AI-Based Race StrategyAssistant and Car data Monitor

Abstract: In the world of motorsport, it is not only the driver's skill that determines the outcome of a race. Race strategy and car setup are two main factors that determine if the driver is competitive or not. Thus, this project focuses on optimizing the strategy part. In some of the motorsport series racing strategy includes tyre compound choices and timings on when to do a pit stop in order to change tyres or repair the car. Optimizing these tasks allows to create an optimal race strategy. Such strategies are data-driven and rely on records of tire usage from practice sessions and modeled data using regressions. The constants produced from the model can then be used, as showed in the project, in quadratic optimization problem in order to create different base strategies. To automate the strategic choices that will be further used in a race can be used artificial neural networks (ANNs). In this project an ANN has been successfully trained on the generated data from the strategies in a simulation environment. Additionally, the whole process of creating a AI-based race strategy assistant for race engineers that starts from the creation of the regression model, followed up by obtaining an optimization formula and finally designing the ANN is introduced. The data collected for the race strategy assistant development have been generated using the driver-in-loop method. The outputs of the race strategy assistant in this project are decisions of tyre compound stint lengths/timings of pit stops and what combination of tyre compounds to use during a race in order to minimize the race time. The evaluation results of the race strategy assistant demonstrated the overall improvement in reducing the racing time by finding the optimal tyre race set.