Being Stationary in the Car reducin Car speed

I sometime on a phone or something just sit and the car before going or after reaching somewhere, Arc is counting those stationary time in the trip time and because of that it reduces the speed per hour because it think I have driven 13km in 41 minutes. I tried to split the stationary session but it just doesn’t split after so many tries.

Hm. If you tap on that 17 minute stationary period, you can’t assign a Place to it? If you assign a place to it then it should be split out as a separate visit in the timeline. Unless perhaps it wasn’t truly stationary, and was instead perhaps a slow but moving period. Though even in those cases there’s usually still enough samples at some particular slow portion of the segment such that the processing engine leaves those as a separate visit.

Try tapping on the stationary segment and choosing “Not stationary”, so that you can see the path of the segment on the map. That way you can tell whether it truly was stationary or not.

There’s also the “Most common speed” detail row on the Trip Details view, which is designed specifically for getting a better understanding of trips where there’s a lot of different speeds involved (and thus the average speed can be something you didn’t expect).

If you tap on the “Most common speed” row it’ll take you to a graph/histogram that shows you a breakdown of the speeds of all the recorded samples within the trip. From there you can see which speeds were more common or less common. For car trips, that often means the slowest speed will be the most common (due to stop-start traffic), but you’ll also be able to see the histogram bars for the portions of the trip where you were moving along at a decent pace. So it’s a much more useful view than the simple average km/h number on its own, which as you can see is often misleading / not very useful for car trips.

Thanks for replying.

  • I tried to set as place but it’s still not splitting. And same as 41 min. It is full stationary I am just sitting in the car and car is on the drive way.
  • Tried tapping the stationary segment and set as not stationary(even though I am 100% I was in car from my camera) to see the route it has hardly 50 m of the drive So I split that and set as cad that worked but I still now have 16m of stationary attached to that drive.
  • Screen shot attached for most common speed. The 7.6/hr shows for sitting in car.
  • Just so know its the story of most of my car trips. Arc always take that time stationary in the car part of trip. I know it’s hard to define stationary in between but At least you should exclude the stationary time in start and end as none of the stationary segment I can assign as location.
  • Airplane is another issue where lot of segment recognize as Airplane and even though I tried to change each to car but common in a 13 km trip car+Airplane+Stationary+(Sometime bus also) 99% of my trips on Car or walking so why don’t just let us choose our common transport mode so Arc engine don’t need to waste time in learning and when in a once a while case we can be allowed to change the mode.

I use other app for location tracking and as attached you can see it exactly taken when I left and when I returned.



Ah! Now I see the problem. Because Arc is very picky about visit / place boundaries, it will be seeing those samples as being outside of your house. So essentially it’s seeing them as a separate visit (ie a visit for “sitting in the driveway”).

Ideally, if you assigned that stationary segment to your Home place, it would just merge it into your Home visit and be done with it. Which is what it does do, but then it rechecks the visit boundaries after you’ve done the edit, and it then moves the samples back outside the visit again, due to them being outside of the place/visit boundary.

This is something I’d like to improve at some stage. If you say “these samples are at X place”, then it should honour that, instead of processing them back out of the visit again. So that is on my todo list!

In the meantime, the easiest way to fix it is to create a new private place, such as “Home Driveway”, and assign the stationary segment to that. That will stop the processing engine from splitting the samples out again, because they’ll have a new “visit centre of gravity” located at your driveway. It also has the benefit of giving you much more detail in your timelines. Though granted, if you don’t want that extra detail it’s hard to convince Arc to do what you want.

Aside: that workaround is something I do daily (although Arc auto recognises it now) for the bike stand outside my condo, and the bike stand at my local mall. I also do similar sometimes for the pool/garden area at my condo. In that latter case, the processing engine doesn’t force me into it - it sees that area as within the place’s radius/centre of gravity - but I like to do it anyway, to get that extra detail for time spent at the pool.

This is odd. Do you live anywhere near an airport? If so, there could be local data for planes on airport runways. (Airplanes tend to spend more than half their time just taxiing around airport runways, so even though we think of them as being very fast and very high up, most of their time is actually at ground level and slow speed).

But even so, unless you live within a few kilometres of an airport, it seems quite odd that it’d continue to throw airplane into car trips. And even if you did live that close to an airport, it still seems … weird. I’m not sure what to make of that. I’ll keep thinking.

What I’d expect you to see is for your car trips to be sprinkled with short segments of another activity type that you personally commonly do, such as cycling or walking for example. To have it throw in an uncommon activity type repeatedly… yeah, I can’t immediately see why it’d do that. I’ll ponder and get back to you.

This is what is is doing already :wink: If you just do a high level activity type confirm, for the entire trip (ie the Edit button or Confirm view on a trip, instead of the Individual Segments view), then it’ll have Car at the top, and Airplane will likely have a score of zero. (Though if you’re seeing a non zero score for Airplane in that trip edit / confirm view, I’d be keen to know! That would be very surprising).

The sprinkle of not-car segments on the Individual Segments view is mostly just noise, that you’ll see in any kind of data if you push deep enough into it. With Arc, that deeper detail of the data is made visible, for the data perfectionists amongst us (eg me :joy:). While other apps would only ever show you the high level detail, ie just the overall best match type for the entire trip, Arc lets you dig down to the individual segments and even individual samples (from the Split Segment view).

So while it does look a bit messy, and I do always want to push that noise floor further down over time, with ML improvements, it’s a mostly unavoidable side effect of allowing us to dig far deeper into the data than most (or any?) other apps of this kind do.

Think of it like zooming in on a flat surface, using a microscope, and seeing that the surface isn’t really flat and is instead a bumpy mess of weird surface artefacts. The high level illusion of simplicity versus the low level / high detail reality of complexity. To get a truly smooth surface is a lot more difficult than to get a superficial appearance of smoothness, if you see what I mean.

On this I disagree. If you travel from point A to point B, covering 100 kilometres, and it takes you 1 hour, that’s 100 km/h. Whether you travelled at 150 km/h for part of it and 10 km/h for another part of it, that doesn’t change the fact that the overall speed of the trip was 100 km/h.

This is exactly why the “Common speed” view/histogram exists - to show us not just the “mean” (ie average) speed, but also the distribution of speeds. So the overall trip speed shown in headers is the mean (average) speed, and the “common speed” listed on the details view is the “mode” speed. Both are relevant, but have different meanings and uses.

I am not at all close to Airport. Atleast 20-30Km away. What is sent you was just one of the example but Arc recognize some segments as Airplane oddly quite often. Once it think in a Airplane while at home without any car. What wonders me is Flying in Airplane is total different parameter and how can it be trigger without even moving from one location to other at a certain speed. Normally people fly in Airplane not just roam around on runway. Feeling bad for the people who work in airport or baggage their activities must be messed up…Arc should be correcting itself if it see car segment before and after in seconds time that what it think is wrong.

I am love in-depth data too but the only issue is it should make some sense. Like driving in car it’s quite normal for the car to slow down. Stop at signals and also as important to learn user visiting a place and dropping off somebody, obviously it is quick stop may be 5-10 sec but those are visits but Arc keep ignoring those needed visit and recognizing all those traffic stops which is good to know but other is more important. Stopping on road and going inside a place and dropping somebody should be differentiated by the engine.

I agree with your point 100% but what I am saying is different why include waiting in the car in the trip. Specially start and end. As for the start you are at your origin and time could be merged in origin and end is desitination and that time could be merged with destination visit rather than trip time.

This turns out to be not true! Airplanes (and passengers in them) genuinely do spend the majority of their time either stationary on runways or moving slowly on runways. If you’re interested, I could screenshot some of the speed and altitude graphs from the global models. It’s an interesting insight into what vehicles really do spend their time doing, versus what we think they do!

Ok those Airplane scores of 3 and 1 look sensible enough to me. The 3 is a bit weird - perhaps there’s been people in that neighbourhood misclassifying trips as Airplane, feeding bad data into the models. But still, 3 is low enough to not cause trouble, just a bit of noise in the Individual Segments view.

And it already is :wink: Anything longer than 2 minutes will be split out as a separate visit. The situation with your 17 minutes on the driveway is something different. That’s a case where Arc does think it’s worthy of being a separate visit, but when you assign the same Place to it as is assigned to the adjacent visit, the processing engine looks at the data and says “nah, that’s outside of the visit, it’s a different place”, so it pushes the samples back out of the visit again.

Unfortunately Arc doesn’t explain that that’s what it’s done, and there’s also no easy way to work around it, other than the steps I described above, where you assign the segment to a different place, such as “Driveway”.

So there’s the speed and altitude histograms for airplanes globally. What it shows is that the overwhelming amount of samples (and thus time spent) are for ground level altitude and for almost stationary / very slow speed. (The speed histogram is in metres per second, but you get the idea - the leftmost bar is slow/stationary speed).

Edit: Those speeds can’t be metres per second, otherwise those numbers are way too high. I must be already changing them to km/h in that graph, even though they’re mps in the database.

Whatever plane Data you show for me it doesn’t make sense to see part of road trip as Airplane. Something need a serious fix.

Today I have another example for you which is more weirdo than previous ones. I was at home this weekend not even got out in backyard or front yard. Pls see how Arc sees that. How would you explain that???

One more example of car trip today morning where you can see how Many times Arc think I am a pilot and driving a airplane.

As long as there isn’t an incorrect assignment of Airplane at the top level (ie at the whole trip level - from the Edit / Confirm view, not the Individual Segments view) Arc is still doing a better, more accurate job of activity type detection than any other app out there that I’m aware of. So it’s still Best In Class.

If another app turns up that gets higher accuracy than Arc, then the priority of ML improvements will go up. But while Arc is still best in class, such improvements are only in the “nice to have” category, not the “needed now” category.

Wow you changed on statements so quickly. It’s hardly 24 hours when you said if I get airplane more than zero it’s unusual and that you want to know now I have don’t hard work to help you fix you are for no reason praising your own app. What I can say is god bless you I rest my case.