A Closer Look at Invalid Action Masking in Policy Gradient Algorithms
This addresses a practical issue for researchers and practitioners in reinforcement learning, but it is incremental as it builds on existing masking techniques.
The paper tackles the problem of invalid actions in deep reinforcement learning for strategy games by investigating the practice of masking out invalid actions in policy gradient algorithms, showing theoretical justification and empirical importance as invalid action spaces grow, with experiments demonstrating performance improvements.
In recent years, Deep Reinforcement Learning (DRL) algorithms have achieved state-of-the-art performance in many challenging strategy games. Because these games have complicated rules, an action sampled from the full discrete action distribution predicted by the learned policy is likely to be invalid according to the game rules (e.g., walking into a wall). The usual approach to deal with this problem in policy gradient algorithms is to "mask out" invalid actions and just sample from the set of valid actions. The implications of this process, however, remain under-investigated. In this paper, we 1) show theoretical justification for such a practice, 2) empirically demonstrate its importance as the space of invalid actions grows, and 3) provide further insights by evaluating different action masking regimes, such as removing masking after an agent has been trained using masking. The source code can be found at https://github.com/vwxyzjn/invalid-action-masking