Optimizing Periodic Operations for Efficient Inland Waterway Lock Management

TL;DR

This paper develops algorithms to optimize periodic water lock schedules, aiming to reduce vessel waiting times and improve inland waterway traffic management by matching irregular vessel arrivals with predictable schedules.

Contribution

It introduces polynomial-time algorithms for estimating periodic vessel arrival patterns and computing optimal lock operation schedules, enhancing efficiency in inland waterway management.

Findings

Intuitive policies often outperform optimal policies trained on periodic data.

Algorithms effectively match vessel arrival patterns with periodic schedules.

Scheduled policies reduce vessel waiting times in numerical experiments.

Abstract

In inland waterways, the efficient management of water lock operations impacts the level of congestion and the resulting uncertainty in inland waterway transportation. To achieve reliable and efficient traffic, schedules should be easy to understand and implement, reducing the likelihood of errors. The simplest schedules follow periodic patterns, reducing complexity and facilitating predictable management. Since vessels do not arrive in perfectly regular intervals, periodic schedules may lead to more wait time. The aim of this research is to estimate this cost by evaluating how effective these periodic schedules manage vessel traffic at water locks. The first objective is to estimate a periodic arrival pattern that closely matches a dataset of irregular vessel arrivals at a specific lock. We develop an algorithm that, given a fixed number of vessel streams, solves the problem in polynomial time. The solution then serves as input for the subsequent part, where we consider algorithms that compute operational schedules by formulating an optimisation problem with periodic arrival patterns as input, and the goal is to determine a periodic schedule that minimises the long-run average waiting time of vessels. We present a polynomial-time algorithm for the two-stream case and a pseudo-polynomial-time algorithm for the general case, along with incremental polynomial-time approximation schemes. In our numerical experiments, use AIS data to construct a periodic arrival pattern closely matching the observed data. Our experiments demonstrate that when evaluated against actual data, intuitive and straightforward policies often outperform optimal policies specifically trained on the periodic arrival pattern.