Integrating Reinforcement Learning with Explainable AI for Real-Time Decision Making in Dynamic Environments
Sandeep Kumar · 2025
Combining Reinforcement Learning (RL) and Explainable AI (XAI) is becoming more popular as a way to improve making decisions in real time in settings that are always changing. RL lets systems learn the best policies on their own because it can adapt to changes that are hard to predict. However, the fact that RL models aren't completely clear makes them hard to use in important situations where decision clarity is very important. This problem can be fixed with XAI methods that give us readable information about how decisions are made. This builds trust and responsibility. This paper looks at a system that combines RL and XAI to help people make better decisions in places where things change quickly and there is a lot of doubt. We use a deep Q-learning method to teach agents how to behave in changing situations like self-navigation and allocating resources. For later understanding, we use SHAP (SHapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations). Benchmark dynamic datasets and real-world case studies are used to test the system. The results show that adding XAI makes RL rules easier to understand without hurting performance. People who have a stake in the system can better understand and change how it works with the help of decision explanations. When compared to standard RL models, the results show that decisions are 15% more likely to be right and mistake rates are 20% lower. Combining RL and XAI has the ability to make strong, understandable systems that can make flexible decisions in real time in complex settings.