01 / What it is
The project
This concept explores a robotic arm that uses one motor and electromagnetic clutches to control multiple joints. Interchangeable arm sections would change the system geometry, so its software must understand the current configuration and coordinate motion accordingly.
02 / My part
What I contributed
- Researching how control software can sequence one motor and multiple clutches to produce useful multi-joint movement.
- Exploring simulation and training approaches that can generalize across different arm lengths, joint counts, and configurations.
- Connecting mechanical constraints to the state, actions, rewards, and safety limits an ML control system would need.
03 / Experience
What I have gained
- Defining an ML problem from physical requirements rather than beginning with a model.
- Thinking about training data, simulation, hardware integration, and testing as one control-system pipeline.
- Balancing lower-cost mechanical choices against the added complexity they create in software.
- Planning experiments for a project that is still in the research and prototyping phase.
04 / Now
Where it is going
Current focus
Defining the control problem and a simulation strategy before committing to a training approach.
Next step
Build a small simulated joint-and-clutch model and use it to compare rule-based and learned control strategies.