INTELLIGENT TRAFFIC MANAGEMENT SYSTEM FOR A MINING ENTERPRISE BASED ON A MULTI-AGENT APPROACH

Abstract

Objective. To develop, formalize, and evaluate a decentralized multi-agent system for real‑time traffic management in autonomous open‑pit mining operations, aimed at minimizing truck idle time, preventing congestion and deadlocks, and increasing overall haulage efficiency. Methodology. The system is designed using a multi‑agent paradigm where haul trucks, excavators,dumping points, and key infrastructure elements are modeled as autonomous agents. Task allocation is performed via a market‑based Contract Net Protocol with bidding based on estimated time of arrival. Path planning employs an A* algorithm on a dynamic graph, while conflict resolution is achieved through a reservation mechanism managed by infrastructure agents. Reinforcement learning (Q‑learning) is integrated to allow truck agents to adapt their bidding strategies over time. The approach is validated through agent‑based simulations under normal, high‑intensity, and disruption scenarios, and compared against a centralized first‑come, first‑served dispatch system. Results. Under normal operation, the multi‑agent system reduced average truck idle time by 44%, cycle time by 17.5%, increased throughput by 21%, and lowered traffic conflict frequency by 66.5% compared to the baseline. In high‑intensity traffic, it prevented congestion and deadlocks, maintaining smooth flow. During unexpected disruptions (excavator breakdown, road blockage), the system autonomously re‑planned routes and reassigned tasks within minutes, effectively isolating the impact and ensuring operational continuity. Scientific novelty. The novelty lies in the holistic integration of decentralized coordination, market‑based task allocation, dynamic path reservation, and reinforcement learning within a unified multi‑agent framework for real‑time traffic management in open‑pit mining. This approach enables emergent self‑organization, robust adaptation to dynamic conditions, and scalability beyond the capabilities of traditional centralized fleet management systems. Practical significance. The proposed system offers a scalable and resilient solution for autonomous haulage fleets, directly reducing operational costs through lower idle times and higher throughput, improving safety by preventing traffic conflicts, and enhancing resilience to equipment failures or route blockages. It provides a clear pathway for transitioning existing mines to fully autonomous, efficient, and safe operations. Keywords: multi-agent system, traffic management, open-pit mining, autonomous haulage, fleet management, reinforcement learning, Contract Net Protocol, path planning, conflict resolution, simulation.