Competitive Mobility Systems Simulator

Inspiration

We were inspired by the Haas F1 team's 2018 rise, where Gene Haas dared to join F1 with a lower budget and still become a major player in the midfield against giants like Mercedes AMG F1 and Red Bull Racing. Also, the idea of doing a hackathon about F1, which has never been done before in India, is an exciting prospect, and the idea that we get to even have a chance to work with an actual F1 team, that too a reputed one like Haas as well as the Mphasis F1 foundation, deeply inspired us to tackle the hackathon.

What it does

We are focusing on holistic sensor integration, which means tracking the car and driver as one. The use of proximity sensors on the cars, embedded sensors on the track, and on the sidewalls, combined with existing tech and racing parameters used by FIA provides us with more accurate telemetry and allows our software to provide real-time driver recommendations and improvements through real time simulations of all possible race paths. This will aid race engineers in providing the drivers with the best strategy possible in real time through actionable information - not just raw data.

The software will use machine learning to detect anomalies in the data feed in real time (for example, unexpected vibrations, tire temperatures, etc. ). It will also integrate biometric sensor data from the driver (heart rate, EMG, etc. ) to predict driver stress and driving performance.

How we built it

First, we listed the parameters such as pit stop average time, safety car cases, lap time varying with respect to DRS, brake setup, battery charge, fuel weight, etc. Then we decided to take all these parameters into account and make some changes of our own as described above. We decided to use fast API for the backend along with React.js for the frontend. For the ML anomaly detection and simulation portions, we used Python.

Challenges we ran into

Varying driver behavior skews the results for the ML, which has to be accounted for. Making the software performant requires immense precision and optimization, managing CPU load, avoiding overfitting on training data, and performing poorly in testing.

Accomplishments that we're proud of

Creating a system which can replicate the systems used by real teams to some extent and enhance them with novel features.

What we learned

How tough the job of an F1 team is and how naturally data intensive and competitive the sport is. Crunching through the massive amounts of data in time spans in the order of microseconds is a huge feat of engineering.