CAUSAL EXPLAINABLE REINFORCEMENT LEARNING FOR RELIABLE AUTONOMOUS DECISION-MAKING IN DYNAMIC ENVIRONMENTS

Autonomous systems require three essential decision-making capabilities which must be both dependable and able to adjust their operation while providing users with system status information. Traditional Reinforcement Learning (RL) methods depend on learning that uses correlation data yet these methods cannot discover genuine cause-and-effect links which causes their performance to diminish during environmental shifts. The lack of interpretability in most RL models prevents their use in critical safety systems which include robotics and smart manufacturing environments. The research presents a Causal Explainable Reinforcement Learning (CERL) framework which combines causal inference with reinforcement learning to create more robust and transparent autonomous decision systems. The model uses structural causal modeling to establish cause-and-effect connections which it combines with risk-constrained policy optimization to achieve stable learning and a causal explanation system that produces decision justifications which people can comprehend. The framework unifies causal reasoning and explainability through its single RL system which delivers improved generalization and adaptability and trustworthiness for dynamic environments The proposed approach is validated through simulation-based experiments in autonomous and human-robot collaborative scenarios which show better reliability and interpretability than traditional reinforcement learning methods. The CERL framework demonstrates its strength through its ability to handle two common challenges which autonomous systems face during real-world operational environments. The system uses intervention-based reasoning together with counterfactual analysis to assess potential outcomes which enables risk mitigation before crucial actions take place. The system provides essential benefits for safety-critical fields because it helps operators make correct decisions which prevent dangerous situations from arising. The combination of causal reasoning with explainability helps build trustworthy AI systems which improve system reliability. The framework improves decision-making accuracy while giving humans monitoring abilities through its capacity to create explanations based on logical principles which match causal relationships. This research aims to bridge the gap between high-performance reinforcement learning and responsible AI deployment, supporting the advancement of next-generation intelligent autonomous systems.