[peppercorn (OP)]

GPS can be patchy - or just plain wrong on occasion.

When the commercial self driving cars do come along, is it likely that they will need an alternative or 'double checking' ground based system in place to remain safe and reliable?

Is it fundamentally important that such systems will need to know their exact position (and velocity) at all times? The present Go*gle car system needs to have the route driven out first but is then safe to navigate even a long trip 'alone'. Problems occur in heavy rain and with snow hiding the landscape.

This seems relevant

[CliffordK]

The technology for GPS Base Stations has been around for quite some time, and significantly improves one's GPS accuracy.  Trimble does maintain a list of base stations around the world, but I don't think there is a unified effort to add base stations.  The advantage of using a base station is that it gives a user millimeter accuracy with one's GPS, and may also aid with improving vertical accuracy.

Some kind of an active sonar system should also allow a dynamic mapping without GPS, and, of course, also using visual stimuli.

For driving, it might be handy to put base stations in tunnels, but I would think that an automated driving system would have multiple ways to find its lanes.  If it requires a "drive-through", how does it deal with unforeseen occurrences such as detours, blocked traffic lanes, finding parking spaces, & etc?

Could one build smart lane markers?

[evan_au]

Knowing where you are...
Civilian GPS only claims 100m accuracy, and can be worse than that in city canyons, heavy rain or thick forest, and non-existent in tunnels. Some GPS algorithm improvements claim better accuracy from the civilian GPS channel.

When you turn a corner, that allows a GPS receiver to calibrate its location to better than the normal 100m range.

When Google collected their streetview data, they also detected the location of WiFi hotspots. Location of mobile base stations provides another point of reference (but its easier for the base station operator to determine your location than it is for you to determine your own location).

Cheap and ubiquitous digital camera chips provide effective sensors to detect the immediate surroundings of a vehicle. This is where you really need to know your position to within centimeters/inches. The image processing to identify lanes, trucks, cars, pedestrians, cyclists, dogs and edges of the road is a real challenge (especially with rain, snow and driving into the setting sun), but some high-end vehicles now have alerts if you start to veer out of your lane without first putting on your direction indicators.

A recent chip announcement for mobile devices supports a third dimension of location data - height. If street mapping data included height information, it is possible to recalibrate position based on hills on the road. A resolution of 1 foot/0.3m is claimed, but it requires calibration to cancel variations in atmospheric pressure.

Knowing what others are doing..
Research vehicles are often fitted with more expensive laser rangefinders or radar to measure the relative speed and distance of nearby vehicles.

Knowing how to get to your destination..
Digital maps are now extensive. But roadworks and detours requires considerable flexibility. Perhaps the location of blockages and detours could be broadcast via the internet, just as information on traffic congestion is available for some cities.

[CliffordK]

Your cell phone may not be very accurate.  However, commercial GPS survey equipment such as the Trimble equipment can get you down to about 30cm accuracy without a base station, and about 1cm accuracy with a base station.

There are export restrictions on the equipment, so I don't know if the same equipment is available in the UK and Australia.

Of course, traveling at 70 MPH around buildings, trees, and other obstructions can put additional demands on the equipment.

It looks like at least Missouri has invested in a statewide GPS base station network.  I'm not sure about other states or a national network.
http://gpsweb.modot.mo.gov/
http://gpsweb.modot.mo.gov/network.php

It looks like their stations are about 30 miles apart or so.  Perhaps not quite high enough density for super-accurate measurements, but it would allow dynamic calibration of one's equipment as one drives. 

It would make sense to put a GPS base station on every cell phone tower.  The government licenses the airwaves, they could add a requirement to add GPS.

[peppercorn (OP)]

I wonder what has driven the US State of Missouri to implement this network? And are similar schemes likely to be rolled out across other states or countries?

What about light aircraft? Can they also make use of these ground based 'beacons'?

There must be a whole host of devices in the 'internet of things' age which could take advantage of knowing their precise position (especially in 3D), I would think.

[evan_au]

Differential GPS has been used for aircraft and boat navigation in foggy conditions, where there is a clear value in improved safety.
The accuracy is still quite good over a radius of 100km or more around the base station transmitting the correction signal.
However, the GPS signal and its correction signal may have trouble reaching tunnels and underground car parks.

See: http://en.wikipedia.org/wiki/Differential_GPS

[CliffordK]

I wonder what has driven the US State of Missouri to implement this network? And are similar schemes likely to be rolled out across other states or countries?

I'm not quite sure.  I just bumped into it.  I don't know if other states are doing it too.  But, it may help with tracking trucks, trains, cargo, etc. 

I believe the typical GPS is unidirectional, but one could easily make a web that was bidirectional, sending position information, but also receiving information about cargo.

What about light aircraft? Can they also make use of these ground based 'beacons'?

I don't know about light aircraft, but the jets certainly could use GPS with autopilots.  One typically gets a clear view of the sky from in the sky, so one should receive the maximum number of satellites.  100 feet accuracy is probably just fine for the jet, except for collision avoidance.  Even 100 feet vertical would generally be ok, except when landing. 

A couple of base stations around an airport would be helpful.  Everything seems to discuss a single base station, but it would seem like one could add 2 or 3 of them in strategic spots around the airport to get good triangulation.

Did you see the Die Hard 2 movie in which somehow the terrorists reset the elevation parameters for the runway?  There is always an advantage of looking out the window to see what is around oneself.

[alancalverd]

Augmented GPS is approved for precision approach at a few airports, and for nonprecision approach in rather more.  It's a lot more flexible than ILS which is runway-specific, and supposedly easier to use - I'll let you know when I can afford it! En-route GPS can talk to your autopilot: it takes the boredom out of longhaul and the fun out of light aviation. But GPS doesn't do collision avoidance, which is more important than position and direction vectors when driving.

[evan_au]

A couple of base stations around an airport would be helpful. Everything seems to discuss a single base station...

The factors that are cancelled by the base station include:

• Intentional uncertainty in the "civilian" GPS data stream, intended to prevent unfriendly military applications
  
• Variation in signal propagation through the atmosphere, due to varying space weather
  
• Clock drift in the satellite atomic clocks
  
• All of these corrections can be measured and broadcast by a single base station.
  
• Of these, only the atmospheric signal conditions change much from place to place, so one base station can easily cover a radius of 30km

GPS does use "triangulation" (or "quadulation"*), but it is triangulation from the position of the satellites, not triangulation from the position of the base station(s). The Base station(s) merely transmit a low bit-rate correction signal.

* Triangulation works if you know the distance to 3 known points on the surface of the Earth; you can solve 3 equations in 3 unknowns to determine your position in 3 dimensions. However, with GPS, you know the position of the satellites (because they continually broadcast it), but you initially don't know the distance to any of them. So you must solve for 3 dimensions of space + 1 of time; you need to monitor at least 4 satellites, and solve 4 equations in 4 unknowns to determine your position and time, hence "quadulation" (if there is such a word).

[evan_au]

An interesting episode on autonomous vehicles, including different approaches to the navigation aspects (2 hour podcast):
http://omegataupodcast.net/2014/02/140-self-driving-cars/