A wearable rehabilitation glove that turns servo motion into assisted finger and hand movement — built to help people with weak hand-muscle control practice everyday grips.
Move past controlling electronic components and use electronics to create real, physical movement.
The goal was a wearable glove that uses servo motors to assist finger and hand motion through a tendon-based mechanism — a practice aid for people with weak muscle-to-hand control, such as the reduced dexterity that comes with nerve or movement conditions. It's framed as assistance and daily hand-muscle practice, not a medical treatment: something to support everyday movement and give the hand structured exercise, with the possibility of gradual improvement.
Servos pull individual tendon lines attached to the fingers; elastic bands on the back of the hand return each finger to open when the servo releases tension. The user picks an exercise and the assistance level with physical controls.
Each exercise follows a move → hold → release cycle with a rep counter, so a session is structured practice rather than a single motion.
A strong MG995 wrist servo pulls all fingers together for close-hand and grasp (pick-up / drop) exercises — the large, full-hand movements.
Five SG90 servos, one per finger, drive the selective modes: hook grip (four fingers, no thumb), pinch (two fingers), and finger training that cycles through each finger one at a time.
An analog joystick selects the exercise and a potentiometer sets the assistance level, so a session can be tuned to the user on the spot.
The mechanism only pulls; elastic bands on the back of the fingers provide the opening motion when tension releases. A design constraint that shaped the exercise set.
An ST7735 TFT shows the active exercise and reps, so the user can follow the move → hold → release rhythm.
Each exercise counts repetitions through its hold-and-release cycle, turning assisted motion into measurable practice.
Chosen so each part of the mechanism could be tested on its own, and revised as the mechanical and power realities became clear.
Built up one finger at a time. Proving a single tendon before adding the rest meant every later problem had a known-good starting point to compare against.
A. Single finger
One servo, one tendon — establish whether a finger could be moved reliably at all.
B. Multi-finger
Expanded to several fingers and checked how the tendon mechanisms behaved together.
C. Control
Tested the joystick and potentiometer for selecting exercises and adjusting the movement.
D. Full hand
Integrated the servos and controls to run complete, whole-hand movements end to end.
The movement was rough and inconsistent, and untangling why took the project through three separate problems — mechanical, electrical, and platform. Each one reshaped the design.
Even with the servos moving correctly, the fingers didn't move smoothly — the motion wasn't reaching the hand.
Two mechanical limits: the glove material was too thick and stiff for the servos to pull the fingers cleanly, and the tendon strings didn't hold enough tension to translate the pull into finger movement.
Reworked the tendon setup and tension; identified a thinner, more flexible glove material as the real path to smooth transfer.
The prototype still moves the hand, but the mechanical system caps how smooth it can get — the biggest single constraint on the final result.
Movement was weak and inconsistent, and the servos buzzed — especially the MG995 doing whole-hand pulls.
The MG995 draws roughly 2A at stall. Every battery pack sagged under that load — a 4×AA / 5.5V pack dropped to about 5.25V and buzzed — so the servo never got the current it needed to pull the whole hand.
Moved to a regulated 5V / 3A wall supply feeding the servos directly, with a common ground to the Arduino — and never routed through the Arduino's own 5V pin.
A motor is only as strong as its supply under load. Total voltage looked fine on paper; it was the current draw at stall that broke it.
On the original ESP32 build, running the servos caused the display to tear and glitch, and the joystick misread its position.
Driving the servos from the ESP32's 5V pin disturbed the shared supply, corrupting the display and the joystick's analog reads.
Switched the controller to an Arduino Uno, where the servos run cleanly and the setup is simpler to drive.
Gave up the ESP32's extra capability for reliable motion — the right call for a device whose whole job is smooth, dependable movement.
The two servo classes also needed separate handling: the small SG90s run the selective moves (pinch, pick-up, drop, finger cycling) while the MG995 runs whole-hand training, so each got its own control logic rather than one shared routine.
The honest limits of a first working prototype.
The servos can only pull; elastic handles the return. That ruled out an active "open hand" exercise, which was replaced with finger-cycling training.
Glove thickness and string tension limit how cleanly motion transfers — improvable with thinner material, but not fully solved in this build.
Dropped the ESP32 for an Arduino Uno to get reliable servo behaviour, trading capability for dependable movement.
The MG995's current draw forced a regulated wall supply, so the current prototype is tethered rather than battery-portable.
