1.5d testing with another GPS Update - 9 June
- Diarmid Mackenzie
Testing with this build:
https://tinyurl.com/covid-safepath-5june
Release notes:
Removed location logs from the lib react-native-background-geolocation - JIRA ticket SAF-494.
Instead of getting 5 location data points, we are now collecting data points until we get one with a high-enough self-reported accuracy ( < 15 m) - JIRA ticket SAF-556.
Main focus so far has been on:
Battery usage - has it reduced?
Reliability - still good?
Accuracy - still good?
All compared to previous build as assessed here: 1 June - 3d testing with latest Android Build
Headline
Significantly improved battery usage. No issues (and possibly even an improvement) on reliability & accuracy. Fantastic!
Battery usage
5% in 27 hours. vs. 8% in 20 hours previously. So we are uing about 50% as much battery as before. Great!
Reliability
As good as previously. Just 1 gap (37 mins). Phone was on low battery with battery saver on, which may have contributed (that’s new - previously I had kepy the phone at high battery, but I didn’t manage it on this occasion).
Metric | 23 to 26 May | 30 May to 1 June | 8 to 9 June |
|---|---|---|---|
% of expected data points | 61% | 91% | 95% |
% > 6 mins | 22% | 16% | 16% |
% > 9 mins | 17% | 2.7% | 2.1% |
% > 14 mins | 3.2% | 0.1% | 0.3% |
% > 29 mins | 2.1% | 0.1% | 0 |
Accuracy
Points still tightly clustered. Maybe even an improvement? Hard to say as these aren’t tightly controlled tests, and there’s some movement. Also there’s 3d data on the left vs 1.5d on the right.
But overall we see data just as tightly clustered with the new build - and possibly even tighter.
Before After
