I have thought about this and I see 2 ways to possibly improve the results:
- Get a better ultrasonic sensor. Apparently there are several sensors that are supposed to be more accurate than the HC-S04, like the US-100 and HC-SR05. The US-100 also works on 3.3V or 5V logic. I am going to order a US-100 and try it.
- Improve the child driver to do better range checking and/or software debouncing. In my case, I only get false presence when the car is not present because the strange readings are only ever less than the actual measurement. Right now, the driver only has a threshold for present/non present. If we said that the present and non-present measurements had to be in a specific narrow range, the driver would not generate presence changes from non-valid distance readings. Coupling that with logic which says that you need two in-range readings in a row before changing the presence status should eliminate the false status changes.
My review of the specs on these modules show that they are supposed to be accurate within 3mm. They are, however affected by temperature which impacts the speed of sound on which the measurements are based. I think the US-100 has a temperature sensor and compensates.
