Train Control Handovers Using Machine Learning Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current train control systems lack the ability to effectively manage handovers between centralized and edge-based models, which can lead to communication breakdowns and synchronization issues, particularly when trains enter geo-fences with limited connectivity.
Innovation Solution
A system and method utilizing machine learning to implement handovers between centralized and distributed train control models, including a centralized cloud-based processing system, edge-based processing systems on-board locomotives, and a data acquisition hub to acquire real-time and historical data for training. This system uses virtual system modeling engines to generate control commands and a machine learning engine to predict communication breakdowns and adjust parameters for optimal control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If centralized control systems are used for train operation, then control precision and coordination are improved, but communication reliability deteriorates in geo-fence areas with limited connectivity
Solution Approach 1:
The control system is segmented into centralized control functions (off-line) and edge-based control functions (on-board). The on-board edge computing device executes control algorithms locally when communication is unavailable, while the centralized system handles coordination when connected. This segmentation allows the system to maintain control precision through centralized planning while ensuring communication reliability through local autonomy in geo-fence areas.
Solution Approach 2:
The system performs preliminary action by pre-loading control parameters, route information, and operational data into the on-board edge computing device before the train enters geo-fence areas. The machine learning model is trained in advance to predict train behavior and generate control commands. This preliminary preparation ensures that when communication is lost, the system can continue operating with high reliability using pre-stored information and local decision-making capabilities.
2Reliability
If edge-based processing systems are deployed on-board locomotives, then communication reliability is improved, but device complexity increases
Solution Approach 1:
The on-board edge computing device is designed with multi-functionality to handle diverse tasks including real-time data processing, machine learning inference, control command generation, and local communication management. By consolidating these functions into a single universal platform, the system improves communication reliability through local processing while avoiding the complexity of multiple separate specialized systems. The device can adapt its functionality based on communication availability and operational requirements.
3Reliability
If machine learning models are used to predict communication breakdowns, then control reliability is improved, but computational requirements and energy consumption increase
Solution Approach 1:
The machine learning model operates with partial action by executing inference at reduced computational intensity when communication is stable, and only increasing computational effort when prediction indicates impending communication breakdown. The system processes only the most critical features and uses simplified models during normal operation, reserving full computational resources for prediction and response generation when needed. This approach maintains control reliability through accurate prediction while minimizing energy consumption during stable communication periods.
Data Source
AI summary
A train control system uses machine learning for implementing handovers between centralized and distributed train control models. A machine learning engine receives training data from a data acquisition hub, receives a centralized train control model from a centralized virtual system modeling engine, and receives an edge-based train control model from an edge-based virtual system modeling engine. The machine learning engine trains a learning system using the training data to enable the machine learning engine to predict when a locomotive of the train will enter a geo-fence where communication between the edge-based computer processing system and the centralized computer processing system will be inhibited.