QMP: Q-switch Mixture of Policies for Multi-Task Behavior Sharing

TL;DR

This paper introduces Q-switch mixture of policies (QMP), a framework for sharing behaviors across tasks in multi-task reinforcement learning to improve sample efficiency and performance.

Contribution

QMP provides a new, principled method for selectively sharing behaviors using Q-functions, enhancing existing multi-task RL algorithms.

Findings

QMP improves sample efficiency in various environments.

QMP outperforms alternative behavior sharing methods.

QMP offers complementary gains over popular MTRL algorithms.

Abstract

Multi-task reinforcement learning (MTRL) aims to learn several tasks simultaneously for better sample efficiency than learning them separately. Traditional methods achieve this by sharing parameters or relabeled data between tasks. In this work, we introduce a new framework for sharing behavioral policies across tasks, which can be used in addition to existing MTRL methods. The key idea is to improve each task's off-policy data collection by employing behaviors from other task policies. Selectively sharing helpful behaviors acquired in one task to collect training data for another task can lead to higher-quality trajectories, leading to more sample-efficient MTRL. Thus, we introduce a simple and principled framework called Q-switch mixture of policies (QMP) that selectively shares behavior between different task policies by using the task's Q-function to evaluate and select useful shareable behaviors. We theoretically analyze how QMP improves the sample efficiency of the underlying RL algorithm. Our experiments show that QMP's behavioral policy sharing provides complementary gains over many popular MTRL algorithms and outperforms alternative ways to share behaviors in various manipulation, locomotion, and navigation environments. Videos are available at https://qmp-mtrl.github.io.