Taught by Dr. Crowthers, this course centers around scientific research, engineering design, and clear technical communication. In this second part of the class, we work in small teams to engineer new products, typically assistive technology devices. We meet with clients, conduct patent searches, design and build prototypes, demonstrate our work to expert judges, and deliver the finished products to our clients.
The Assistive Technology project is the central deliverable of STEM II. Students partner with a client, identify a specific need, and apply the engineering design process to develop a working prototype. The sections below describe my project's problem statement, design approach, and prototype.
Around 80% of Americans experience low back pain at some point in their lives, contributing to an estimated $60 billion in annual economic loss in the United States (Simpson et al., 2019; Martin et al., 2014). Poor posture is a major contributor, as it increases mechanical stress on the thoracic and lumbar spine (Du et al., 2023). The problem is that postural deviation often goes unnoticed because it is tied to individual habit, and the wearables that are currently on the market rely on a single fixed-threshold alert with no per-user calibration. The clinical research devices that do provide accurate, personalized feedback are inaccessible to everyday consumers. There is a clear need for an affordable, comfortable, real-time posture monitor that delivers personalized haptic biofeedback based on each user's own baseline.
The engineering design process began with concept evaluation using a Pugh chart, where four sensor architectures and two attachment mechanisms were scored against weighted criteria and benchmarked against existing devices including the Upright GO 2, Straight Plus, and Z-Spine. From there, the device went through four CAD iterations in Onshape, progressing from a multi-encasement, screw-joined v1 to a single closed-encasement v4 with an integrated adhesive slot and snap lock. Each iteration was structurally validated using ANSYS Mechanical FEA across multiple candidate materials with a 20 lb load test and a 1 mm displacement test; Nylon was ultimately selected because it could withstand the snap feature's elastic strain with minimal deformation. The attachment mechanism was determined through a wear study comparing adhesive options, in which Nexcare Waterproof Tape received the highest composite comfort score. The full design was constrained by four Level One Requirements: at least 75% posture-detection accuracy with per-user calibration, haptic feedback of at least 150 Hz for 2 seconds when the curvature threshold is exceeded, zero assembly upon delivery other than the adhesive, and a baseline calibration completed within 15 seconds of power-on.
The final prototype is built around an ESP32 microcontroller running MicroPython, paired with an inertial measurement unit (IMU), a vibration motor, a LiPo battery, and USB-C charging, all housed in a single Nylon encasement attached with Nexcare Waterproof Tape. On power-on, the firmware runs a baseline calibration and then compares live gravity-axis readings against that per-user reference, triggering the haptic alert after a sustained deviation. Data is logged to a microSD card at one-second intervals for review. Testing followed a per-user 50-second protocol — 10 seconds of baseline calibration, 20 seconds of good posture, and 20 seconds of bad posture, all sampled at 10 Hz. Across four pilot users, the device produced no false positives during good posture and correctly flagged bad posture 80.2% of the time, clearing the 75% accuracy requirement and providing a low-cost alternative to existing consumer wearables and clinical research devices.