Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

A Survey on Explainable Reinforcement Learning: Concepts, Algorithms, Challenges

Created by
  • Haebom

Author

Yunpeng Qing, Shunyu Liu, Jie Song, Huiqiong Wang, Mingli Song

Outline

This paper provides a comprehensive review of explainable reinforcement learning (XRL) to understand the inner workings of reinforcement learning (RL), especially deep reinforcement learning (DRL), and to enhance the reliability. To address the black-box problem of deep neural network-based DRL agents, we propose a novel taxonomy that categorizes existing XRL works into model, reward, state, and task description methods. In addition, we review and highlight RL methods that utilize human knowledge, which is often overlooked in the XRL field, to improve the learning efficiency and performance of agents. Finally, we discuss the challenges and opportunities in the XRL field and call for future research on more effective XRL solutions. The related open source code is available at https://github.com/Plankson/awesome-explainable-reinforcement-learning .

Takeaways, Limitations

•
Takeaways:
◦
We systematically classify and comprehensively review existing studies in the field of XRL to clearly present the current status of XRL research.
◦
We present the importance of RL methodologies that leverage human knowledge in the field of XRL.
◦
We suggest future research directions in the field of XRL and support follow-up research by providing related open source code.
•
Limitations:
◦
Although this paper provides a comprehensive review of XRL, it may lack an in-depth analysis of specific methodologies.
◦
Although a new classification system has been proposed, it is unlikely to fully encompass all existing studies.
◦
There may be a lack of concrete discussions about real-world applications of XRL.
👍
Made with Slashpage