3 ms·

If you're navigating on a boat, every boat has it's own deviation chart. When you then use the compass you factor in variation (what the chart says the differen

If you're navigating on a boat, every boat has it's own deviation chart. When you then use the compass you factor in variation (what the chart says the difference is between mag north and true north) and the personalized deviation chart for that boat. To set this up, some guy comes down to the dock, takes your boat out, does a few 360's with known landmarks in site and compiles a table of how your boat's compass varies from standard magnetic north. Big things like the engine, generator, keel, etc will influence what your deviation chart looks like. Usually it's less than 4 degrees at each point of the compass IIRC. So every place in the world has metal objects that create that place's own distinct deviation from where magnetic north should be. So perhaps that's what these guys are using. A known table of variation and then looking at the deviation. What I wonder about is what happens when someone turns on a 12 amp vacuum cleaner, a monitor nearby goes into powersaving mode, someone moves their laptop or someone with their own cellphone walks past. These things emit magnetic fields, so I wonder how they've solved the problem of the local deviation changing constantly.

famousactress 14y ago
It's possible that constant change would be dealt with by probability and consensus. Compared to your best historical guesses at this location, how likely is it that the reading you're getting is accurate? You can establish high or low confidence and disregard sudden changes. Also, the more nodes (people, phones) you have the more you can build consensus between them, use the higher confidence predictions as waypoints for 'fixing' lower confidence ones.
baddox 14y ago
> What I wonder about is what happens when someone turns on a 12 amp vacuum cleaner, a monitor nearby goes into powersaving mode, someone moves their laptop or someone with their own cellphone walks past. These things emit magnetic fields, so I wonder how they've solved the problem of the local deviation changing constantly. Are any of those magnetic disturbances even within an order of magnitude of a smartphone's compass threshold? I'm surprised that the disturbance from a skyscraper's skeleton would be detectable, and a computer monitor seems like even more of a stretch.
- kmm 14y ago
  It's clear some disturbances must be measurable because as I walk around the room, the magnetic field measured by my smartphone constantly changes and since the magnetic field of the Earth is practically uniform over small distances, this must mean the disturbances come from my surroundings. However, when toggling some apparatuses in my room, nothing happened. My laptop doesn't have a hard drive though, but even only centimetres from it, no change was measurable. Neither did anything happen near my lamps or the AC/DC transformator. The only place I measured a difference (<10%) was on the stove (why yes I'll put my smartphone worth hundreds of euros on a hot stove). However, this faded into the background at a distance of about 10 cm.
endtime 14y ago
As famousactress may have been alluding to, this is equivalent to the well understood problem of robot localization in dynamic environments. If my robot has a map of the building and it thinks it knows where it is, will it get confused if someone walks in front of it? Not if I reserve some probability mass for the possibility of seeing an obstacle at any distance (in the range of the sensor). For some background, see these slides (PPTX): http://pub1.willowgarage.com/~konolige/cs225B/slides/09-sensor_model.pptx http://pub1.willowgarage.com/~konolige/cs225B/slides/09-sens... (via http://pr.willowgarage.com/wiki/CS225B/Handouts http://pr.willowgarage.com/wiki/CS225B/Handouts)

famousactress 14y ago

It's possible that constant change would be dealt with by probability and consensus. Compared to your best historical guesses at this location, how likely is it that the reading you're getting is accurate? You can establish high or low confidence and disregard sudden changes. Also, the more nodes (people, phones) you have the more you can build consensus between them, use the higher confidence predictions as waypoints for 'fixing' lower confidence ones.

baddox 14y ago

> What I wonder about is what happens when someone turns on a 12 amp vacuum cleaner, a monitor nearby goes into powersaving mode, someone moves their laptop or someone with their own cellphone walks past. These things emit magnetic fields, so I wonder how they've solved the problem of the local deviation changing constantly. Are any of those magnetic disturbances even within an order of magnitude of a smartphone's compass threshold? I'm surprised that the disturbance from a skyscraper's skeleton would be detectable, and a computer monitor seems like even more of a stretch.

kmm 14y ago
It's clear some disturbances must be measurable because as I walk around the room, the magnetic field measured by my smartphone constantly changes and since the magnetic field of the Earth is practically uniform over small distances, this must mean the disturbances come from my surroundings. However, when toggling some apparatuses in my room, nothing happened. My laptop doesn't have a hard drive though, but even only centimetres from it, no change was measurable. Neither did anything happen near my lamps or the AC/DC transformator. The only place I measured a difference (<10%) was on the stove (why yes I'll put my smartphone worth hundreds of euros on a hot stove). However, this faded into the background at a distance of about 10 cm.

endtime 14y ago

As famousactress may have been alluding to, this is equivalent to the well understood problem of robot localization in dynamic environments. If my robot has a map of the building and it thinks it knows where it is, will it get confused if someone walks in front of it? Not if I reserve some probability mass for the possibility of seeing an obstacle at any distance (in the range of the sensor). For some background, see these slides (PPTX): http://pub1.willowgarage.com/~konolige/cs225B/slides/09-sensor_model.pptx http://pub1.willowgarage.com/~konolige/cs225B/slides/09-sens... (via http://pr.willowgarage.com/wiki/CS225B/Handouts http://pr.willowgarage.com/wiki/CS225B/Handouts)