Explainable multi-agent task scheduling in vehicular edge computing using fuzzy logic and interpretable deep reinforcement learning

The swift expansion of latency-sensitive applications inside the Internet of Vehicles (IoV) necessitates that task scheduling exhibits both high efficiency and transparency, as well as interpretability. This research presents an explicable multi-agent framework that combines fuzzy logic with interpretable deep reinforcement learning (DRL) for intelligent job scheduling in vehicular edge computing contexts. Each agent (vehicle or edge node) assesses its local conditions—such as latency, workload, and energy status—using fuzzy logic, while the interpretable deep reinforcement learning model dynamically establishes optimal offloading and scheduling strategies. The proposed framework improves performance and explainability by utilizing explainable deep reinforcement learning architectures, including attention-based and decision-tree-embedded networks. Simulation outcomes under the evaluated simulation settings indicate that the model diminishes system latency and enhances task completion rates while preserving over 90% decision interpretability, providing a framework that demonstrates improved transparency and interpretability under the evaluated simulation settings.