An AI Approach for Analyzing Driving Behaviour in Simulated Racing Using Telemetry Data

The emerging and rapid progress of esports (competitive computer gaming) currently lacks approaches for ensuring high-quality analytics to augment performance in professional and amateur esports teams. In this paper, we demonstrate the application of Artificial Intelligence (AI) and Machine Learning (ML) techniques in the esports domain, particularly in simulation (sim) racing, for analyzing drivers' behaviour based on telemetry data from race drivers. To achieve this, we used a professional racing simulator to collect a wide range of feature-rich telemetry data from 93 participants through MoTec telemetry software and the ACC sim racing gaming platform. An objective assessment of the characteristics of the driver's behaviour was then obtained through a set of predefined lap-based metrics derived from telemetry data. Additionally, a comparison of driving styles was carried out using machine learning approaches for grouping the acquired laps based on performance (lap time). The findings from our analysis contribute to a better understanding of how elite drivers differ from low skilled drivers based on their telemetry. Furthermore, our findings provide researchers with key metrics to develop more efficient training tools and techniques to improve sim racing performance.