ReachARM was developed during the WearHacks Montreal hackathon in October 2015. The project aimed to help people with disabilities regain mobility through advanced robotics, using muscle motion detection to control a robotic arm that simulates natural movement patterns. You can find more details about the project on our DevPost submission and check out our code on GitHub.
The Sprint Begins
One of our team members had developed a robotic arm as part of his Master's thesis. Having a pre-existing robotic arm gave us a head start, but integrating it with wearable technology presented its own set of challenges. Our first few hours were spent understanding the Myo armband's capabilities and limitations. While the robotic arm was already capable of precise movements, we needed to create an interface that could interpret human motion and translate it into mechanical control signals.
Technical Deep Dive
We broke down the project into three main components:
- Motion Capture: We used the Myo armband's EMG sensors to detect electrical signals from arm muscles, combined with Metawear's accelerometer data for precise position tracking. Getting these two devices to work in harmony was our first major challenge.
- Real-time Processing: We built a processing pipeline that could translate these biological signals into movement commands with minimal latency - every millisecond counted when trying to create natural-feeling movement.
- Mechanical Control: The servo system needed to be both precise and responsive, capable of smooth transitions between positions while maintaining stability.
Challenges
As with any hackathon project, we faced several significant challenges that forced us to adapt our plans on the fly. Our original design called for an independent camera system that would move separately from the robotic arm, adding an extra dimension of control and functionality. However, the lack of necessary hardware components meant we had to pare back this feature, focusing instead on perfecting the core arm control system.
The Myo armband presented its own set of challenges. Getting reliable data from the device proved more difficult than anticipated. We spent considerable time fine-tuning the calibration process to properly interpret the user's movements. One tricky aspect was matching the system's response to users' movement speed.
Key Takeaways
Our success in this hackathon largely came down to efficient team organization. We got our development environment up and running quickly, setting up our IDE, GitHub, and Slack workflows from the start. This early infrastructure allowed us to begin collecting device data almost immediately, which proved crucial for validating our approach.
Perhaps most importantly, we learned valuable lessons about working with new hardware under time constraints. The experience taught us how to quickly understand external APIs, analyze device capabilities, and identify the specific data points needed for our application.
Beyond the Hackathon
While ReachARM began as a hackathon project, it opened our eyes to the possibilities in assistive robotics. We identified several directions for future development:
- Creating more affordable motion capture alternatives to expensive medical-grade devices
- Developing specialized versions for different medical conditions and mobility needs
- Exploring applications in remote operation for hazardous environments
- Integrating machine learning to improve movement prediction and response
See It In Action
This demo video captures our first successful integration test, showing the real-time response and precision we achieved in just 48 hours: