Constraints as Rewards: Reinforcement Learning for Robots without Reward Functions
This work addresses the problem of tedious reward tuning for roboticists, offering an incremental improvement by reformulating constraints to simplify multi-objective balancing in reinforcement learning.
The paper tackles the challenge of reward engineering in reinforcement learning for robotics by proposing Constraints as Rewards (CaR), which uses constraint functions instead of reward functions to automatically balance multiple objectives via Lagrange multipliers, and demonstrates its success in generating standing-up motions for a six-wheeled-telescopic-legged robot where manual reward design fails.
Reinforcement learning has become an essential algorithm for generating complex robotic behaviors. However, to learn such behaviors, it is necessary to design a reward function that describes the task, which often consists of multiple objectives that needs to be balanced. This tuning process is known as reward engineering and typically involves extensive trial-and-error. In this paper, to avoid this trial-and-error process, we propose the concept of Constraints as Rewards (CaR). CaR formulates the task objective using multiple constraint functions instead of a reward function and solves a reinforcement learning problem with constraints using the Lagrangian-method. By adopting this approach, different objectives are automatically balanced, because Lagrange multipliers serves as the weights among the objectives. In addition, we will demonstrate that constraints, expressed as inequalities, provide an intuitive interpretation of the optimization target designed for the task. We apply the proposed method to the standing-up motion generation task of a six-wheeled-telescopic-legged robot and demonstrate that the proposed method successfully acquires the target behavior, even though it is challenging to learn with manually designed reward functions.