Learnable Behavior Control: Breaking Atari Human World Records via Sample-Efficient Behavior Selection
This work addresses the exploration challenge in RL for Atari game agents, achieving state-of-the-art performance without sacrificing sample efficiency.
The paper tackles the exploration problem in deep reinforcement learning by proposing Learnable Behavioral Control (LBC), which enlarges the behavior selection space and uses learnable processes for behavior selection, achieving a mean human normalized score of 10077.52% and surpassing 24 human world records in the Arcade Learning Environment within 1B training frames.
The exploration problem is one of the main challenges in deep reinforcement learning (RL). Recent promising works tried to handle the problem with population-based methods, which collect samples with diverse behaviors derived from a population of different exploratory policies. Adaptive policy selection has been adopted for behavior control. However, the behavior selection space is largely limited by the predefined policy population, which further limits behavior diversity. In this paper, we propose a general framework called Learnable Behavioral Control (LBC) to address the limitation, which a) enables a significantly enlarged behavior selection space via formulating a hybrid behavior mapping from all policies; b) constructs a unified learnable process for behavior selection. We introduce LBC into distributed off-policy actor-critic methods and achieve behavior control via optimizing the selection of the behavior mappings with bandit-based meta-controllers. Our agents have achieved 10077.52% mean human normalized score and surpassed 24 human world records within 1B training frames in the Arcade Learning Environment, which demonstrates our significant state-of-the-art (SOTA) performance without degrading the sample efficiency.