E911: Indoor Positioning is Embarking on a New Arms Race

If you ever need to make a call to the emergency services, and you have a choice, use a traditional landline.

That’s the most important thing to remember from this post. “But why?”, I hear you cry. Well, it’s pretty simple.

The land line is at a known, fixed location – the location that a telecommunications engineer went to at some point in the past to set up the telephone socket. That location is held on a database by the telecommunications operator. If the emergency caller cannot pass on any useful location information to the 999 or 911 operator, then the telecommunications provider can pass on the location of the telephone in a few seconds. In most areas this entire process has been automated and the emergency operator sees the details of the location of the telephone as soon as the call is connected.

When an emergency call is made on a mobile cellular telephone, all bets are off. There is no guarantee that the billing address of that mobile telephone is where the call is currently coming from – and of course there is no billing address associated with pay-as-you-go SIM cards. The inherent positioning performance associated with a cellular connection is given by the coverage area of the serving cellular base station. The identity of the service base station (the Cell-ID) is known to the telecomms operator, and they have some idea as to where this base station is. The cellular telephone is therefore within range of this transmitter. Well, the range of that transmitter can be as small as a few hundred metres in a busy city, or as large as 35 kilometres in a rural area. Either way, the location of the emergency caller cannot be confidently determined.

This has been a known problem for decades, and the original E911 directive in the USA set goals on the American telecomms operators in the 1990s to ensure provision of a positioning capability for 911 calls with an accuracy of 300 metres or better. We have a similar push in Europe: the E112 directive. As GPS chips have become standard in cellular phones, E911 capabilities have converged on this single technology as the provider of a position fix. This was okay until people started removing landlines completely and moved to using their cellphones indoors and outdoors as their main telephone. The problem being of course, that GPS receivers do not normally work indoors, as they require an unobstructed view of the satellites.

So where do we stand today? The FCC estimates that over 70% of all emergency calls are now made on cellular phones ,with over half of those calls being made from an indoor location. Further, we know that around 2% of all emergency calls need to be located by the operator because the caller is unable to provide any location information (they have no idea where they are, or are panicking, or are a very small child, or cannot speak the language well enough, or their speech is incapacitated by their medical emergency, etc.). These numbers lead to an incredibly sobering realisation – around 10,000 people die in the USA every year now because they are making emergency calls on un-locatable cellphones. That figure extrapolates up to over a million people worldwide each year. That’s unacceptable.

Thankfully, things are changing. The FCC released an updated mandate last month. New targets have been set for mobile operators. Those targets are:

• Wireless 911 Calls must be located horizontally to within 50m at the following success rates over the coming years:

• 40% of all calls within 2 years

• 50% of all calls within 3 years

• 60% of all calls within 4 years

• 70% of all calls within 5 years

• 80% of all calls within 6 years

• The vertical location of these calls must be provided using barometric data in the short term (3 years) but a new "z-axis technology" is to be developed and implemented within 8 years.

It is excellent news that there is a new mandate, and that enhanced indoor positioning technologies will be developed. This E911 mandate of course only affects the USA, but the forced development of new positioning capabilities for emergency services within the USA will inevitably permit these capabilities to extend to cellphones and cellular operators globally. With luck the E112 mandate for Europe will be updated in the coming months to also address this serious location problem in Europe.

“Hang on!”, I hear you cry. “My smartphone has that blue dot thingy indoors.”, I hear you assert. Yes, smartphones already provide indoor positioning solutions. However, there is no mechanism available, or planned, that can pass this third-party-derived position solution to the E911 operator. And should there be? Those positioning systems have not been designed for safety-of-life purposes, and they never should be used for this purpose, because their integrity cannot be guaranteed. Fingerprinting is not a reliable, dedicated positioning system with quantifiable performance statistics, unlike systems based on global positioning satellites or on telecommunications networks. Let me explain:

Fingerprinting involves recording a scan of WiFi, BLE and Cellular measurements at a location and storing this information in a database. Later on, a scan can be triggered and the measurements compared to this database, and the location can be estimated as long as there are matching readings in the database. So the only way this technology works is by surveying everywhere you might need to be located in the future. This is a chicken-and-egg problem, you need a way of locating the receiver in order to correctly geo-reference the fingerprint measurement. You cannot easily survey the planet in this way – sure, large retail outlets, main shopping streets, airports, museums, are all fair game. But every home in the USA? Every home on the planet? That involves a serious surveying capability. Companies like Skyhook, Nokia, Apple, Google, etc. have been generating these vast databases through manual surveys and crowdsourcing for decades, and they are certainly worth a lot of money in the indoor positioning market. Google has used its streetview cars to generate and maintain its fingerprint databases at the same time as generate its streetview mapping imagery. They also crowdsource data from users of Google Maps by surveying fingerprints with smartphone GPS position fixes. The databases need to be constantly re-surveyed to ensure accuracy and integrity, they need to be “healed” and regularly updated as time goes one – WiFi access points move to new locations as people and businesses relocate, some cellular operators regularly change their Cell-IDs (to prevent third parties generating fingerprint databases!), and surveying can be flawed (the database is only as good as the location estimate during surveying).

Large corporate entities spend money generating and maintaining these databases continuously because they can generate revenue from the data and by providing location based services to their subscribers. Can telecomms operators perform similar surveys at great cost when they cannot monetise this data so easily? Will the big commercial database owners allow their databases to be used for safety of life purposes? I doubt it, that would be opening themselves up to a flurry of court cases. If a fingerprint location is incorrect and an ambulance is sent to the wrong part of a city, and someone dies, will attempts be made to sue the fingerprint database provider? It is very unlikely that fingerprint database owners will want to get into this space when they do not need to and when the risks for them are likely to outweigh any licence fees for the position fixes. Crowdsourcing fingerprint locations deep indoors a long time after the most recent GPS fix is also still a research problem, not a solved problem. Techniques involving Simultaneous Localisation and Mapping have been demonstrated by myself and others to provide a usable solution, along with pedestrian dead reckoning schemes, but the only way to really test the overall performance of these systems with real end users is by accessing the vast amounts of crowd-sourced data available to Apple, Google, Nokia, Microsoft, etc., so only those companies know for sure. The information on the performance of their systems is a closely-guarded secret.

What are the alternative technologies? There are proposals to deploy dedicated positioning beacons in cities to provide the indoor positioning coverage but this approach is infeasible. The need for new infrastructure plus new technology in the phones represents too great a cost for this solution to be viable. So what about existing infrastructure? Use of the mobile phone signals themselves is certainly possible and is a well-trodden path in radio positioning. The issue there is that there are so many different types of cellular signal to support. GSM 2G, 2.5G, CDMA, 3G WCDMA, UMTS, 4G LTE. A practical solution for the USA would be to convert completely over to 4G LTE and make use of its dedicated ranging signal, but this again constitutes great cost in the form of replacing all of the legacy transmitters in the country that provide service in many areas using older technologies (2G and 3G). Timing-based positioning also provides high positioning accuracy outdoors, but the multipath interference experienced indoors reduces the positioning performance. There is a risk that the FCC performance requirements of 50 metre positioning might not actually be met by 4G LTE cellular positioning using time-of-flight measurements.

So what’s my assessment? I think there is only one way to really solve this problem such that accuracy, availability and integrity are at high enough levels for safety-of-life purposes – and that is to pick up GPS and other satellite positioning signals indoors, and fuse the information with cellular timing measurements. We know that the GNSS signals are too weak to be picked up indoors at the moment, and this is the true challenge that needs to be overcome. Boosting the satellite signals with Earth-wide coverage is not possible without providing power sources at the satellites much more capable than solar panels. GPS military upgrades are likely to include spot-beam antennas able to boost power to a small region of the Earth during times of war, but not capable of boosting global signal power. That leaves receiver technology. High sensitivity receivers exist but increasing sensitivity increases susceptibility to multipath interference, reducing the positioning accuracy. Long signal integration times an be used by a static receiver, but not a mobile one. These issues will be the focus of my research for the second half of this year and beyond, and their resolution will have a huge impact on the E911 problem. I believe is it possible to make new gains in receiver signal processing and sensor fusion schemes in order to increase the gain of GNSS receivers while reducing the susceptibility of GNSS and cellular timing measurements to multipath interference. So watch this space…

The FCC have provided important new goals in addressing a serious safety-of-life issue, and with radio-positioning at the heart of the problem. The race for the solution is on.

#E911 #E111 #IndoorPositioning #PNT #Smartphone