An Architecture: Solving The Lock Scheduling Problem At Artificial Waterways With Reinforcement Learning

Kurzfassung

This study proposes an architecture to address a critical challenge in inland waterway transport systems: the Lock Scheduling Problem (LSP). Locks and ship lifts are essential components of waterborne infrastructure, yet they frequently act as bottlenecks in the inland waterway network. These bottlenecks can lead to traffic congestion, increased waiting times, and economic losses for industries relying on waterborne transport. In practice, the First-Come-First-Served (FCFS) principle is often used for scheduling vessels. However, this approach often yields suboptimal and inefficient lock utilization. Alternative methods based on mathematical optimization, such as Mixed-Integer-Programming (MIP), offer improved scheduling capabilities but suffer from long computation times and limited adaptability to dynamic and uncertain operational conditions. To overcome these limitations, this study introduces an architecture that applies Reinforcement Learning (RL) to the LSP.