Open to new roles
01 · Wearable rehabilitation robotics · modeling + evaluation
M.Sc. thesis · 2026
An M.Sc. engineering feasibility study spanning simplified mechanics, repeatable benchtop experiments, and an on-hand wearable prototype.
Role
M.Sc. thesis · 2026
What I built
Built reduced-order models for finger kinematics, tendon routing, passive torque, leverage, stroke, and tension.
Selected evidence
1,000 / 1,000 rigid-fixture cycles
The problem
The thesis turns an assistive-device concept into a working system of tendon routing, mechanics, control, and repeatable measurement. The final prototype moved from rigid benchtop tests to trials on the hand.
Technical ownership
System architecture
Reduced-order routing model → Arduino Uno motor/encoder control in C → camera and encoder measurement.
Evidence + outcome
Tools + systems
Case study
The model did not hand me a finished design. I used it to choose what to build next, measured the drive and load sides separately, and changed the routing when the hardware disagreed.
Before I modeled the mechanism, I spoke with clinicians and patients about rehabilitation needs and the practical limits of existing devices. Those conversations set the direction for a wearable design that could be tested on the hand.
Measurement loop
The Arduino commanded the motor while the encoder measured the drive side and the camera measured the finger side. That separation mattered. Motor motion alone could not show whether the tendon path actually transferred motion to the load.
Control architecture
The first rig could wait for the camera. The wearable could not, so timing moved into firmware and vision became a slower correction channel.
The webcam measured angle online between small encoder jogs. Each move ended with a pause and another measurement.
Useful for calibration, too slow for tappingA half-cosine trajectory ran on the Arduino while encoder and velocity feedback set motor PWM. The camera moved offline.
Deterministic motion on the rigid fixtureThe encoder handled the fast inner loop. A bounded camera loop near 50 Hz trimmed the reference as the wearable load path changed.
Fast motor control with load-side correctionRoute A to Route B
A reduced-order screen pointed toward moving the fixed guide. I used that result to revise the routing and chose a 29 mm shift near the high end of the screened range.
Rigid validation
The endurance sequence ran 20 consecutive sets of 50 cycles without changing the installation or configuration. Across the run means, measured excursion reached 97.5% of the 49.93 degree engineering reference, with about 0.26% mean period error.
Wearable evolution
The hand interface went through rings, straps, a failed resin transmission, and removable guides before the final build.
Printed rings tested simple ways to route the tendon around the finger.
Straps made the finger interface easier to fit and reposition.
The rigid resin path failed during development and was dropped.
A removable guide made placement and rework faster between tests.
The final build joined the wrist unit, tendon path, and finger interface.
Working prototype
The encoder handled the fast motor loop while the camera measured what reached the finger. This run shows the reinforced wearable cycling with the acquisition overlay recording commanded and measured angle.
Wearable transfer
On-hand measurements fell below the commanded excursion even when the motor-side system continued to move. Slack, compliance, calibration, and the changing load path absorbed part of the motion. The wearable load path, rather than motor travel, became the limiting part of the system.
