On 3/30/22 23:15, Michael Balzer wrote:
I've had no GPS false alerts since Craig added the NMEA checksum validation and I fixed the VW e-Up state detection.
That is, until war on Ukraine began. Since then, I've had multiple alerts, mostly late night. I wouldn't have thought military GPS jamming or spoofing would have an effect with that range, the next war region is ~ 1.200 km from here.
I've added some more logging to see if we can detect and filter this.
Reduced log from the latest event below. HDOP jumps from 1.8 (good) to 5.6 (bad), but only after the alert has been triggered.
Should we generally scale the alarm threshold by HDOP (if available)?
If we did scaling based on HDOP I think we could also lower alarm distance from 500m (250m was my first guess). Here's a quick spreadsheet: https://docs.google.com/spreadsheets/d/1C9mvMw3PSl58qmg-lRHdzMXzB2S4glE4vitd... It's unfortunate that HDOP lags the increase in distance, maybe you could adjust the smoothed calculation to damp that? Craig