Learning-based Velocity Estimation for Autonomous Racing

Our lab's recent work with AMZ Racing on leveraging deep learning to accurately estimate the velocity of racecars has been featured in an article on IEEE Spectrum.

If you're interested to learn how we managed to obtain estimates that rival those coming from expensive, highly specialized sensors, using affordable sensors combined with machine learning - and the relevance of our approach to self-driving cars - please read the Spectrum article external page here.

For more technical details, the original paper is now available on external page IEEE Explore and external page ArXiv. In case you're curious to find out how an entire autonomous racing system fits together, you might also like our recent paper in the Journal of Field Robotics external page here, or its preprint on external page ArXiv.