Mathematical Model for Mitigating Cascade Train Delays

A transformative railway management approach that shifts from isolated local decisions to a global, network-wide strategy, leveraging machine learning for dynamic optimization of secondary delays. This innovative approach was recognized with 2nd place in the Polish State Railways innovation competition.

Challenge

Train delays often escalate due to a cascade effect, where local scheduling decisions at rail junctions fail to account for the global network impact. Current systems prioritize local management over network-wide optimization, leading to inefficiencies in long-distance passenger transport.

Goal

Develop a machine-learning-driven approach to manage random disruptions and improve passenger information, shifting the paradigm from local decision-making to a global, network-wide management strategy.

Our Approach

  • Railway Network Model: Represented as a weighted matrix, where historical travel data determined expected travel times.
  • Dynamic Schedule Adjustments: Used gradient-based optimization to adapt train schedules in response to delays while maintaining operational constraints.
  • Real-Time Data Handling:
    • Implemented a tensor-based approach to store real-time arrival data.
    • Enabled the algorithm to redistribute secondary delays to minimize total disruption while ensuring collision-free scheduling.
  • Additional Constraints: Incorporated maximum allowable secondary delay for enhanced control.

Results

  • Successfully demonstrated a novel approach to handling cascading delays.
  • Optimized train movements to minimize total delay impact.
  • National recognition: Secured 2nd place in the Polish State Railways innovation competition (Check here!)

Future Plans

  • Integration into a Passenger Application:
    • Seamless connection with existing railway systems.
    • Travelers can dynamically change seat assignments during the journey.
    • Real-time rebooking options to adjust connections and minimize overall travel time.

Team Expertise

  • Data Scientist: Specializing in machine learning and optimization.
  • Railway Communication Expert: Provided critical domain knowledge on scheduling, constraints, and operational feasibility.