Our goal with Pocket Racer is to enable researchers to tackle difficult problems concerning high-speed autonomy through scalable methods. To achieve that, we've made our miniature platforms open-source as to encourage autonomous races to race together and have fun!
Keck Data Science Institute, Pepperdine University, Malibu, California, USA
Abstract— Head-to-head autonomous racing can engage undergraduate students in artificial intelligence (AI) and robotics education through the excitement of maneuvering sharp corners and overtaking opponents. In particular, the construction and programming of miniature mobile racing robots can facilitate head-to-head racing competitions for use in indoor classrooms and hallways. Unfortunately, existing racing platforms remain inaccessible for overtaking maneuvers in such confined settings because they are physically large or expensive due to the computational cost of enabling such high performance in autonomy. In an attempt to address these issues, we present Pocket Racer, an open-source, pocket-sized racing robot capable of head-to-head racing within a typical hallway or classroom. Inspired by how humans visually perceive the sequential progression of the road while driving, we propose an efficient yet effective approach to end-to-end autonomous driving. We demonstrate head-to-head autonomous racing with our Pocket Racer platforms, enabling high speed overtaking within hallways upwards of 15 km/h. Designed to be easily assembled with off the shelf components and a low-cost edge device (Raspberry Pi Zero 2 W), our Pocket Racer platform is made fully accessible through an open source website detailing build instructions. By making head-to-head autonomous racing accessible for undergraduate students, our work hopes to further artificial intelligence and robotics education.
We introduce Pocket Racer: A low-cost, open-source platform for multi-agent autonomous racing. With a budget under $250, it can autonomously navigate high-speed tasks such as overtaking and maneuvering within typical university hallways and classrooms. Powered by Vision Transformers (RaceViT) and designed for accessibility, Pocket Racer makes advanced autonomous racing technology available for research and education.
The videos below show real-time demonstrations of Pocket Racer in action. Single-agent scenarios feature one car autonomously navigating high-speed tasks, while multi-agent scenarios showcase multiple cars autonomously performing maneuvers such as overtaking. Each video highlights the capabilities of the RaceViT algorithm, processing images at 50Hz for precise steering and navigation. For perspective, the typical episode length ranges between 600 and 1000 frames. The success rates for various tasks in both single-agent and multi-agent settings demonstrate the effectiveness of the Pocket Racer platform in different environments.