TerraSense
A prototype insole designed to turn foot pressure into sensory feedback for people with amputation or diabetic neuropathy. I co-founded TerraSense and wrote its embedded firmware.
- Role
- Co-founder and lead embedded systems engineer
- Where
- TerraSense, Georgia Tech Create-X
- When
- May 2025 to present
- Tools
- C++, ESP32-C6, ESP-IDF, DMA, MCPWM, BLE, logic analyzer
- Result
- 2.36 ms from sensing to feedback output, within a 5 ms target
The problem
We wanted to address lost ground feedback for people with amputation or diabetic neuropathy. Existing feedback systems start at about $900; our price target was under $100.
We began with an adaptive prosthetic ankle. User interviews pointed us toward ground sensation, so we changed the design to an insole: pressure sensors, an ESP32-C6, and a custom electrical-stimulation circuit in a 3D-printed housing.
TerraSense was accepted into Georgia Tech Create-X Startup Launch. We demonstrated the prototype at Demo Day in Summer 2025.
The firmware
Each pressure sample starts a 5 ms window to classify stance and produce feedback. I kept Bluetooth telemetry and output work separate from pressure sensing so wireless activity would not delay the sensing loop.
- A hardware timer triggers pressure sampling at 200 Hz. DMA transfers samples into alternating buffers without CPU copying. An int8 neural network classifies stance in about 2.2 ms, using bit shifts for rescaling.
- Bluetooth LE sends telemetry, while the ESP32's motor-control PWM peripheral drives haptic and electrical output.
How I tested it
Measured latency from pressure sensing to feedback output was 2.36 ms, within our 5 ms target. I checked it with microsecond hardware timers in the firmware and GPIO toggles captured on a logic analyzer. Both methods agreed.