CarMind R.E.V and CarMind E.V#

OpenCarMind EdgeOS carries both on-board models. Both answer the same questions: what you are doing (stopped, braking, accelerating or cruising) and the three estimates (brake pressure, engine torque and fuel rate). One model runs at a time.

CarMind R.E.V, the default#

R.E.V learns in your car. It corrects its fuel figure from what your car reports while you drive. It also learns an airflow constant that gives your engine size. The rest of the model stays as shipped. The board saves what it learned.

CarMind E.V#

E.V is fixed. It runs as shipped and never learns. While E.V runs, learning is off and the Learning app says so. What R.E.V learned stays saved, and switching back to R.E.V turns learning on again.

Switching models#

Switch only while parked. All three ways do the same thing:

  • tap Use R.E.V or Use E.V in the model’s app

  • open Settings > MODEL and tap R.E.V or E.V

  • use the model switch in the desktop app

During a drive the board refuses and says “A drive is active. End the drive first, then switch models.” When the car’s computer answers but no drive is recorded yet, it says “Not now: the car link is up. Try again later.” While the board only searches for the adapter or connects to it, it stays Parked and the switch works. The board keeps your choice after a restart. Every model mark and name on the screen follows the model in use.

The board checks that a model fits#

A model has to give its answer within 1 second, while the board also reads the car and R.E.V learns. Before it loads a model, the board takes the model’s measured run time from its self-test, doubles it for safety, and checks that every job on that core still finishes on time. If one would not, the board keeps the model it has and says, for example, “Not switched: CarMind E.V refused: ml would miss its deadline.” Here “ml” is the name of the model’s job.

You should not see this with the two models that ship. Measured on the board, OpenCarMind EdgeOS 2.0.0 4aeff43: the check gave R.E.V a worst case of 4.2 ms against its 1,000 ms deadline.