Data Collection and Mathematical Modeling of Wheelchair Dynamics

Brief

A summer internship with R2D2 / Thryv Mobility deriving the mathematical model behind a manual wheelchair’s push dynamics, from real push-input data rather than assumed parameters.

Problem

Manual wheelchair push behavior varies with the user’s physique and push style. Before any control (like cruise assistance) can be added, the system needs a state-space model calibrated against real strain-gauge readings, which first have to be cleaned of sensor bias, slope error, and noise.

Approach

• Derived the mathematical model and state-space representation of wheelchair dynamics.
• Aggregated push-input data from 40 subjects of varying physiques.
• Estimated the first-order equation of the strain gauges and calibrated it to remove bias and slope error.
• Filtered signal inputs with a moving-average filter and implemented a cruise-control algorithm as a proof of concept.

Results

A calibrated first-order strain-gauge model and a state-space representation of push dynamics grounded in the 40-subject dataset, with a working cruise-control proof of concept on the filtered signal.

Status / Next

Completed as a fixed-term internship.