Parrot: Data-Driven Behavioral Priors for Reinforcement Learning
This work tackles the problem of high data requirements in reinforcement learning for agents learning new tasks, providing a method for more efficient learning.
This paper addresses the extensive data collection required for reinforcement learning by proposing a method for pre-training behavioral priors from previously collected datasets. This pre-trained prior enables rapid learning of new tasks in robotic manipulation domains, outperforming prior works by a substantial margin.
Reinforcement learning provides a general framework for flexible decision making and control, but requires extensive data collection for each new task that an agent needs to learn. In other machine learning fields, such as natural language processing or computer vision, pre-training on large, previously collected datasets to bootstrap learning for new tasks has emerged as a powerful paradigm to reduce data requirements when learning a new task. In this paper, we ask the following question: how can we enable similarly useful pre-training for RL agents? We propose a method for pre-training behavioral priors that can capture complex input-output relationships observed in successful trials from a wide range of previously seen tasks, and we show how this learned prior can be used for rapidly learning new tasks without impeding the RL agent's ability to try out novel behaviors. We demonstrate the effectiveness of our approach in challenging robotic manipulation domains involving image observations and sparse reward functions, where our method outperforms prior works by a substantial margin.