Smart Sleeve

Pulley bracket fit to potentiometer
Device hardware
Device around the strap
Original pulley design
Final pulley design
App data view section
App data view section

Inspiration

In Australia, many people live with chronic pain because rehabilitation is often delayed or too expensive to access consistently. Without regular feedback, recovery becomes guesswork.

We built Smart Sleeve to make rehabilitation more accessible by giving users real-time insight into how they move, not just how much they move.

What it does

Smart Sleeve is a wearable knee rehabilitation device that tracks both movement and movement quality in real time.

Instead of only counting repetitions, it evaluates how well each step is performed. The system detects issues like:

stalling mid-step
uneven acceleration
lack of control during descent
stiff or inconsistent motion

These signals are converted into simple scores so users can understand their progress and improve technique without constant supervision.

How we built it

Smart Sleeve is a tightly integrated hardware and software system.

Hardware

We designed a custom two-pulley sensing mechanism using:

a spring-loaded pulley for constant tension
a secondary pulley coupled to a potentiometer
a string attached to the user’s leg

This converts knee motion into continuous rotational data.

A major challenge was durability. Early prototypes failed under real movement forces, so we redesigned the system with:

soldered connections
heatshrink reinforcement
strain relief loops

This made the device stable enough for repeated use.

Embedded Processing

An ESP32 continuously reads the potentiometer and processes the signal into more than just angle data.

We built logic to analyse movement quality by tracking how motion changes over time. Instead of just detecting movement, the system evaluates:

smoothness
control
consistency

These are combined into step-quality metrics and progress scores that users can act on.

Custom Embedded Compiler

One of the most technical parts of the project was bringing a modern web interface onto a microcontroller.

Frameworks like Angular and Express are too heavy to run on an ESP32, so we built a custom compiler that:

takes Angular-style components and EJS-style server code
tokenizes and parses them in C
compiles them into static C/C++ assets at build time

This required translating object-oriented JavaScript concepts into C:

component state → runtime slots
bindings → direct lookups
data models → tagged C structures (objects, arrays, primitives)

The compiler also pre-renders server-side pages by evaluating route data and resolving templates ahead of time.

The result is a fully embedded web app served directly from the ESP32 without running a heavy framework at runtime.

Toolchain Integration

We integrated the compiler the PlatformIO toolchain using a pre-build script, so every firmware build automatically recompiles the web interface.

To make development manageable, we built and tested the compiler on a desktop environment with unit tests before deploying to hardware. There are two compilers in the github, windows_compiler and embedded_compiler

Challenges

Hardware durability: early designs failed under real movement forces
Embedded constraints: delivering a modern UI on limited hardware
System coupling: hardware reliability directly affected analytics accuracy

We solved this by iterating on mechanical design and shifting complexity from runtime into the compiler.

Accomplishments

Built a full end-to-end rehabilitation system
Designed custom mechanical sensing hardware
Implemented real-time movement quality analysis
Developed a custom compiler for embedded web apps
Delivered a modern interface from an ESP32

What we learned

Rehabilitation tech requires tight integration between hardware, signal processing, and user experience
Devices attached to a human body can experience significant forces in unexpected ways
Moving complexity into build-time systems enables richer embedded applications
Testing complex systems off-device is critical for fast iteration