Explainable Federated Deep Reinforcement Learning for Energy Efficient Resource Allocation in 6G Edge Enabled Smart Cities

Abstract

The exponential growth of intelligent Internet of Things (IoT) applications in 6G smart cities demands effective, fast, low latency, privacy preserving, and explainable resource management frameworks. An Explainable Federated Deep Reinforcement Learning (X-FDRL) framework for intelligent edge computing in 6G enabled smart city environments was proposed. The proposed model integrates Federated Learning (FL), Deep Reinforcement Learning (DRL) and Explainable Artificial Intelligence (XAI) to optimise task offloading, energy consumption and latency while preserving user data privacy. An explainability module based on SHAP is integrated to improve transparency and interpretability of DRL based decisions. Simulation results show that X-FDRL can reduce energy consumption by 31.4% and significantly reduce latency compared to conventional centralised DRL approaches. The framework allows for secure decentralised learning without sharing raw user data. Moreover, the ablation studies show the success of the combination of FL, DRL and XAI in improving the overall system performance in next generation smart city networks.