Constraints as Rewards: Reinforcement Learning for Robots without Reward Functions

TL;DR

This paper introduces Constraints as Rewards (CaR), a novel reinforcement learning approach that uses constraints instead of reward functions to automatically balance multiple objectives in robotic tasks, reducing the need for reward engineering.

Contribution

The paper proposes CaR, a method that formulates task objectives as constraints and employs Lagrangian optimization, enabling automatic balancing of objectives without reward tuning.

Findings

Successfully applied to robot standing-up motion generation

Achieved target behaviors without manual reward design

Demonstrated effectiveness on complex robotic tasks

Abstract

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.