A Survey on Transformers in Reinforcement Learning
This is an incremental survey that organizes existing research for researchers in reinforcement learning and AI.
The paper surveys the application of Transformers in reinforcement learning, reviewing motivations, progress, and providing a taxonomy to address the lack of systematic understanding in this evolving field.
Transformer has been considered the dominating neural architecture in NLP and CV, mostly under supervised settings. Recently, a similar surge of using Transformers has appeared in the domain of reinforcement learning (RL), but it is faced with unique design choices and challenges brought by the nature of RL. However, the evolution of Transformers in RL has not yet been well unraveled. In this paper, we seek to systematically review motivations and progress on using Transformers in RL, provide a taxonomy on existing works, discuss each sub-field, and summarize future prospects.