Date of Award

6-8-2026

Document Type

Thesis

Publisher

Santa Clara : Santa Clara University, 2026

Departments

Computer Science and Engineering; Computer Engineering; Bioengineering; Mechanical Engineering

First Advisor

Salem Al Agtash

Second Advisor

Michael Abbott

Third Advisor

Unyoung Kim
Michael Schimpf

Abstract

Physical therapy is a branch of modern healthcare that holds crucial importance across multiple demographics, ranging from injured athletes to seniors. However, the recovery process often suffers due to a lack of quantitative progress tracking and the use of static equipment that does not adapt to the user. Furthermore, there are noticeable communication faults between the therapist and client, especially between sessions. Our dynamic rehabilitation cable machine addresses a lot of these limitations in a compact integrated system that provides adjustable motor-driven resistance, real time force measurements, and computer-vision-based exercise feedback. The main device is composed of two subsystems that were finalized through a tradeoff analysis, a motor-drum configuration and a tension-sensor that were constructed through machined and 3D-printed parts. The motor-drum system serves to smoothly transmit the torque from a CubeMars actuator into a drum that wounds the main cable in which the client uses for resistive exercises. The tension-sensor is a 3-pulley setup that is collinear with the motor-drum system and utilizes an S-beam load cell with HX711 signal amplification, serving to accurately measure force with future plans of formulating adaptive closedloop feedback. A STM32H753ZI microcontroller drives the actuator through a UART port and reads cable tension on each control cycle so the controller can verify the force being delivered during shoulder rehabilitation exercises. As part of the design process, most of our structural components with heavy force concentrations were validated through hand calculations and finite element analysis to confirm its safety and rigidity under expected loading conditions. Final testing demonstrated asymmetric resistance during shoulder abduction, inducing around 5.25 lbs concentric and 7.75 lbs eccentric force against a target prescription of 5 lbs and 7 lbs that were derived from published EMG ratios. In conjunction with our device, a computer-vision form tracker was built to compute joint angles for shoulder external rotation, internal rotation, and static holds to produce a per-frame score rate from 0 to 100 with corrective warnings. A dashboard displays the live skeleton overlay, form score, repetition counter, pacing guide, and multistep protocol progress to both the patient and the therapist. In the end, our final prototype exhibited an integrated rehabilitation system with quantitive force sensing, real-time motion tracking, and a portable form factor that could be applied in home-based rehabilitation settings. This device not only improves recovery efficiency through accurate force adjustments, but also solves the disconnect between in-office and at-home rehabilitation with its portability and general functionality. Our future work involves closing the firmware-level loop between form score and motor torque, full-data acquisition logging, and a clinician-facing remote communication channel.

Download

Share

COinS