Technical Deep Dive
The mjbots/quad A1 is a masterclass in modular, real-time robotics design. At its core, the robot uses four mjbots qdd100 servo actuators, each integrating a brushless DC motor, a magnetic encoder, and a custom field-oriented control (FOC) driver on a single PCB. This is a critical architectural choice: by embedding the motor controller directly into the joint, the system eliminates the need for external servo drives and complex wiring, reducing latency and simplifying assembly. The qdd100 communicates over a CAN bus (Controller Area Network), a robust, deterministic protocol well-suited for real-time control loops. The central controller is the mjbots pi3 hat, which mounts on a Raspberry Pi 3B+ or 4. The hat provides CAN bus transceivers, power regulation, and an IMU (Inertial Measurement Unit) for orientation estimation.
Control Architecture: The software stack runs on the Raspberry Pi, using a real-time Linux kernel (PREEMPT_RT) to achieve control loop frequencies of 1 kHz or higher. The primary control loop is implemented in C++ using the mjbots moteus library, which handles low-level joint position, velocity, and torque commands. Higher-level gait and trajectory planning is done in Python, using the `quad` repository's `quadruped` library. This library implements a simple but effective state machine for trotting, bounding, and standing, with inverse kinematics computed on the fly. The system supports both position-based and torque-based control, allowing researchers to implement advanced techniques like model predictive control (MPC) or reinforcement learning (RL) by replacing the default gait planner.
Hardware Openness: The project's standout feature is its complete hardware disclosure. The GitHub repository includes:
- KiCad PCB design files for the qdd100 servo and pi3 hat
- STEP files for the 3D-printable chassis and leg components
- Bill of materials (BOM) with Digi-Key and McMaster-Carr part numbers
- Firmware source code for the servo's STM32 microcontroller
This level of openness is rare in robotics. For comparison, Unitree provides only mechanical CAD files and a proprietary SDK, while Boston Dynamics offers no hardware documentation. The mjbots/quad allows users to modify the motor winding, change the gear ratio, or even redesign the leg kinematics without reverse engineering.
Performance Benchmarks: While the A1 is not a speed demon, its specifications are adequate for research and education.
| Metric | mjbots/quad A1 | Unitree Go1 Edu | Boston Dynamics Spot |
|---|---|---|---|
| Weight | ~2.5 kg | ~12 kg | ~32.5 kg |
| Payload | ~0.5 kg | ~3 kg | ~14 kg |
| Max Speed | ~1.5 m/s | ~4.7 m/s | ~1.6 m/s |
| Battery Life | ~20 min | ~30 min | ~90 min |
| Joint Torque | 3.5 Nm (peak) | 23.7 Nm (peak) | 50 Nm (peak) |
| Control Frequency | 1 kHz | 1 kHz | 1 kHz |
| Open Hardware | Full (CAD, PCB, firmware) | Partial (CAD only) | None |
| Cost (approx.) | $1,500 (DIY) | $3,600 | $75,000 |
Data Takeaway: The mjbots/quad A1 is dramatically cheaper than commercial alternatives, but with significantly lower torque and payload. Its value proposition is not performance but transparency and customizability—ideal for labs that need to iterate on hardware designs or teach robotics from the ground up.
Relevant GitHub Repositories:
- `mjbots/quad`: The main repository (95 daily stars) with all hardware and software for the A1.
- `mjbots/moteus`: The firmware and library for the qdd100 servo (over 1,000 stars), providing low-level control APIs.
- `mjbots/pi3-hat`: Hardware design files for the Raspberry Pi interface board.
Key Players & Case Studies
The mjbots ecosystem is the brainchild of Josh Pieper, an engineer with a background in robotics and embedded systems. Pieper's vision is to create a complete, open-source robotics platform that rivals commercial offerings in quality but not in price. His work is heavily influenced by the open-source hardware movement, drawing inspiration from projects like the OpenDog and Stanford's Doggo.
Case Study: Stanford Pupper / Stanford Doggo
The Stanford Doggo project, developed by the Stanford Student Robotics club, shares a similar philosophy: an open-source, 3D-printable quadruped. However, the Doggo uses hobby-grade servo motors (e.g., Dynamixel) and an Arduino-based controller, limiting its torque and precision. The mjbots/quad A1 improves upon this by using custom FOC servos with higher torque density and a real-time control stack. The Doggo's GitHub repository (around 1,500 stars) has been a stepping stone for many hobbyists, but the mjbots/quad offers a more professional foundation.
Case Study: Unitree Robotics
Unitree, a Chinese company, has disrupted the quadruped market with its affordable Go1 and A1 models. The Go1 Edu, at $3,600, is a polished, ready-to-run platform with a developer SDK. However, its hardware is closed-source. Researchers who need to modify the motor controller or add custom sensors must work within Unitree's API. The mjbots/quad offers an alternative for those who prioritize hardware access over convenience. Several university labs, including the MIT Biomimetic Robotics Lab, have used mjbots components for custom robot projects, citing the ease of integrating the qdd100 servos into novel leg designs.
Case Study: OpenDog Project
The OpenDog project, led by James Bruton, is a well-known open-source quadruped that uses linear actuators. While impressive, its design is more mechanically complex and less suited for high-speed locomotion. The mjbots/quad's rotary joint approach is simpler and more scalable, making it a better starting point for gait research.
| Platform | Open Hardware Level | Motor Type | Control Complexity | Community Size |
|---|---|---|---|---|
| mjbots/quad A1 | Full (CAD, PCB, firmware) | Custom FOC servo | High (real-time Linux) | Small but active (95 daily stars) |
| Stanford Doggo | Full (CAD, firmware) | Hobby servo (Dynamixel) | Low (Arduino) | Medium (~1,500 stars) |
| Unitree Go1 Edu | Partial (CAD only) | Proprietary | Medium (SDK) | Large (commercial) |
| OpenDog | Full (CAD, firmware) | Linear actuator | High (Arduino/ROS) | Medium (~1,000 stars) |
Data Takeaway: The mjbots/quad occupies a unique niche: it offers the highest level of hardware openness among viable research platforms, but requires the most technical skill to build and operate. Its community is smaller but more technically focused.
Industry Impact & Market Dynamics
The robotics industry is bifurcating. On one side, companies like Boston Dynamics and Agility Robotics target high-value industrial and logistics applications with expensive, closed systems. On the other, a growing ecosystem of open-source hardware projects is lowering the barrier to entry for education and research. The mjbots/quad is a key player in this second trend.
Market Size and Growth: The global quadruped robot market was valued at approximately $120 million in 2023, with projections to reach $1.2 billion by 2030 (CAGR of 39%). This growth is driven by applications in inspection, surveillance, and research. However, the high cost of commercial units (Spot at $75,000) limits adoption. Open-source platforms like mjbots/quad could capture a portion of the education and low-budget research segment, which is currently underserved.
Funding Landscape: The mjbots project is not a startup; it is a passion project funded through sales of kits and components on the mjbots website. Josh Pieper sells fully assembled qdd100 servos and pi3 hats, generating revenue to sustain development. This model mirrors that of Arduino and Raspberry Pi: sell hardware, give away the design files. The project has not taken venture capital, which allows it to maintain its open-source ethos but limits its ability to scale production or marketing.
Adoption Curve: The mjbots/quad is still early in its adoption curve. Its GitHub star count (95 daily) suggests a steady but small influx of new users. For comparison, the popular ROS 2 repository has over 2,000 daily stars. The barrier to entry is high: building the robot requires soldering surface-mount components, 3D printing, and configuring a real-time Linux kernel. This limits its audience to experienced makers and researchers. However, as the project matures and pre-built kits become more available, adoption could accelerate.
| Metric | mjbots/quad (2024) | Arduino Uno (2005 launch) | Raspberry Pi (2012 launch) |
|---|---|---|---|
| Initial Cost | $1,500 (full build) | $30 | $35 |
| Time to Build | 40-80 hours | 0 hours (pre-assembled) | 0 hours |
| Target Audience | Advanced hobbyists, researchers | Beginners, educators | Makers, educators |
| Ecosystem Maturity | Low (few third-party add-ons) | High (thousands of shields) | High (vast software library) |
Data Takeaway: The mjbots/quad is analogous to the early Arduino or Raspberry Pi—a disruptive platform that lowers the cost of entry but requires significant user effort. Its success will depend on whether a community emerges to create tutorials, add-on modules, and pre-built kits.
Risks, Limitations & Open Questions
1. Scalability of Hardware: The qdd100 servo is a marvel of integration, but its 3.5 Nm peak torque is insufficient for larger robots or heavy payloads. Scaling up would require a new motor design, which is non-trivial. The project currently lacks a high-torque servo option.
2. Real-Time Linux Complexity: Configuring a Raspberry Pi with a real-time kernel is a known pain point. Many users report stability issues with the PREEMPT_RT patch, and the control loop can drop frames under heavy load. This limits the platform's reliability for critical applications.
3. Lack of Advanced Locomotion Algorithms: The default gait library is basic—it can trot and stand, but it cannot handle stairs, rough terrain, or dynamic recovery from pushes. Researchers must implement their own algorithms, which is a high barrier for newcomers.
4. Community Fragmentation: The mjbots ecosystem is small. There are few third-party tutorials, no official forum, and limited ROS integration. Users often rely on the project's GitHub issues for support, which is not scalable.
5. Ethical Concerns: Open-source quadruped hardware raises the specter of weaponization. While the A1 is too small to carry a meaningful payload, the design could be scaled up. The robotics community must grapple with the dual-use nature of such projects.
AINews Verdict & Predictions
The mjbots/quad A1 is not a product for the masses; it is a toolkit for the dedicated. Its true value lies in its role as a teaching platform and a foundation for custom research. We predict three developments over the next 18 months:
1. A Pre-Built Kit Version: The biggest barrier to adoption is the DIY build process. We expect mjbots to release a semi-assembled kit (pre-soldered PCBs, pre-printed parts) that reduces build time to under 10 hours. This will expand the user base from dozens to hundreds.
2. ROS 2 Integration: The lack of ROS 2 support is a major gap. A community-driven effort to create a ROS 2 driver for the mjbots/quad will emerge, making it compatible with the broader robotics ecosystem (SLAM, navigation, manipulation). This will unlock use cases in research labs.
3. Competition from China: Chinese manufacturers will clone the mjbots design and sell fully assembled clones for under $1,000, similar to what happened with the 3D printer market. This will commoditize the hardware but also validate the open-source model.
Our editorial judgment: The mjbots/quad is a vital project for the health of the robotics ecosystem. It forces commercial players to justify their high prices and provides a fallback for researchers who need hardware access. However, without a significant community growth or a commercial partner, it risks remaining a niche curiosity. The next year will be decisive: if the project can cross the chasm from 'hobbyist' to 'research tool,' it could become the Arduino of legged robotics. If not, it will be a footnote in the history of open hardware. We are cautiously optimistic, but the clock is ticking.