotic control, yet practical application is often hindered by the difficulty of designing effective reward functions. Real-world tasks typically require optimizing multiple objectives simultaneously, necessitating precise tuning of their weights to learn a policy with the desired characteristics. To address this, we propose a two-stage reward curriculum where we decouple task-specific objectives from behavioral terms. In our method, we first train the agent on a simplified task-only reward function to ensure effective exploration before introducing the full reward that includes auxiliary behavior-related terms such as energy efficiency. Further, we analyze various transition strategies and demonstrate that reusing samples between phases is critical for training stability. We validate our approach on the DeepMind Control Suite, ManiSkill3, and a mobile robot environment, modified to include auxiliary behavioral objectives. Our method proves to be simple yet effective, substantially outperforming baselines trained directly on the full reward while exhibiting higher robustness to specific reward weightings.