Most car data loggers stream your drives to a server and do the thinking there. That costs you a phone app, an account and a data connection, and it puts a record of where and how you drive on someone else's computer. OpenCarMind takes the other route. It is an open-source OBD2 logger and dashboard for the ESP32 whose neural network runs on the chip itself: on-board AI, in the literal sense. It is part of Mr Pirate's tools and MIT licensed.

What the on-board AI does

The board reads your car through a Bluetooth LE OBD-II adapter. A pruned, knowledge-distilled multi-task 1D CNN looks at the last 5 seconds of driving and does two jobs at once:

  • It classifies the moment as stopped, braking, accelerating or cruising.
  • It estimates three values the OBD port does not report: brake pressure, flywheel torque and fuel rate.

You choose between two on-board models in the desktop app:

CarMind E.V CarMind R.E.V
Size on the chip about 89 KB about 66 KB
Parameters 38,679 19,815
Time per run on the ESP32-S3, in the car about 3.5 ms about 2.9 ms (median)
Learns in your car No, fixed and tested Yes, its fuel estimate and an airflow constant

CarMind R.E.V is the interesting one. The board updates part of the model from what your car reports while you drive, with no computer or cloud in the loop, and what it learns stays on your SD card. It suits older cars, roughly pre-2020, that have no fuel-rate PID.

Why on the board, and not in the cloud

Privacy you do not have to trust. Everything OpenCarMind records goes to the board's own storage, an SD card on the ESP32-S3 board or the 128 MB NAND on the ESP-Mosaico, and to the desktop app on your own computer if you pair them. Wi-Fi is optional and used for one thing: asking one HTTPS time server for the clock.

It works where your phone does not. No signal in a car park, no app update, no server outage. The model answers on the chip whether or not anything else is online.

Honest numbers. Fuel economy comes from the best source your car has. A car that reports nothing shows "--", never a made-up figure, and fill-ups you enter in the desktop app correct the board's fuel estimates.

On-board AI that stays on time

On Espressif's ESP-Mosaico touch-screen board, OpenCarMind EdgeOS 2.1.0 treats every model as a soft real-time job. A model is admitted only when every deadline on its core still holds, and when deadlines slip the board gives up the screen rate first, then learning, then the model, so the drive log keeps running. Version 2.1.0 adds AI services for the battery and the radios: a state-of-charge estimate from an extended Kalman filter, a small autoencoder that flags a battery behaving unusually, and a guard that watches for floods of new Bluetooth addresses and rotating Wi-Fi access points. A Health app explains each one in plain words. The EdgeOS article covers the admission check in detail.

What you need

  • A Waveshare ESP32-S3-LCD-1.47B board and a FAT-formatted SD card, or an Espressif ESP-Mosaico V1.2 (ESP32-S31).
  • A Bluetooth LE adapter that speaks ELM327 commands. The LELink is the one driven with the board so far.

The firmware and the desktop app for Windows, macOS and Linux are on the releases page, and the OpenCarMind documentation walks through flashing, pairing and the first drive. The code is on GitHub.