Solution Study
DECLASSIFIED OMI DOSSIER MC-04099
***
Drivers and cars are laden with dozens of sensors that provide data related to driver biometrics, automotive performance, and environmental conditions. But the biggest gap in information was related to the physical conditions of the track.
Specifically, McLaren was unaware of the coefficient of friction around turns on tracks around the world. Enhanced understanding of this missing element could lead to simulators that are more reflective of real world conditions thereby fulfilling their primary purpose more effectively.
An overly aggressive drive towards data normalization was affecting McLaren’s simulations. In turn, this was distorting various key points of information that could affect driving style or vehicle maintenance and improvement. Because racing is a pursuit in which victory can be achieved or loss over the space of a split-second, accuracy within the simulation was critically important.
Hypergiant approached this as a data-science driven endeavor and we trained a Support Vector Machine Classifier, a random Decision Forest, and a k-Nearest Neighbor (k=3). Using the minimum number of data features, environmental conditions, and leveraging a Machine Intelligence model, we provided valuable insight about the racetrack. Our algorithms provided 100% validation accuracy in predicting the sine of the derivative. This suggested that the data from the simulation was not capturing the true complexity of the real-world situation. We have a world-class data science team, but achieving 100% success on the first attempt suggests that something about the data is amiss. As illustrated in the screen designs, Hypergiant proposed a dashboard that leverages AI-driven predictive capability across multiple vectors:
1) The track represents a potential future state: Here a Machine Intelligence provides real-time predictions of the track grip from passively collected information from the vehicles and historical knowledge. This would require the simulated data to contain a higher degree of resolution of the behavior of the coefficient of friction than our current data provides.
2) The dial represents how this can be visualized for a particular car where additional information about the vehicle can be leveraged to produce a richer tool set for the engineers. This would require a more in depth understanding of how the car parameters are changing over the track and how this interacts with track friction.
3) The time-series represents a long term view of how the track friction has changed for a particular car, providing more information about current track conditions to the engineers, and in a future state provide predictions of future track conditions in order to make more informed decisions involving car behavior.
“Technological change is never an isolated phenomenon. This revolution takes place inside a complex ecosystem which comprises business, governmental and societal dimensions. To make a country fit for the new type of innovation-driven competition, the whole ecosystem has to be considered.”
