Roadmap
The model proves the logic. The next steps are the bench run of firmware 2.0.0, then one real entrance.
Next on the bench
- Flash firmware 2.0.0 to both boards and the updated camera sketch (how).
- Repeat the bench tests, then run the whole system with
python -m smartparking run. - With the barrier up, pull the computer's USB cable: the barrier must come down within about 4 seconds.
- Record the run and add it to testing.
From model to pilot
Each hobby part has a standard equivalent. The decision rules, the record and the screens carry over; each new part needs a small adapter.
| Job | On the model | In a pilot |
|---|---|---|
| Barrier | MG90S hobby servo | A standard barrier arm, opened by the same signal through a relay |
| Plate and face | ESP32-CAM | A plate-reading camera at the lane |
| Card | RC522 desk reader | A weatherproof reader at the driver's window |
| Bay sensing | Two IR sensors per bay | Per-bay ultrasonic or camera sensors, a type already used in commercial bay guidance |
| Site computer | Laptop | A small fanless computer in the gate cabinet |
A 13-week pilot
| Weeks | Step | Done when |
|---|---|---|
| 1โ2 | Agree one entrance and its car park; register the drivers who agree | Drivers, cards and plates in the registry |
| 3โ6 | Install beside the existing gate; record what it would decide before it controls anything | Its decisions match the guard's for two weeks |
| 7โ12 | Run and measure: reading by day, night and in fog; time at the barrier; refusals and why; bay use by hour | Six weeks of logged data |
| 13 | Report, with a cost per bay from real supplier quotes | A go or no-go decision |
A host provides power, a network point and one contact. Card and plate need no biometric processing, so a pilot can start with those two and add the face check only where the law and the site allow it.
Later
- An exit lane: a second gate unit running the same software with a simpler rule.
- More bays per controller, and more controllers per site (the software already accepts several bay controllers).
- A plate detector trained on local plates, running on the site computer.