This unique project is a collaboration with STARX, an MSU engineering club that designs and fabricates its own mobility-enhancing exoskeleton suit. Their legacy suit uses active actuation (via linear actuators), which are controlled by an array of accelerometers.
This original movement system uses a conventional algorithm to translate accelerometer data into a linear motion for the actuator; in a reactionary method.
The objective for this project is to conduct a Proof of Concept (PoC) to test the feasibility of replacing the conventional algorithmic solution with a predictive, machine learning-based system.
The main reasoning for exploring a machine learning solution is the potential for efficiency uplifts. Breaking the motivation down into parts:
Latency: Utilizing ML would allow the system to predict rather than react; creating potential for a more responsive experience.
Comfortability: Overcorrection in stand-still situations can cause jitters due to the suit picking up small changes from its own movement.
Innovation: Switching to ML could tap into existing extensive research on human movement prediction and help foster new ideas.
