Community builds of Mobile ALOHA

Bimanual mobile manipulation with 4 robot arms on a wheeled base for teleoperation

License

MIT

• DIY Mobile ALOHA at 5% of Original Cost (Jonathan Randall): A complete DIY recreation of Stanford's Mobile ALOHA robot at 5-6% of the original cost — built for $1,500-2,000 instead of $32,000 — with full ROS2 integration and ACT imitation learning capabilities. What Was Built Jonathan Randall, based in Sydney, Australia, set out to prove that the Mobile ALOHA concept could be built affordably by anyone with access to common maker tools. His build replaces Stanford's expensive components with accessible alternatives: - GoBilda frame chassis instead of custom aluminum extrusion - 2x Hiwonder xArm ESP32 arms instead of ViperX 300 arms ($2,400 each) - Raspberry Pi 5 as the main compute platform instead of a desktop workstation - Custom PCB designed for motor control and sensor integration - Stereo camera system for binocular depth perception - ESP32-based control with ROS2 micro-ROS bridge Technical Approach The robot uses a modular architecture split across multiple ESP32 microcontrollers and a Raspberry Pi 5: 1. Mobile base — 4 DC geared motors with encoder feedback, controlled via ESP32 2. Dual arms — Hiwonder xArm ESP32 6-DOF arms with custom firmware modifications 3. Perception — Stereo camera system for depth estimation and scene understanding 4. Learning pipeline — ACT (Action Chunking with Transformers) for imitation learning from teleoperation demonstrations The entire codebase is organized across 10 GitHub repositories covering hardware design, firmware, ROS2 packages, and training scripts. Why It Matters Stanford's Mobile ALOHA paper generated enormous excitement in the robotics community, but the $32,000+ price tag put the physical platform out of reach for most researchers and hobbyists. Jonathan's DIY version achieves a 94-95% cost reduction while maintaining the core capabilities: bimanual mobile manipulation with imitation learning. This build proves that meaningful robotics research platforms don't require institutional budgets. The detailed documentation, open source code, and use of readily available components make it one of the most accessible paths to experimenting with mobile bimanual manipulation. Build Details - 10 GitHub repositories covering all aspects of the build - Custom PCB designed specifically for this project - ROS2 integration for the full software stack - ACT training pipeline for learning manipulation tasks from demonstrations