When I arrive at certain places, like home, my trip often includes some stationary time after arrival. While part of the trip is correctly categorized as “Stationary,” I can’t add this time to the place visit.
If I use “Extract Brief Visits” or manually adjust the segment location in Arc, the issue is temporarily fixed. However, Arc recalculates and reverts to the incorrect data.
For example, a 3-minute car trip is incorrectly recorded as lasting 30 minutes:
This is a pretty common problem, and a weakness in the timeline processing logic unfortunately. The easiest way to work around it is by assigning that stationary segment to a different place.
The reason why the processing engine is rejecting your fix shortly after you do it is because it calculates the centre and radius of visits based on the recorded samples within them, then uses a kind of “gravity well” system of deciding whether to keep edge samples inside the visit or whether to eject them out into the adjacent trip item.
The samples in that stationary segment you’re trying to assign to the visit must be just slightly too far from the gravity well of the visit (even with them included in the centre/radius calculation), so they get ejected back out again.
This can happen when there’s things like a hotel reception, building lobby, building parking area, etc. Those environments will create a brief cluster of samples on the peripheral of the place, and then the processing engine will be inclined to think “these edge samples aren’t quite inside the visit, so I’ll push them out”.
So that’s where the workaround comes in: Create a new place, eg “Parking”, “Lobby”, “Reception”, “Front room”, etc. Then when you assign the samples to that, they’ll stay in a new visit to that newly created place. Not ideal, and not what you really want the timeline to look like, but better than having the samples incorrectly pushed back in the car trip item.
Eventually I want to fix this problem at the processing level. I’ve got various ideas for how to do that, but it’ll take time until I can get to doing that task. So for now this Parking / Lobby / Front room approach is the best option.
Oh, the situation can also happen sometimes even when there isn’t such a separate location in the place, like when there’s no carpark or lobby or front room. Sometimes the location data just isn’t accurate enough, and incorrectly clusters in the wrong place initially. Such is life with unreliable and low accuracy location data - it isn’t always as high quality as we’d like!
