INTELLIGENT RAILWAY CROSSING BASED ON NEURO-FUZZY LOGIC MODELS

ABSTRACT

Railway crossings are among the most vulnerable locations in transportation networks, where safety and efficiency must be managed simultaneously. Conventional railway gate control systems often rely on fixed-time logic or sensor-based mechanisms, which may not perform reliably under dynamic traffic conditions, adverse weather, or system faults. This research presents an intelligent railway crossing system that leverages a hybrid neuro-fuzzy logic model, integrating the adaptive learning capabilities of neural networks with the human-like reasoning of fuzzy logic.

АННОТАЦИЯ

Железнодорожные переезды являются одними из наиболее уязвимых участков транспортной инфраструктуры, где необходимо одновременно обеспечивать безопасность и эффективность. Традиционные системы управления железнодорожными шлагбаумами часто полагаются на таймеры или простые сенсорные механизмы, которые могут работать ненадёжно в условиях изменяющегося трафика, плохой погоды или системных сбоев. В данной работе представлена интеллектуальная система управления железнодорожным переездом, основанная на гибридной нейро-нечёткой логической модели, объединяющей обучающую способность нейронных сетей с логическим рассуждением нечётких систем.

Keywords: Neuro-Fuzzy Logic, Intelligent Railway Crossing ANFIS, Transportation Safety, Real-Time Control Systems, Fuzzy Inference Systems.

Ключевые слова: Нейро-нечёткая логика, Интеллектуальный железнодорожный переезд, ANFIS, Безопасность транспорта, Системы управления в реальном времени, Нечёткая логическая система.

Introduction

Railway crossings are critical points in transportation infrastructure where the safety of rail and road users intersects. Traditional railway crossings often rely on mechanical systems or simple sensor-based logic that can fail in dynamic or uncertain environments. As transportation systems grow more complex, the need for intelligent, adaptive, and highly reliable crossing systems becomes more pressing.

This research introduces an intelligent railway crossing system based on neuro-fuzzy logic models, combining the learning capabilities of neural networks with the reasoning strength of fuzzy logic. This hybrid system adapts to real-time traffic and train data, improving response accuracy and enhancing safety.

Literature Review

Numerous studies have explored automation in railway systems. Conventional systems use electromagnetic sensors or timing-based controls, but they often suffer from a lack of adaptability in unusual or unforeseen scenarios.

Fuzzy logic systems have been employed in railway automation due to their robustness in handling imprecise data (Zadeh, 1965). Studies such as Chen et al. (2003) showed improved decision-making in automated railway gates using fuzzy controllers.

Neural networks, on the other hand, have been used in pattern recognition and traffic prediction (Krose & van Dam, 1993). They can learn from historical data but lack explainability and logic interpretation.

Neuro-fuzzy systems, like Adaptive Neuro-Fuzzy Inference Systems (ANFIS), combine the strengths of both. Jang (1993) proposed ANFIS for function approximation, and its applications have expanded to traffic management and robotics.

Recent advancements have applied these models to intelligent transport systems (ITS), yet limited work specifically targets railway crossings, leaving a gap this research aims to fill.

The field of intelligent transportation systems (ITS) has seen substantial progress, particularly with the integration of artificial intelligence techniques for enhancing safety and efficiency. Within railway automation, the domain of intelligent railway crossings is critical due to its direct impact on accident prevention and traffic management.

Main Body

This section presents the proposed model, its components, and operational logic.

The intelligent railway crossing system consists of:

• Sensors and IoT devices to detect train position, speed, and vehicle presence near the crossing.
• A neuro-fuzzy inference engine that processes sensor input and decides on gate operation, alert mechanisms, and emergency overrides.
• Traffic data integration, considering vehicle density and weather conditions.

The system responds adaptively to complex scenarios such as:

• Trains approaching at varying speeds.
• Vehicles stalled on tracks.
• Emergency vehicles nearby.
• Adverse weather conditions impacting sensors or braking distance.

The neuro-fuzzy system ensures the gate logic is non-binary and context-aware, unlike traditional systems with fixed timers.

Methodology

The methodology followed in this research includes:

Data Collection

• Real-time data was collected from a simulated railway junction using virtual sensors.
• Scenarios included normal operations, emergency stops, weather disruptions, and traffic congestion.

Model Development

• A fuzzy inference system (FIS) was designed with input parameters: train speed, train distance, vehicle presence, and traffic density.
• Membership functions were defined for each input (e.g., Low, Medium, High).
• Rules were created (e.g., "IF train speed is High AND vehicle present is True THEN close gate immediately").
• A neural network was trained to fine-tune the fuzzy rule weights and membership functions, using backpropagation.
• ANFIS was implemented using MATLAB/Simulink to combine the systems.

System Integration

• The model was integrated into a crossing control unit with simulated sensor feedback loops.
• Performance was tested under various conditions.

Analysis

Performance Evaluation

The intelligent system was evaluated based on:

Accuracy of gate operation timing, false positive/negative rates in gate closure, response time to emergency events, system adaptability to new data.

                                                                                                Table 1

Performance Evaluation

Case Study Analysis

• In a simulated vehicle-stuck-on-track scenario, the neuro-fuzzy system correctly kept the gate open and activated alarms.
• Under foggy conditions, the system relied on historical learning patterns rather than just real-time sensor feed, improving safety.

Conclusions

This research demonstrated that integrating neuro-fuzzy logic into railway crossing systems significantly enhances their intelligence, adaptability, and safety. The proposed model reduces errors and improves response time under unpredictable real-world conditions.

References:

1. 1. Ali A., Prasad P. Automated railway gate control using fuzzy logic // International Journal of Computer Applications. – 2010. – Vol. 1(13). – Pp. 23–27.
   2. Chen H., Hwang J.N., Wang Y. A fuzzy logic based method for intelligent train gate control system // Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. – 2004. – Vol. 4. – Pp. 3321–3326.
   3. Gupta N., Sharma R., Saxena M. IoT-based automatic railway gate control system // International Journal of Engineering and Advanced Technology (IJEAT). – 2020. – Vol. 9(4). – Pp.  2134–2138.
   4. Jang J. S. R. ANFIS: Adaptive-network-based fuzzy inference system // IEEE Transactions on Systems, Man, and Cybernetics. – 1993. – Vol.  23(3). – Pp. 665–685.
   5. Krose B., van Dam E. Neural networks for control and system identification in railway traffic management // Neural Networks. – 1993. – Vol. 6(6). – Pp.  971–979.
   6. Kumar A., Singh V. Traffic prediction using hybrid neuro-fuzzy model // International Journal of Intelligent Transportation Systems Research. – 2019. – Vol. 17(2). – Pp.163–172.