GRadient

• 

  Web app home screen

• 

  Configure car parameters

• 

  Select Track to simulate car telemetry values

• 

  Added telemetry reading simulation

Inspiration

GRadient is a simulation tool designed to gamify data science and encourage users to learn about Toyota GR's racing database. With the rising popularity of motorsport, fans around the world are increasing steadily and excited to learn more about the hard work designers put into optimizing performance. GRadient simplifies these concepts and gives users a quick jump into experimenting with parameters and information.

What it does

GRadient utilizes Toyota GR's telemetry datasets for every given track to correlate different cars with different parameters (Brake, Engine, Chassis, Steering). Then using GRU Recurrent Neural Networks, GRadient associates these parameters with telemetry readings at every timestamp in a race. Through GRadient's user interface, users are able to customize parameters and a racing track and observe their car's performance.

How we built it

This project started data first, with ML libraries like Scikit and Pytorch spearheading the exploration of data and what it can do. Then GRadient's web app was developed to showcase these tools afterwards, using Typescript and React to safely simulate features.

Challenges we ran into

Due to hardware limitations, the quality of the model that could be trained in time for the hackathon was unideal. Currently due to hosting limitations the displayed input params are incorrect and the simulation time is longer than preferred.

Accomplishments that we're proud of

Working with real data through Toyota GR's telemetry datasets has been an extremely interesting challenge. Learning to find real world patterns to create informed insights is something we are very proud of.

What's next for GRadient

In the future we hope to develop a more refined model and add features such as comparison in performance against other users

Built With

• jupyter
• matplot
• node.js
• numpy
• python
• pytorch
• react
• scikit-learn
• shadcn
• three.js
• typescript