Causal Reinforcement Learning: A Survey
Abstract: Reinforcement learning is an essential paradigm for solving sequential decision problems under uncertainty. Despite many remarkable achievements in recent decades, applying reinforcement learning methods in the real world remains challenging. One of the main obstacles is that reinforcement learning agents lack common knowledge of the world and must therefore learn from scratch through numerous interactions. They may also struggle to explain their decisions and generalize the learned knowledge. Causality, on the other hand, has a distinct advantage in that it can formalize knowledge and utilize structural invariance for efficient knowledge transfer. This has led to the emergence of causal reinforcement learning, a subfield of reinforcement learning that seeks to improve existing algorithms using structured and interpretable representations of the data generation process. In this survey, we comprehensively review the literature on causal reinforcement learning. We first introduce the basic concepts of causality and reinforcement learning, and then explain how causal modeling can address core challenges in non-causal reinforcement learning. We categorize and systematically review existing causal reinforcement learning approaches based on their target problems and methodologies. Finally, we outline open issues and future directions in this emerging field.
Submission Length: Long submission (more than 12 pages of main content)
Assigned Action Editor: ~Bo_Dai1
Submission Number: 788
Loading