MuSHR: A Low-Cost, Open-Source Robotic Racecar for Education and Research

Siddhartha S. Srinivasa, Patrick Lancaster, Johan Michalove, Matt Schmittle, Colin Summers, Matthew Rockett, Joshua R. Smith, Sanjiban Choudhury, Christoforos I. Mavrogiannis, Fereshteh Sadeghi

Abstract

We present MuSHR, the Multi-agent System for non-Holonomic Racing. MuSHR is a low-cost, open-source robotic racecar platform for education and research, developed by the Personal Robotics Lab in the Paul G. Allen School of Computer Science & Engineering at the University of Washington. MuSHR aspires to contribute towards democratizing the field of robotics as a low-cost platform that can be built and deployed by following detailed, open documentation and do-it-yourself tutorials. A set of demos and lab assignments developed for the Mobile Robots course at the University of Washington provide guided, hands-on experience with the platform, and milestones for further development. MuSHR is a valuable asset for academic research labs, robotics instructors, and robotics enthusiasts.

Introduction

In the field of autonomous systems, the "barrier to entry" is frequently financial. High-performance platforms like the MIT RACECAR have long served as the gold standard for research and education, but their approximately $2,800 price tag often precludes deployment in underfunded classrooms or by individual hobbyists. Furthermore, many existing platforms lack a cohesive "full-stack" integration, leaving researchers to struggle with bridging the gap between hardware assembly and software deployment.

The MuSHR (Multi-agent System for non-Holonomic Racing) project addresses these challenges by providing a unified ecosystem. It combines affordable, off-the-shelf hardware with a complete ROS-based software stack and extensive "do-it-yourself" documentation. This approach allows users to transition from a box of parts to a fully navigating robot in under an hour, significantly lowering the threshold for high-level robotics research.

System Architecture

The MuSHR platform is built upon a 1/10 scale RC car chassis (specifically the Redcat Racing Blackout SC). The hardware selection is guided by a "smart" economy, prioritizing components that provide sufficient performance for research tasks at a fraction of the cost of industrial-grade alternatives.

The MuSHR hardware ecosystem includes budget-friendly sensors, compute, and power management components.

Hardware Components

  • Compute: An Nvidia Jetson Nano serves as the central processing unit. While less powerful than the Jetson TX2, it provides ample GPU capability for real-time machine learning and localization tasks at roughly one-third of the cost.
  • Sensing: The robot utilizes an Intel RealSense D435i for depth sensing and a YLIDAR X4 for 2D laser scanning.
  • Power and Drive: Power delivery is managed via a simple buck converter, and the drive system utilizes a standard BLDC motor and servo controlled through a VESC (Vedder Electronic Speed Controller).

Software Stack

The software architecture is built on the Robot Operating System (ROS). The core navigation component, mushr_rhc, implements a Model Predictive Controller (MPC) for trajectory generation and obstacle avoidance. For localization, the system employs a Particle Filter approach, enabling robust positioning within a known map.

Design Philosophy and Key Insights

The primary innovation of MuSHR lies in its system integration and the "Good Enough" philosophy. The authors demonstrated that budget sensors, such as the YLIDAR X4—which costs 16 times less than the Hokuyo LIDAR used in more expensive platforms—are entirely sufficient for typical localization and navigation tasks in indoor environments.

Beyond hardware, the project treats documentation as infrastructure. By providing comprehensive videos, tutorials, and lab assignments, the team ensures that the platform is truly open-source and accessible. This focus on usability allows researchers to spend less time on platform maintenance and more time on high-level algorithmic development.

Results and Impact

The MuSHR platform achieves a significant reduction in cost without sacrificing the core functionality required for robotics research. A basic configuration can be assembled for approximately $600**, while a fully-loaded version with both LIDAR and RGBD cameras costs around **$900. This represents a ~67% cost reduction compared to previous industry standards.

The affordability of the MuSHR platform enables the deployment of large-scale multi-agent fleets for research and education.

The platform's educational value has been proven through its adoption as the primary tool for three distinct robotics and AI courses at the University of Washington. Its affordability enables "multi-agent" research, allowing a single laboratory to operate a fleet of ten or more cars for the price of just three high-end alternatives.

Conclusion and Future Work

MuSHR successfully democratizes access to autonomous vehicle research. While currently optimized for indoor environments, future work includes integrating high-accuracy indoor motion-capture systems for ground-truth experimentation. Additionally, the team is developing a custom gripper attachment to extend the platform's capabilities into the realm of decentralized multi-robot manipulation. While the Jetson Nano provides a high performance-to-cost ratio, the team acknowledges that extremely intensive deep learning tasks may eventually require hardware upgrades as research complexity increases.

Summary updated

Abstract

We present MuSHR, the Multi-agent System for non-Holonomic Racing. MuSHR is a low-cost, open-source robotic racecar platform for education and research, developed by the Personal Robotics Lab in the Paul G. Allen School of Computer Science & Engineering at the University of Washington. MuSHR aspires to contribute towards democratizing the field of robotics as a low-cost platform that can be built and deployed by following detailed, open documentation and do-it-yourself tutorials. A set of demos and lab assignments developed for the Mobile Robots course at the University of Washington provide guided, hands-on experience with the platform, and milestones for further development. MuSHR is a valuable asset for academic research labs, robotics instructors, and robotics enthusiasts.
BibTeX
@misc{srinivasa2019mushr,
  author = {Siddhartha S. Srinivasa and Patrick Lancaster and Johan Michalove and Matt Schmittle and Colin Summers and Matthew Rockett and Joshua R. Smith and Sanjiban Choudhury and Christoforos I. Mavrogiannis and Fereshteh Sadeghi},
  title = {{MuSHR: A Low-Cost, Open-Source Robotic Racecar for Education and Research}},
  year = {2019},
  eprint = {1908.08031},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1908.08031},
}
Figure