GOAT RACER ONE – Master’s Project 2025

A living document. Last Updated December 4, 2025 (FINAL)

Executive Summary

“Everything that moves will be autonomous someday, whether partially or fully. Breakthroughs in AI have made all kinds of robots possible.”

— Jensen Huang, NVIDIA CEO [@nvidiagtc2020]

With advances in sensor technology and GPU-accelerated computing, a revolution in autonomous mobile robotics (AMR) is taking place. The GOAT RACER ONE project builds upon the shared heritage of F1TENTH [@f1tenth2020] and NVIDIA’s Project Leatherback [@nvidialeatherback], transitioning from traditional control methods to AI-driven approaches that use reinforcement learning to create a more capable and adaptive autonomous racing platform capable of operating in diverse environments.

The GOAT RACER ONE uses a digital twin approach with reinforcement learning to train behaviors or policies. This addresses the messy real world and the curse of dimensionality through training in essentially a video game. The environment can be changed so that the model can learn a generalized view of the world. This will help the robot overcome lighting and sensor noise present in the real world. The system uses sim-to-real transfer so policies trained in simulation work effectively on real hardware.

The hardware platform uses a modified Traxxas RC car with an NVIDIA Jetson Nano computer and Intel RealSense camera. This setup balances performance with cost, making it accessible for education and research while still capable of complex autonomous behaviors.