NIVA dock

Live dashboard · Firmware · Rev C.1 hardware · Scoring rules

Niva is a smart insole for gait screening. An ESP32 reads four force pads, a piezo disc under the heel and an MPU6050, then streams 20 frames a second to a React dashboard and a Flutter app. Each gait phase, score and alert comes from a threshold in one TypeScript file, so you can point at the exact reading that set off an alert. Niva was shown at Automation Expo in July 2026.

Status: the bench build runs today. The Rev C.1 wearable pod is designed, routed and DRC-clean in KiCad, but it hasn't been built or measured yet. Niva is a research and screening prototype. It isn't a medical device.

Bench build

The active sketch is niva arduino/niva_hardware/niva_hardware.ino.

What the firmware does each frame:

• reads every ADC channel 16 times at 12 bits and averages the result
• subtracts a zero-load tare (60 samples, saved in NVS, redone when you send calibrate or tare over WebSocket or Serial)
• reads pitch, roll and accZ from the MPU6050
• sends one frame every 50 ms to your WebSocket clients and the USB port
• shows Wi-Fi state, tare progress and live channel bars on the OLED

You set up Wi-Fi through a WiFiManager captive portal (NIVA-GaitGuard-AP), and the board answers at ws://niva.local:81 over mDNS.

JSON over WebSocket (ws://<esp32-ip>:81):

{"heel":42.00,"inner":68.00,"outer":31.00,"toe":27.00,"piezo":176.00,"pitch":-2.40,"roll":1.30,"accZ":9.72}

CSV over USB Serial at 115200 baud:

H,I,O,T,Piezo,Pitch,Roll,AccZ

The simpler sketch, esp32_gaitguard_wifi_manager, uses hardcoded Wi-Fi credentials and doesn't drive the IMU or the OLED. BLE isn't in any sketch, and the IMU in firmware is the MPU6050 (you won't find BMI270 code here).

How a step gets scored

Both apps run the same engine on every frame: biomechanicsEngine.ts in the web app and biomechanics_engine.dart in Flutter, and there's no trained model in it. Here are the rules, with the default values from the code, so you can check any number on the dashboard by hand:

Rev C.1 wearable pod

The next build swaps the dev board for a pod you wear on the lower shin: ESP32-C3, an MCP3208 ADC, an LSM6DSO32 IMU, a removable battery cartridge, one button and one light. A desk dock charges the cartridges. Every part comes out of a script (FreeCAD 1.1, KiCad 9, Blender 4.5), so you can rebuild the whole design from hardware/work/.

These are CAD renders, not photos. The signal chain inside the pod:

What's still open: nothing has been printed, built, powered or measured. RF, sealing, skin contact, cleaning and FSR characterisation are untested, and nobody outside the project has reviewed the design. A clean ERC and DRC is an automated check. It isn't a design review. The pod has a bare-PCB first-article Gerber set for bench bring-up; the rest are review-only. The full list is in VERIFICATION-PLAN.md and the hardware README.

Run it

Don't have the hardware? Open the live dashboard and switch to Simulation. It streams generated gait for six modes (Normal, Parkinson, Stroke, Neuropathy, Foot Drop, Ataxia) through the same engine. The simulated patterns are synthetic, so don't read them as patient data.

Install only what you need for the part you're running:

• Web: Node.js 22+ and npm
• Flutter: Flutter SDK (Dart 3.3+)
• Firmware: Arduino IDE (or another ESP32 Arduino toolchain) with ESP32 board support

Web dashboard (niva web)

cd "niva web"
npm install
# if peer dependency conflicts appear:
npm install --legacy-peer-deps
npm run dev

Other scripts:

npm run build
npm run preview
npm run lint

Optional environment file, niva web/.env.local:

VITE_ESP32_WS_URL=ws://<esp32-ip>:81
VITE_ESP32_WS_RELAY_URL=wss://<your-relay-domain>/ws
VITE_ESP32_WS_RELAY_TARGET_KEY=target
VITE_DATASET_UPLOAD_URL=https://your-api.example.com/upload
VITE_DATASET_UPLOAD_TOKEN=your_optional_bearer_token

Optional WebSocket relay, needed when an HTTPS dashboard has to reach a ws:// ESP32:

cd "niva web/relay"
npm install
npm start

Flutter app (niva flutter)

cd "niva flutter"
flutter pub get
flutter analyze
flutter test
flutter run

The ESP32 endpoint, relay URL and upload URL are runtime settings on the Device tab, not build-time env files.

Firmware (niva arduino)

Open a sketch in Arduino IDE, install the libraries listed in niva arduino/README.md (WiFiManager, WebSockets, U8g2), select an ESP32 board and upload:

• Active hardware sketch: niva arduino/niva_hardware/niva_hardware.ino
• Simpler hardcoded Wi-Fi sketch: niva arduino/esp32_gaitguard_wifi_manager/esp32_gaitguard_wifi_manager.ino

Open Serial Monitor at 115200 baud, note the printed IP, and point the Flutter Device tab or the web Settings page at ws://<ESP32_IP>:81.

Stack

• Web: React 19, TypeScript, Vite, Tailwind, Recharts, IndexedDB, Web Serial
• Flutter: Riverpod, Hive, web_socket_channel, fl_chart
• Relay: Node.js and ws
• Deploy: GitHub Pages through .github/workflows/deploy-pages.yml, which builds niva web

There's no application server or database. Your datasets stay in your browser (IndexedDB) or on your phone (Hive). CSV upload to an external research API is optional.

Repository layout

Documentation

Niva is an engineering and research screening tool. It is not a medical diagnosis device.