UKC

How do smartphone walking apps measure mileage?

The one I'm using has a map on the screen so it cant just be footsteps?

Today I repeated my Sunday morning walk  to the nearest possible footstep and today's read 3.00 miles while Sunday's was 3.06 miles. I calculate the variation as being 100 yards over the three miles.

Is this a normal thing and am I expecting too much for the readings to match, more or less?

In reply to Tom Valentine:

I'm not sure but I know that I can average a good 15kph on Strava whilst remaining perfectly still ☺️

Edit. It's improved. Just checked and I'm averaging only 5-8kph.

In reply to Siward:

I've nothing to complain about then. Thanks.

In reply to Tom Valentine:

A variation of less than 2% doesn't seem worth worrying about, you could always use a piece of string instead. 

In reply to Tom Valentine:

Yeah on strava, if running on a phone, when you stop the signal bounces around so it'll add distance. That doesn't sound too far out. 

I generally find a watch more accurate than a phone, my iphone always overestimated mileage by 3-5%.

In reply to Tom Valentine:

Your phone records your location using GPS at regular intervals, and does a sophisticated version of joining the dots to calculate how far you travelled. The GPS locations are subject to errors of a few metres, so the resulting track length will also have some error (which will be different from one day to the next) and also tends to be a slight overestimate. 1-2% variation seems to be pretty typical. Further explanation and some interesting data collected from marathon runners here: https://www.theguardian.com/lifeandstyle/the-running-blog/2016/apr/15/cours...

In reply to Mark Bull:

Yep - no one else had actually answered Tom's question of "How do smartphone walking apps measure mileage?" until Mark neatly did above.

Tom - I take it you have a vague understanding of how GPS works?  My understanding doesn't go past vague, but your phone (or any other GPS device) picks up signals from I believe geo-stationary satellites (so they don't go round the earth, they are always above the same spot) and does some trigonometry to work out where you are as a result.

In my head my phone is taking a number of bearings and the drawing the lines back, and where they cross is where you are - like I learnt to do in scouts 35 years ago! But rather than a church tower and hill top it's "sighting" on satellites a long way above us.

edit: just looked on wikipedia, which led me to https://www.cfa.harvard.edu/space_geodesy/ATLAS/gps.html which explains they are not geo-stationary, which just makes it even more impressive and magic.

In reply to Mark Bull: Also the GPS will use some sort of best fit line to calculate a distance.  If you want to deliberately wobble that line around it will be less accurate as you're not going in a straight line.  How did your number of steps compare between the two measurements

In reply to TobyA:

I think I was a bit swayed by the fact that GPS can position my car as accurately as one car park space  and that led me to see it as infallible but I didn't think about cumulative  small inaccuracies.

I'm still puzzled about accuracy varying from one day to the next. I've read about tree  cover affecting the signal but I walked under the same trees on Sunday as I did yesterday.

In reply to Tom Valentine:

It's the random, fractal nature of GPS tracks, that Mark explained rather well.

Next, ask yourself why the elevation figures aren't the same as you get by counting contour lines on your map...

In reply to captain paranoia:

that's rather a sore point: sometime over the past few weeks they have removed things like elevation from the (free) app and made them available only on paid upgrade.  

In reply to Tom Valentine:

The number of satellites your GPS could 'see' and whereabouts in the sky they were won't have been the same though.

In reply to TobyA:

Interesting. I would argue that they do, in fact, go round the earth - it is just that the earth is going round too. If they were not going round they would fall out of the sky. And no, it is not like two trains going at the same velocity because velocity is entirely relative whereas acceleration is not.

Anyway, back on topic. I measure mileage by roughly counting the squares on the map. I used to use a piece of string, but these days I can't be arsed.

In reply to deepsoup:

So If i walk in a dead straight for a mile, then retrace my steps, I shouldn't expect it to give the same reading because of the different positions of the measuring satellites? 

Task for this afternoon

In reply to Tom Valentine:

My Garmin 235 measures my regular 6-point-something run (which I repeat to the metre every time I do it) as between 6.18 and 6.26 kilometres.

It also solemnly tells me I’ve ascended 13 metres and descended 20, without me starting my run from my rooftop.

Just GPS margins of error or a quirk in the algorhythms, I suppose.

In reply to wbo2:

Yes, absolutely: that's what I meant by "sophisticated version of joining the dots".  This is inevitably a compromise between not smoothing enough (and leaving some noise in) and smoothing too much (and removing some signal). 

In reply to deepsoup:

Atmospheric conditions are also a source of error - for example, the amount of humidity in the atmosphere affects the speed of light (and therefore radio signals) slightly, which results in an error calculating the distance to the satellites. This can be corrected for by using a fixed base station in the local area with a known position to broadcast the estimated error (differential GPS), but hand held devices like smartphones don't use this. 

In reply to captain paranoia:

Thank you!

Cumulative elevation gain is a harder problem than distance. The raw GPS elevations have larger vertical than horizontal errors (due to the geometry - most of the satellites visible at any one time are near the horizon). It is better to use the horizontal GPS positions and get the elevation data from a digital terrain map, but you have to be very careful about how you do interpolation to avoid accumulating errors (the errors from each "leg" of the track add up rather than cancel out). Some interesting stuff here: https://www.gpsvisualizer.com/tutorials/elevation_gain.html 

In reply to Tom Valentine:

You know how the saying goes that you can't cross the same river twice?  You also can't walk a mile twice under the same sky.

I was just talking about the trees though really - you said you walk under the same trees each time, but whatever effect they had on the signal wouldn't be the same each time even if you could ignore everything else.

In reply to deepsoup:

This is a discussion to file under Useful stuff learned on UKC. Grateful to all.

As an aside, one of my junior school mates had a dad who was a surveyor. One day he brought in an actual surveyor's chain to show us (its relevance greater to kids in those days who chanted their mutiplication tables and weights and measures tables every morning., the "22 yards equals 1 chain" part actually being brought to life). it was a cumbersome thing!#

Just one more thing!

How accurate in terms of pinponting rolling locations can a very sophisti cated navigation system be?

I'm thinking about automated driving: if a driverless car is moving at 50 mph along a curved line is a once per second update (as mentioned in one of the links) good enough to avoid a massive threepenny bit ride or even to stay in lane: or are the necessary updates made in tenths of a second or similar?

In reply to Robert Durran:

I actually understand this, I guess that's why its called geo-stationary orbit I guess, but after reading the Harvard thing saying they are not geo-stationary it all seemed a bit moot.

In reply to TobyA:

I hadn't realised that either. I suppose it is communications satellites that need to be geo-stationary - it's not much good if the satellite bringing you live TV from the US is somewhere above Japan.

In reply to Tom Valentine:

Self-driving cars don't just use GPS - as you have seen it's not accurate enough. They also use inertial navigation to augment the GPS data,  lidar/radar/cameras to image their environment and machine learning algorithms to decide what to do next.   

In reply to Mark Bull:

yes I've just watched an autonmous Audi TT climb Pike's Peak and it's obvious that GPS alone wouldn'tbe up to that. I've heard of inertia navigation so that's something else to google and I've already seen a couple of articles about radar and cameras  working some fuzzy neural magic into the system.

Still, Pike's Peak has hundreds of curves and no two are alike. Very im pressive but I think I'd have sat that one out.

In reply to Tom Valentine:

Correct. That and other errors in the system, due to atmospheric conditions, local ground environment, noise, etc.

If you reduce your sampling rate, you will reduce the straight line error (as each inter-sample section is longer, so the error from the straight line is much less significant). However, it will increase curved path error, as you 'cut more corners'...

In reply to Mark Bull:

I know; that's why I wryly suggested it...

...and referred to contour lines. When previously explaining the apparent elevation discrepancy, I suggested thinking about walking over sand dunes, each of which didn't quite cross a contour; you'd do an awful lot of up and down, but counting contours on a map would suggest you shouldn't...

In reply to colinakmc:

Do you live somewhere really flat, or is that just a fictional example? Sounds great if you are going for a pb.

In reply to Robert Durran:

Back in the days of Telstar you'd have people in the studio waiting for the footprint to come over the receiving station and see the pictures appearing through the noise and then eventually disappearing again, just like most ham satellites even now.

A miracle of the early 60s, though Arthur C Clarke had written a revolutionary article in Wireless World in 1945 laying out the principles of geostationary communications satellites occupying what became known as the "Clarke Orbit".  He worked on aviation electronics for the RAF during the war.

http://lakdiva.org/clarke/1945ww/

[factoid 16536 in a series of 100000]

In reply to captain paranoia:

Couldn't find a straight mile nearby so settled for a shorter stretch  (with a couple of trees covering part of it)

which produced  0.31 miles and 0.32.

I paced it and my stride was out by two steps. Perhaps I should stick with the Roman army  system.

In reply to mountain.martin:

Not particularly, I suspect the Garmin tells lies! My start point is 106 metres ASL and highest point about 113m; lowest point 100m. However it’s never actually flat, so I suspect my aggregate ascent is more like 40-50m. Ive never measured all the wee bits of vertical interval on the Connect readout graph (can’t be bothered) but heart rate vs speed certainly would hint at that.

I have worked out my steps for walking in poor visibility to get to features. I use 66 steps per 100 metres. although not an exact calculation I work on 15 minutes a Km and break that down if needed to smaller measures i.e 3 mins = 200 metres. I could use 20 mins a mile, but maps are in Km squares these days. My 15mins per Km is calculated on daily routes over all terrain and including rest stops.

My MemoryMap app also gives an accurate distance travelled which I would use as well as the above- mentioned in challenging conditions i.e to be sure to find a change of direction/important feature.

In reply to Diddy:

If your memorymap app is measuring distance travelled, it will be using GPS to do so (since the drift on its accelerometer sensors makes those unusable for the purpose). So you might as well just use the GPS to locate yourself...

And, as the OP observed, and has been explained, GPS path length suffers from inaccuracies due to position noise (although it's likely to be more accurate over pace counting in difficult terrain).