A Protocol for Comparing Sailboat Routing Algorithms

Ship weather routing is crucial today to optimize trajectories, whether for cargo ships, racing sailboats or drones. The well-known Isochrone method provides the fastest route. Multi-Objective Optimization (MOO) approaches can handle extra goals specific to each context. For example, during sailing races, strong winds can induce high speeds but present risks of damage to the vessel and cause exhaustion to the crew due to the maneuvers. In our work, these factors are collectively referred to as stress and more generally, it is interesting to take into account objectives other than time. MOO algorithms produce sets of solutions corresponding to trade-offs between the different objectives, rather than a single solution like time-oriented methods such as the Isochrone algorithm. Furthermore, it is necessary to compare MOO algorithms across all the objectives they address. Currently, the various types of routing algorithms are often evaluated using simple protocols, such as comparisons with existing algorithms. They apply on a reduced number of test cases using a single weather forecast file at the departure time. However, in real conditions, sailors take advantage of the latest grib files to periodically recalculate their trajectory. This paper addresses this weakness in the evaluation and comparison of algorithms. We have developed a protocol based on simulation i.e. the use of series of grib files. Routing algorithms are evaluated using real race data while they are compared using a MOO-specific metric. This method is illustrated on the Retour à la base 2023 race for two MOO algorithms.