Indoor Navigation On Your Smartphone, Using the Earth's Magnetic Field

MrSeb writes "Researchers from the University of Oulu in Finland has created an indoor navigation system (IPS) that uses the Earth's innate magnetic field to ascertain your position — just like a homing pigeon or spiny lobster. According to IndoorAtlas, the company spun off by the university to market and sell the tech, its system has an accuracy of between 0.1 and 2 meters. The Finnish IPS technology is ingenious in its simplicity: Basically, every square inch of Earth emits a magnetic field — and this field is then modulated by man-made concrete and steel structures. With a magnetometer (compass), which every modern smartphone has, you can first create a magnetic field map — and then use that map to navigate the shopping mall, underground garage, airport, etc. Compared to most other IPSes, which require thousands of WiFi or Bluetooth base stations to achieve comparable accuracy, IndoorAtlas' infrastructure-free approach sounds rather awesome."

Comments

good in theory, bad in practice

[X0563511]

How exactly is one expected to create these maps then? Every time something metallic or magnetic is moved in the vicinity the previous map is now invalid.

Re:good in theory, bad in practice

[DeeEff]

It is simple. We build our malls out of plastic and we'll navigate just fine.

Another plus is that you'll be able to use this to find your kid in the ball pit at McDonalds next time they disappear.

Re:good in theory, bad in practice

[Professr3]

Simultaneous localization and mapping (SLAM).

Re:good in theory, bad in practice

[Russ1642]

I think it's more like bad in theory, worse in practice.

Re:good in theory, bad in practice

[John Hasler]

Especially after pranksters start leaving magnets in random places...

Re:good in theory, bad in practice

[postbigbang]

And rife with problems, including transient ones. The earth's magnetic field, while kind of constant at each point, can vary wildly in just a few meters. Most of the earth has lots of iron in it, and it doesn't take much to shift it around to the point where Hall sensors go whack. Nice idea, but in practice it will be really difficult. I'll get a nice hand-wound field coil, put on a randomly generated VCO, and watch their devices implode. Yes, they can degauss your ancient monitor, and wreak havoc on Hall devices.

Re:good in theory, bad in practice

[fuzzyfuzzyfungus]

For magnets of reasonable size, even the most powerful ones that you can reasonably buy have a surprisingly small area of noticeable field strength. A rare-earth puck will be good for a bit over a Tesla at the surface, and two of them can make friends with one another hard enough to take your finger off; but would drop into the background surprisingly quickly as you moved further away...

Re:good in theory, bad in practice

[bhcompy]

What about magnets of unusual size?

Re:good in theory, bad in practice

[jamesh]

No, thats what I thought but watch the video, they do something much simpler (i.e. have the user directly input where they are going on a map, while taking magnetic measurements on the phone).

Move smartphones have an accelerometer, so it should be possible to build this map by dead reckoning, especially if you can accumulate the data from enough different devices and/or different journeys...

Re:good in theory, bad in practice

[Jarik C-Bol]

the variance in the magnetic field is the POINT here. the are mapping these subtle/dramatic variations, and using them as measurable waypoints for navigation. (ie, there is a lot of magnetic field outside the GAP in this mall, but a lot less in front of the dip'n'dots, and there are a series of spikes by the fountain across from the dillards etc. etc. etc.) yeah, sure you can wipe out the magnetometer with a field coil, or (maybe) make the map return bad results by throwing 50 hard drives in a trash can, but for the most part, this seems like a smart way to map indoor areas for electronic navigation.

Re:good in theory, bad in practice

[postbigbang]

We agree on the point. My contention: it changes, and therefore the data is a variable. Yes, some places have lovely static fields. You can walk from points a-z and little changes....

Until something changes, and it will. A simple GPS is likely to have more immunity from fluctuations as its point source is an enormous distance away, which is precisely what makes GPS tough to resolve at many thousands of miles. The magneto sensors in a cellphone are going to get Hall Effect distortions pretty regularly, and those distortions aren't going to make the maps bad, just not as usable as GPS. Indoor navigation is subject to penetration indoors, I grant you. Iffy. Magnetic fields variability, IMHO, will thwart this.

Re:good in theory, bad in practice

[fuzzyfuzzyfungus]

I suspect that once your phone has encountered an MRI, small navigational errors will not be high on your list of concerns... The videos of hardware that can put a 3 Tesla field across an entire patient ingesting ferromagnetic objects are... dramatic.

Re:good in theory, bad in practice

[Gadget_Guy]

I assume that nobody here has actually used this system, so surely we can only say that theoretically it is bad in practice. In practice, they may have accounted for these problems in their theory.

My guess is that their software would not assume that people are lurching tens of meters in a single moment just because they pass something magnetic. They would use the same smoothing algorithm that GPS mapping uses. Have you ever noticed when you first load up a map on your GPS position is often quite inaccurate initially before eventually pin-pointing your location a few seconds later. They smooth out any anomalous readings after this, which you can see when your position briefly pauses while you are moving at a constant speed. During those pauses, the system has received new location that differs significantly from the last reading. These are obviously ignored to give the illusion of accuracy.

This magnetic system could do the same. With bidirectional communication, the software could report back anomalies due to changed environments and incorporate them into the self-correcting maps. Given that shopping centers do want to track their shoppers, it seems quite likely that there would be bidirectional communication happening.

Re:good in theory, bad in practice

[viperidaenz]

Until someone walks past you carrying something magnetic like.... their own smartphone

Re:good in theory, bad in practice

[viperidaenz]

start position = outside where you had GPS signals. Landmarks can be used to correct for drifting. That's how decent navigation software works, Accelerometer corrected with GPS locations when available (which is usually once per second when a signal is available). How else can it work in a tunnel or a bad urban canyon?

Re:good in theory, bad in practice

[scdeimos]

"Decent navigation software" just uses your last calculated velocity from your last successful GPS fix to move you along the calculated route. You can demonstrate this yourself by using underground tunnels and intentionally taking the wrong turn or even just slowing down - the GPS doesn't have any fancy accelerometers to notice you've changed course or speed, it continues to show you moving along the calculated route at your old speed, at least until you come back out into the open air to get a new GPS fix.

Re:good in theory, bad in practice

[viperidaenz]

That would be the difference between "decent" and "average". After a quick google search it turns out most of the implementations use a gyroscope and not an accelerometer. Here's one

Re:good in theory, bad in practice

[__aaltlg1547]

These people never heard of signs?

Re:good in theory, bad in practice

[X0563511]

Except if you build a building, it doesn't cause all the features upstream/downstream to reflow.

awesome

[twohands]

yay....one more way for them to track me **puts on tinfoil hat**

Re:awesome

[leuk_he]

Don't worry,

Some Chinese people will sell you coated iron-foil as aluminum foil, because steel is a few cent cheaper than alumunium. (they already do so for Utp just test with a magnet if you fail to get gigabit speeds on your network)

Am I the only one...

[medraut]

that immediately had Fog of War come to mind?

Join the army they said, see the world they said, I'd rather be sailing!

Just maybe :-)

Re:Am I the only one...

[medraut]

Well now that my brain caught has up with my knee-jerk comment, fog of war may not be quite accurate. More akin to exploring your map. But yes, Warcraft 1 definitely came to mind!

Pretty nifty in any event.

Smartphone compass sensitive enough

[MetalliQaZ]

I thought that the API for a compass in a smartphone really just supplied a bearing, but I'll assume they are assuming access to the raw data that comes from the hardware.

Still, are these magnetometers anywhere close to sensitive enough? I would think they are built to be as cheap as possible. I would expect them to be only accurate enough to determine which side of the road you're facing. This application must require amazing accuracy. I'd be amazed if they can get it out of a smartphone.

Re:Smartphone compass sensitive enough

[X0563511]

My phone (HTC EVO 3D) has a compass, accelerometer, gyro, and flux sensors. It's sensitive to changes as small as a single microtesla that I can tell, though noise usually means your sample resolution is about 5, instead. Some filtering would probably do nicely, since I'm only able to look at the raw reading.

Re:Smartphone compass sensitive enough

[vlm]

Bored? Grab your phone and run the app "GPS status" or probably a million similar apps, maybe even some free ones. Then move stuff around on your desk to see how the field changes. I can vary it about 20% by waving my steel clipboard around the phone. Now its possible with enough filtering you can assume changes are solely due to movement rather than me trying to sabotage the data gathering, and perhaps the map is actually of the 1st (or 2nd?) derivative of the field around my desk rather than just mapping the raw data so it doesn't matter if I'm IronMan and you're not, or if our phones do not have absolute calibration.

If I had more time on my hands I'd throw a fridge magnet on the floor, and try to "find the titanic" using the magnetometer and some string and graph paper and walking a grid pattern, or maybe pulling my phone along the floor on the grid pattern. Very much like the movie, I'll probably get bored halfway thru this titanic experiment. But it would probably work. Someone out there in /. land oughta try this, maybe try a big chunk of ferrous metal too, like a manhole cover (try not to get run over...)

Re:Smartphone compass sensitive enough

[fa2k]

Still, are these magnetometers anywhere close to sensitive enough?

it was bad on the nexus s on android 2.3 (?, the one before 4.0). In the compass app i could turn my phone around 180 degrees, and the needle almost followed along , now pointing almost in the opposite direction. I just tried it again now, on 4.0, and the compass works pretty well, it seems to have a precision of a few degrees when rotating the phone. I imagine there's some smart signal processing going on. Btw, I thought the exact same as the parent(and I don't have a good answer, just my experience)

Re:Smartphone compass sensitive enough

[bemymonkey]

The absolute funniest is pedestrian navigation using the built in compass - every time a car drives by (massive metal object), the screen rotates a little. We're going to need a better system...

Re:Compass

[X0563511]

It's called reading. It's been in use for like 1000 years.

(read the summary at least, please? this is doing far more than telling you your orientation.)

Re:Compass

[Baloroth]

Basic compass navigation has two data points: North and South. This has millions. It uses a compass, true, but it uses a sophisticated vector map of magnetic fields which normal compass navigation does not.

Something looks... not quite right about the video

[mark-t]

I can't pinpoint it exactly, but the way the phone is being held doesn't seem quite right. Like it was faked or done on a computer. Is this really just a concept video?

Re:Compass

[vlm]

this is doing far more than telling you your orientation

And no, this isn't a dupe about that fake gaydar app either.

Re:Compass

[fuzzyfuzzyfungus]

Yes and no. A compass just gives you your heading, relative to magnetic north. This will presumably be able to do that; but(depending on how much variation is available in the local magnetic field) it may also be able to give you an approximate location by detecting local magnetic anomalies.

The ground-level strength and orientation of the earth's magnetic field varies a bit naturally, for geological reasons, and our dense masses of ferrous structures and AC wiring probably provide a considerable amount of texture on top of that, in densely built areas.

Re:Compass

[ildon]

Some people have no sense of humor.

BS meter pegged

[frovingslosh]

There is just no way that one would get enough information from a magnetometer to give you the information to do this, any more than a compass in the great outdoors can tell you where you are, it can only tell you headings. Of course, there are all of the other issues that people bring up also, like metal or electrical things moving in the area and changing (effectively randomizing) the minimal information that you have. But to focus on that only ignores the greater problem, any simple vector from a magnetometer (even if it included a vector strength) can't tell you a location in 2D or 3D space. And unless you somehow magically know the correct way to orient your magnetometer when you are holding it, then just moving it as you move through the structure could give you any magnetometer direction at any point.

Well, human bodies disturb magnetic fields...

[Assmasher]

...so malls are going to problematic, no? Airports? Supermarkets? Retail stores that aren't going under...?

Lol.

Perhaps they have methods for mitigating these things or they are less problematic than I expect, but just changing the shelving and orientation in a store would screw this thing up, doesn't it seem?

Re:Well, human bodies disturb magnetic fields...

[JanneM]

The whole point is that the magnetic field is locally disturbed by buildings and other structures. The variation in a small area is what identifies the place.

This is like saying maps are useless because there's buildings placed in the terrain.

Re:Well, human bodies disturb magnetic fields...

[Assmasher]

Not really, it's more like saying "maps are useless if a lot of things that seem to look like things on the map are moving around you at all times."

Re:Well, human bodies disturb magnetic fields...

[JanneM]

Buildings and other large structures rarely change much. The magnetic field in any particular spot may change due to objects moving about, but that acts as spatial high-frequency noise over a much more stable lower-frequency signal.

The simily is imperfect, but when you compare the map with reality you ignore the small stuff like parked cars, curbside junk, temporary road-works structures and so on and focus on the stable larger-scale structures in the environment. In the same way, this would loook at the larger-scale structures and pinpoint you within that. A corollary is that while the fix may be quite good, you won't actually get a fix until you've moved about a fair bit. so the device can sample the signal across a larger spatial area.

Then you have the SLAM aspect as well: as you move about, the incidental changes get tentatively incorporated into the updated map. Those changes that prove stable over time become a permanent part of the map. For fairly busy areas you would have quite high accuracy at all times, even if people did try to actively sabotage it.

Re:Well, human bodies disturb magnetic fields...

[Assmasher]

I understand what you're saying, but I have to disagree because you appear to be quite narrowly stipulating the usage of the technology for the purpose of your analogy.

The magnetic field in any particular spot can change for any number of reasons including, for example, the simple scenario of someone using electrical equipment. Something as simple as an electrical device being plugged into the wall, someone placing a computer based kiosk in the area, a television or monitor being turned on, someone using a floor polisher, fans being turned on overhead, lights being turned on/off, et cetera, ad nauseum. This ignores people in retail environments moving displays, furniture, and product relocation.

You also appear to treat the system as a pattern recognizer that grows in accuracy as samples increase and are correlated as 'accurate' - where are you getting this from? It seems very unlikely that without a beacon based system where the app could recalibrate as it approached (not a terrible idea really) how could you confirm the validity of the introduced changes to the magnetic field map? In computer vision we do this with training that requires user interaction.

I think it could have some interesting uses in a broader environment but claim sub-meter accuracy in an active location seems hyped (although I could be totally wrong - in fact, I hope they're on to something because I occasionally do IPS related work and I'd love to be able to push a good mobile outdoor-indoor transitioning system.

BTW, enjoyed your blog :).

Re:Well, human bodies disturb magnetic fields...

[JanneM]

Good points.

I think the main observation to take is that very rapid, unpredictable changes (a floor polisher, say) also tends to introduce only very _local_ changes. The effects stretch for decimeters, rather than meters, and for seconds or so when the machine is close. A larger, fixed machine that is either on or off, on the other hand, would essentially introduce two alternate local maps, both of which could be estimated and learned.

Remember that these sensors are fairly insensitive to small changes; they have to happen really close for it to have any effect.

"how could you confirm the validity of the introduced changes to the magnetic field map? In computer vision we do this with training that requires user interaction."

To be fair, there's plenty of systems that do estimate validity of changes autonomously. Remember that vision poses a rather different set of issues such as rapid and radical lighting changes that are absent here.

"[...] but claim sub-meter accuracy in an active location seems hyped"

I think the sub-meter accuracy is completely achievable. But don't expect to, say, drop a robot in a spot and it will know where it is immediately or anything. I would assume it would take several minutes of moving about and sampling a largish area before you could identify the approcimate position, separate out local, transient changes from the underlying signal and get a good fix. The move variable the loval environment, the larger a total area would you need to sample to become sufficiently confident you are looking at the right place.

re: blog - thanks! Though I mostly post smaller stuff on Google+ nowadays: https://plus.google.com/105059362788808645801

Re:Well, human bodies disturb magnetic fields...

[Assmasher]

I agree with regards to the size of the changes and the ways they could be overcome; however, that is an academic exercise that assumes many conditions. This is being presented as a consumer technology for mapping interiors for consumption by consumer mobile devices.

Regarding computer visions, I wasn't referring to background learning, I was referring to kernels we use to recognize objects in the scene (speed samples, orientation change rates, sizes in given dimensions against camera orientation,et cetera.)

I think sub-meter accuracy is possible in academic conditions, I certainly don't think it is possible with current mobile devices in a 'noisy' environment though.

Let's hope I'm demonstrably wrong! :)

Re:Something looks... not quite right about the vi

[million_monkeys]

I can't pinpoint it exactly, but the way the phone is being held doesn't seem quite right. Like it was faked or done on a computer. Is this really just a concept video?

The camera is fixed on the phone like they had some kind of brace mounting on his arm. It makes for a steady shot, but it doesn't look natural at all. It looks like a 3D shooter where the view is always centered on the gun.

I also noticed that when focused on the phone, the rest of the world is a blur. I'm not going to enjoy dodging people who are staring down at their phones while walking, completely oblivious to what's going on around them. It would, however, be hilarious to mess with the magnetic field enough to disrupt the navigation and see how many people you can get to walk straight into a wall.

Re:Compass

[X0563511]

It isn't a joke if it's not funny.

Use case

[Narrowband]

Reliable indoor positioning is probably one of the key pre-requisites to building workable augmented-reality apps like games and such. There's probably a real payoff for the long term, but maybe not for the group that invents the underlying concepts/tech, unless they find some way to see it through to applications.

You are currently in...

[wonkey_monkey]

...the basement. You are likely to be eaten by a grue.

Re:Something looks... not quite right about the vi

[mark-t]

To be fair, you can't significantly alter the earth's magnet field with anything particularly portable... while the field strength of the earth's field is quite low compared to what we can generate artificially, the size of the field is so many orders of magnitude larger than anything man-made, that owing to how fast strength drops off with distance, I simply don't think there's too much to worry about with regards to man-made objects changing what the perceived field is unless you are within mere inches of an alternate magnetic field source.. Theoretically, at least, I think the general concept behind this is valid.

who knows?

[farble1670]

Second, and perhaps more importantly, what about the effect of objects that aren’t part of the original magnetic field map, such as cars, moving lifts, and electric motors? Who knows, though — perhaps the effect of these objects is negligible.

well, most everyone knows actually. the effects of even a slab of metal is significant. the effects of powered devices like a monitor, or an electric motor, or a stationary magnet are massive. anyone can see it. grab an app that displays the magnetic levels from your smart phone's sensor, and watch them fluctuate wildly as you move it past various electronic devices.

i guess it's okay though because places like malls aren't known for having electronic devices about ... oh wait. maybe they have some great technology to get around these problems, but TFA doesn't provide any inkling of that.

Re:Use Case

[zlives]

this does not detect the higgs field

Re:Something looks... not quite right about the vi

[hamjudo]

A "Helmholtz coil", is actually a pair of coils, that will produce a uniform magnetic field in a cylindrical region between the coils. A "Maxwell coil" is a pair of coils wired to produce a cylindrical magnetic field with a linear gradient between the coils.

Make a pair of big coils, put some power through it, and you can make a big electromagnetic field. Depending on how you connect the coils, the magnetic field will have interesting properties. With simple electronics, you can vary the field strength between your coils.

Re:Something looks... not quite right about the vi

[million_monkeys]

I'll admit i wasn't sure exactly how to make those changes, but it would be hilarious none the less. Probably much easier to just put up a temporary wall that isn't on the phone's map. I'm sure some people would be so focused on starting at their phones that they would walk right into it. For wild animals, natural selection deals with that sort of obliviousness for any species that has predators. Sometimes I worry that technology lets us get away with being more incompetent that we should really be allowed to be.

Earth inductor compass

[aklinux]

This sounds like it might be an extension of the technology behind the Earth Inductor Compass that's been around since the early 20th century. http://en.wikipedia.org/wiki/Earth_inductor_compass

Easy!

[MarcoPon]

I suppose they modulate the magnetometer with a transphasic ringtone, while diverting power from the non essential functions to the positioning system, and firing a synchronized serie of photon beams trough the back facing camera's LED. To activate, just say: "Engage".

Re:Ummmm, the Earths magnetic field is NOT static.

[Enokcc]

You know, nothing is static. Some things are just static enough for certain applications. You could also argue that geographic maps are useless since continents drift all over the place over time.

inertial navigation

[dirtyhippie]

my phone has a couple of gyroscopes. is the error from these so bad that it can't keep track of my position while i'm inside a mall? if so, why is it there at all?

Earth's magnetic field stable ?

[Lennie]

I thought the Earth's magnetic field isn't really stable, it changes slightly over time.

Re:Compass

[nospam007]

Pigeons are way ahead of them, they follow the highways, much easier than using tiny thingies in their cherry-sized brain.

"Pigeons take the highway
Some birds follow roads instead of flying direct."

http://www.nature.com/news/2004/040210/full/news040209-1.html

Re:Compass

[camperdave]

Basic compass navigation has two data points: North and South.

You've never used a compass for navigation, have you?