Abstract
This article proposes a high-precision map-based traffic flow optimization model for unmanned mining road networks. By integrating centimeter-level map data with graph theory and Long Short-Term Memory (LSTM) machine learning algorithms, the model aims to reduce congestion and enhance transportation efficiency. The effectiveness of the model was validated through four experiments. Results demonstrate significant improvements: average speeds increased by 50%, congestion relief times were nearly halved, and safety metrics—including accident rates and emergency braking events—showed dramatic reductions. These findings validate the model's application value in the process of mining automation and intelligence.
1. Introduction
Open-pit mining environments are characterized by constant change, where excavation activities and weather conditions frequently alter road networks. Traditional traffic management systems often struggle with this volatility, leading to bottlenecks at loading points and intersections. For unmanned vehicle fleets, the lack of granular, real-time environmental data results in low average speeds and high frequencies of emergency braking. This paper addresses these challenges by treating the mine road network as a dynamic graph optimized through high-precision (HP) maps and predictive algorithms.
2. Methodology
The proposed system utilizes a multi-layered approach to optimize traffic flow:
2.1 High-Precision Map Construction
The foundation of the model is a "machine-oriented" HP map providing centimeter-level accuracy. This map integrates data from ground surveys, aerial photography, and vehicle sensors to offer a multi-dimensional view of lane information and road components.
2.2 LSTM-Based Optimization and Management
A central processing system predicts traffic patterns using an LSTM network, which is well-suited for processing sequential sensor data. The system employs a flow optimization function ($Q_{opt}$) to dynamically assign weights to road segments based on real-time congestion levels. Using graph theory, the model calculates optimal traffic flow allocation ($f_{ij(t)}$) between nodes, allowing for proactive route adjustments rather than reactive stop-and-go logic.
3. Experimental Results
The model was evaluated through four key experiments focusing on efficiency, congestion, dynamic response, and safety.
3.1 Transportation Efficiency
The implementation of the optimization model led to a substantial increase in fleet performance. Average vehicle speeds rose from 20 km/h to 30 km/h, while the average travel time was reduced from 30 minutes to 20 minutes.
3.2 Congestion and Dynamic Response
Under high-flow conditions (150 vehicles per hour), the model reduced congestion relief time from 45 minutes to 25 minutes. Furthermore, the system demonstrated superior agility during unexpected events. For road closures (Event 1), response time dropped from 45 to 30 minutes; for speed restrictions (Event 2), it fell from 40 to 25 minutes.
3.3 Safety Performance
Safety is a critical byproduct of smoother traffic flow. By minimizing sudden stops and congestion, the model significantly improved safety metrics across the board, reducing the accident rate from 0.8 to 0.3.
4. Conclusion
The high-precision map-based optimization model proves highly effective in enhancing the "digital arteries" of unmanned mines. While the system shows great promise, future work will focus on improving robustness in extreme weather conditions and navigating high-complexity terrain. Overall, the results confirm the model's value in increasing the efficiency and safety of autonomous mining operations.
Summary updated