WeatherRouting Bench 1.0: Towards Comparative Research in Weather Routing

Jiménez Javier a Precioso Daniel b Bu Louis c Redondo M. Victoria a Milson Robert c Ballester-Ripoll Rafael b Gómez-Ullate David b
a Universidad de Cádiz, b IE University, c Dalhousie University
Paper Code Dataset
Comparison of two routes

Route from Hamburg (DEHAM) to New York (USNYC) during week 6 of 2023, at a sailing speed of 6 knots. The figure compares the orthodromic route (magenta) and the optimized route using A*-FMS (green). It can be seen how the optimized route takes less time, being 200Km larger.

Abstract

This paper introduces WeatherRouting Bench 1.0 to fill a gap in weather routing research: the lack of a common standard for comparing algorithms. We provide a website with code in Python that will assist researchers in using weather data, testing routes, and employing a fair cost function for comparing different methods: http://benchmark.weathernavigation.com/. We also suggest optimization problems based on real shipping routes, each including a standard route as a baseline for comparison. Furthermore, our paper investigates how the choice of a problem affects the optimization by analyzing how the average cost of a journey varies across different sailing speeds, regions of the ocean and times of the year. By offering this benchmark, we aim to make it easier for researchers to compare their work in weather routing. As a first entry to the benchmark, we propose A⋆-FMS: an algorithm based on A⋆ graph search followed by a discrete Newton-Jacobi method, which combines the power of graph methods for global search and gradient descent methods for local optimization. The hyperparameters of this new method are discussed. With the best configuration and compared to the orthodromic route, the method reduces travel time by 1.4% when sailing at 12 knots, and 3.6% at 6 knots.

WeatherRouting Bench 1.0 addresses the lack of a common standard for comparing weather routing algorithms by proposing a benchmark with varied optimization problems.

This benchmark includes:
  • A GitHub repo with Python code for utilizing weather data, testing routes, and employing a fair cost function for method comparison.
  • A new optimization algorithm named A*-FMS, which combines A* graph search with the FMS variational method.

Dataset

currents gif
waves gif


The meteorological data used in WeatherRouting Bench 1.0 is sourced from the Copernicus Marine Service, incorporating high-resolution ocean currents and wave models. Specifically, the data is derived from the Nucleus for European Modelling of the Ocean (NEMO) model for ocean currents and the Météo-France Wave Model (MFWAM) for wave dynamics. The datasets are available on a 1/12° horizontal grid, ensuring detailed spatial resolution, which is crucial for accurately modeling and optimizing maritime routes, taking into account the dynamic effects of waves and currents on vessel speed and route efficiency.

Dataset

Optimization

waves gif


The optimization process in WeatherRouting Bench 1.0 uses a two-step approach to enhance route planning. Initially, the A* graph search algorithm generates a feasible path by discretizing the ocean into a grid and finding an optimal route considering constraints. To refine this, the Ferraro-Martín de Diego-Sato (FMS) variational algorithm smooths the path, minimizing travel time and ensuring a continuous, practical route. This combination leverages the global search capability of A* and the local optimization strength of FMS, resulting in efficient and realistic routing solutions.

Results

results image


The application of the A*-FMS optimization algorithm on WeatherRouting Bench 1.0 demonstrated significant improvements in maritime route planning. The algorithm effectively reduced travel time by leveraging detailed meteorological data, including ocean currents and wave models. On average, the A*-FMS algorithm achieved a travel time reduction of 1.4% at a cruising speed of 12 knots and 3.6% at 6 knots compared to the orthodromic route. The results highlighted the algorithm's robustness across different vessel speeds, ocean regions, and seasonal variations, showcasing its adaptability to various maritime conditions.

Paper

GitHub

The code for WeatherRouting Bench 1.0 is available on a private repository in GitHub. We will review your request and provide access to the repository as soon as possible. Thank you for your interest in our work.


BibTeX


          @article{
            Jimenez2024WeatherRouting,
            title={WeatherRouting Bench 1.0: Towards Comparative Research in Weather Routing},
            author={Javier Jim\'enez, Daniel Precioso, Louis Bu, M. Victoria Redondo-Neble, Robert Milson, Rafael Ballester-Ripoll and David G\'omez-Ullate},
            journal={Preprint submitted to Ocean Engineering},
            year={2024},
            keywords={Weather Routing, Benchmark, Optimization, Github},
            institution={Universidad de C\'adiz, IE University, Dalhousie University}
          }