Evolutionary F1 Race Strategy

What is it about?

Formula 1 is a highly competitive and ever-evolving sport, with teams constantly searching for ways to gain an edge over the competition. In order to meet this challenge, we propose a custom Genetic Algorithm that can simulate a race strategy given data from free practices and compute an optimal strategy for a specific circuit. The algorithm takes into account a variety of factors that can affect race performance, including weather conditions as well as tire choice, pit-stops, fuel weight, and tire wear. By simulating and computing multiple race strategies, the algorithm provides valuable insights and can help make informed strategic decisions, in order to optimize the performance on the track. The algorithm has been evaluated on both a video-game simulation and with real data on tire consumption provided by the tire manufacturer Pirelli. With the help of the race strategy engineers from Pirelli, we have been able to prove the real applicability of the proposed algorithm.

Featured Image

Photo by Jonathan Cosens Photography on Unsplash

Photo by Jonathan Cosens Photography on Unsplash

Why is it important?

F1 is one of the most difficult race competitions, and it offers a number of incredibly interesting challenges from a computational perspective. One of these problems is the optimization of the race strategy, that is precisely the focus of the present paper. As a matter of fact, modern F1 cars have hundreds of sensors that stream data from the car to the team garage. Team engineers need to understand those data and the correlation between them, in order to develop an effective strategy for the race. Therefore, computational intelligence approaches can be a valuable tool in this domain. However, F1 is an extremely closed world, where data are usually kept secret, and it is very difficult to find state-of-the-art works on this topic, with publicly available results and data. Still, it is possible to identify three main approaches that can be used for this type of problem: Monte Carlo (MC) sampling, Reinforcement Learning (RL), and Evolutionary Algorithms (EAs). In this paper, we attempt to go beyond the previous literature by presenting a custom Genetic Algorithm (GA) with much more complex features and degrees of freedom, which takes in consideration the telemetry data with an appropriate representation. As we will show in the paper, the proposed algorithm has competitive performances both on simulated and real data (the latter provided by engineers from Pirelli) and is very fast in terms of compute time.