Hi @matt, I’m a bit confused about my data yesterday which coincides with my first 24 hrs with Arc Timeline Recorder replacing Arc Mini.
I installed and made the change yesterday morning before walking to work (3km), spending the day in the workplace, then walking home (3km). Previously this has accurately recorded as walking both ways. During the day I sometimes have a straight line GPS anomaly that shoots out 150-450m from the office, puts me in a whole other part of the city and then shoots back to the office. I mark them as bogus locations each time.
Yesterday’s data in Arc shows me with only three locations and no movement data.
-
Home 9pm the night before (correct)
-
Random point on my walking commute to work about 7:30am (earlier than I was there)
-
Home again about 9pm (correct)
When I click on the random point in my commute to see more detail, the entire walking route to and from my workplace is shown, and including my workplace, all within a large orange circle spanning about 3km.
It also shows me in my workplace experiencing the GPS drift to a location and back from a location that isn’t there any more because I marked it as bogus. This is something I’ve done previously for ~ 5-6 days per week (Arc never seems to learn or the GPS weirdness is super strong) but yesterday it’s had a weird outcome when I marked it as a bogus location.
Maybe it’s related to Arc Recorder.
Maybe it’s coincidence.
Either way, it’s odd though that my travel is a single data point in a location but not at the right time, and contains 11 hours of walking data and other locations not appearing in the timeline. I’d appreciate your advice in resolving it, thanks.
Hmm. Yeah, weird!
Have you managed to be able to clean it up properly yet? If not, let me know where you’re up to and I’ll see if I can figure out what else you can do to clean it up.
As to why it happened / how to stop it happening again… I’m working on that part!
Unfortunately, because location data is so unpredictable, the same problem rarely happens exactly in the same way (except perhaps for those bogus GPS data jumps you’ve been seeing, which can be frustratingly consistent in happening every day in some locations). But what I’m working on at the moment is trying to pin down any communication breakdowns between Arc Timeline and Arc Recorder, which might be creating these potential new messes.
I still haven’t pinned down any recognisable / consistent patterns in the weirdness yet. And frustratingly, my own data isn’t showing any problems lately. (Good for me, but not good for trying to debug the problem!)
I’m going to ship an update to both apps soon, which improves the communication between them. There’s been a long standing problem where that communication sometimes breaks down, getting delayed / out of date, for unknown reasons. So I’m going to do it a different way, such that that breakdown becomes impossible. I’m not sure if that’s actually the cause of any of this weirdness yet, but it’ll be a step in the right direction.
Anyway, yeah, if your data is still a mess and it’s a struggle to clean it up, feel free to DM me screenshots or whatever. I might be able to spot a way around it, to clean it up.
1 Like
Bit of a tangent, but for this one… how long has that been going on? One common cause of that is a wifi hotspot being moved between buildings. Like, if some company moves offices, 400 metres down the road, and they take their wifi equipment with them, then for some days or weeks their wifi routers will be wrongly located in Apple’s wifi hotspot triangulation database. That then confuses all the nearby iPhones that are using wifi triangulation, until the database eventually updates to the new location.
I’ve seen that one a bunch of times when a nearby new construction project completes. A bunch of companies move in to the new building, and for some days or even weeks, the wifi hotspot triangulation nearby becomes a complete nightmare.
Another cause can be building reflections messing up GPS, in built up city areas. Though most of the time I see that kind of effect only being as bad as maybe 100 metres or so - basically reflecting across the road from a tall building. In those cases, the iPhone itself tends to eventually learn to get around some of it, ironically by making use of wifi triangulation. Though it never tends to fully self correct - built up city areas are just plain nightmarish for quality location data.
The final cause I sometimes see for those problems is intentional government GPS interference / offsetting. Like, a few times over the years I’ve stayed in hotels that happen to be near to a US embassy or some such. At this locations, the GPS data will periodically jump hundreds of metres away, presumably because the US govt are fudging the GPS data around their embassies.
Not as yet. I am still in the same situation of seeing Home > Partial commute location > Home. I can only edit once I click on the Partial commute location.
About 6-8 months.
One common cause of that is a wifi hotspot being moved between buildings. Like, if some company moves offices, 400 metres down the road, and they take their wifi equipment with them, then for some days or weeks their wifi routers will be wrongly located in Apple’s wifi hotspot triangulation database.
That’s a possibility, I suppose, though fairly low probability. There could be another building which my company occupies and which has reused a wifi hotspot but it’s an additional building not a move, so they would likely just buy new equipment. The ownership of the building is not new and the occupation of it by my company has been in place for the past 10+ years, so it seems out of character for the organisation to move the hotspot.
Maybe it’s just attracted to a tall building
Heh. Yeah, unfortunately if it’s been going on 6-8 months, it could be something unavoidable and unfixable like that. Some combination of nearby buildings creating GPS reflections in such a way that the device is going to get confused, and there’s no way around it.
If it were moved wifi hotspots, you’d expect it to self correct within some days or weeks. (Well, you’d really hope so!)
Aside: I’ve made a subtle change in LocoKit2 in Arc Recorder, to hopefully avoid or reduce some of these situations.
LocoKit1 uses a technique to reduce energy consumption of dynamically adjusting the desiredAccuracy setting periodically, taking into account currently achieved accuracy. Like, if it’s requesting 1 metre accuracy and all the incoming data is worse than 10 metres accuracy, it’ll adjust to requesting 10 metre accuracy. If it’s requesting 10 metres, and all the incoming data is worse than 100 metres, it’ll adjust to requesting 100 metres. That way it’s only ever requesting one magnitude higher accuracy than what the device seems able to achieve, which should stop the phone from wastefull pushing the GPS receiver too hard.
But years back an Apple engineer told me (paraphrasing) “be careful about that, because each time you change the desiredAccuracy it resets some internal state, which might have unintended side effects”. So my hunch is that that dynamic accuracy changing trick is sometimes confusing the location system, making it incorrectly jump between GPS / wifi hotspot triangulation / cell tower triangulation, or to combine the data sources in a non-optimal way, which then results in it jumping way off target.
Purely a hunch - I’ve got no convincing evidence for it. But in LocoKit2 I haven’t put in that dynamic accuracy system, and instead leave it always at 10 metres (except in sleep mode and standby). Hopefully over the coming months it’ll become clear whether that’s helping or not. So far it doesn’t seem to be hurting battery life / energy use, so if nothing else it means a possibly unnecessary bit of complexity is removed.
Hmm. Ok. I think the first thing I’d try is going into the partial commute item and splitting the segment inside there, then extracting part of it out as a new item. Basically try to trick the processing engine into reordering the items then correctly merging them, by breaking up the existing dodgy item into smaller parts.