This paper addresses the critical 'black box' problem in Deep Reinforcement Learning (DRL) by surveying comprehensive approaches to explain AI decision-making processes.

• Categorizes XRL methods across feature-level, state-level, dataset-level, and model-level explanation techniques
• Enables greater trust and transparency in AI systems that make sequential decisions
• Facilitates adoption of DRL in high-stakes applications where decisions must be accountable
• Particularly valuable for security applications where understanding AI reasoning is essential for detecting vulnerabilities and ensuring adversarial robustness

A Survey on Explainable Deep Reinforcement Learning