Decoupling Task and Behavior: A Two-Stage Reward Curriculum in Reinforcement Learning for Robotics

TL;DR

This paper introduces a two-stage reward curriculum for reinforcement learning in robotics, decoupling task and behavior objectives to improve training stability, robustness, and performance across various environments.

Contribution

The authors propose a novel two-stage reward curriculum that separates task and behavioral objectives, enhancing training stability and robustness in robotic reinforcement learning.

Findings

Outperforms baseline methods trained on full rewards.

Improves training stability and robustness to reward weightings.

Effective across multiple robotic environments.

Abstract

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.