Article by Simon Taylor

Introduction

This paper investigates a methodology that can be applied to the development of an Automatic Vehicle Location (AVL) System. In essence, any AVL system can be considered as two main centres: the vehicle or mobile system and the central site or control system. This paper will consider components that can be used in both of these systems, the latency of location and the costs of the communication systems involved.

Recent events in the UK have generated a lot of interest in tracking systems which are designed to be worn or can locate children. This is a very interesting area, as considering these applications causes one to look very closely at power, antenna and implementation issues, this is discussed later in this paper.

The Mobile System

The mobile system will normally contain three core components: The location system, a controller and a communications system to allow the central site to contact the mobile. By far the best-known and best-established location system is the Global Positioning System (GPS). The features of this system have been widely publicised, but some are worth re-iterating here as well as making clear some limitations of the system for the best performance.

Although we will discuss the GPS system here, it is worth mentioning other location systems and their comparative benefits and disadvantages.

Cellular location

It is theoretically possible to quite accurately locate a mobile or radio user by triangulation and received signal strength from three radio receivers. Cambridge Positioning Systems have developed the service based upon timing of the mobile telephone signals to base stations. However, providing this as a commercial service has presented a challenge to mobile operators and is not generally available. Other problems exist in remote locations where in many cases only one base station is available, so preventing normal triangulation.

Other methods include Cell ID, but currently this system only operates when a circuit-switched call is initiated, and reports the currently logged cell into the mobile telephone network. This has been announced as a simple ‘get me a Taxi’ style of service, but as the location can be quite inaccurate (Orange have quoted up to 1km in suburban areas). Most UK networks are offering this service but for true location information again, the accuracy can be very restrictive.

Global Positioning System

GPS operates by receiving time information from initially four, but in operation at least three satellites. The satellites are continuously orbiting the earth, and with 24 satellites in the system as specified provide a theoretical maximum visible number of satellites of 12 from any location on the planet due to their high orbit (20,000km). As only three satellites are required for navigation and in fact at the time of writing there are 28 operational satellites in use, coverage is now very good.

To get a good idea of the geometries involved, NASA have a very useful JAVA application which shows satellite positions in 3D at their ‘lift-off’ web site. This application allows the user to see the current position of many satellites, including geo-stationary, LEO (Low Earth Orbit) and GPS satellites, clearly showing their unusual elliptical orbits.

The signals from these satellites used to have intentional errors or “selective availability” (SA) introduced on them which produced position information with accuracy of 100 metres in the horizontal plane, and 150 metres in the vertical plane 95% of the time. This was originally introduced to avoid the misuse of the globally available system, but since the ‘falling of the iron curtain’, the US Department of Defense who provide the system (they have their own high accuracy system for military use) have removed this error, now giving typical accuracies of 10 metres.

The commercial (L1) service is provided without fees or a requirement for a licence.

If an improved accuracy signal is required, then it can be obtained using differential error cancellation techniques, but many vehicular applications can avoid this using simple tracking algorithms.

For example, consider a vehicle, which has broken down and is stopped on the hard shoulder of a motorway. The driver is confused, and does not know if he is travelling north or south. There is a location system which sends the cars location to the breakdown service and possibly offers a voice communication with the driver. It could be said that a possible error of 100 or even 10 metres would not pinpoint the vehicle on either of the two carriageways. True enough, but if the system had logged the positions of the vehicle while it was travelling, just before it stopped, then the last known direction of travel is known, and hence the appropriate carriageway.

For AVL systems which do require more accurate positions, differential GPS can be employed. These systems normally employ the transmission of correction information to the GPS receiver, this correction information has corrections for each satellite in view. This is done because each satellite has it’s own error, the error in GPS is not simply a X-Y error which will be the same for all receivers. The error on any two given receivers will only be the same if those receivers are using the SAME satellites. This can’t normally be guaranteed as satellites may be obscured at one location, making the error slightly different for two receivers.

Just becoming widely available are the WAAS (Wide Area Augmentation Service) and EGNOS (European Geostationary Navigation Overlay Service) systems which provide differential calculations free of charge, many GPS receivers will take advantage of these signals. The interesting aspect of these services is that they are based on geo-stationary satellites (GPS satellites are continuously moving) which transmit on exactly the same frequency as the GPS satellites themselves. This gives the user the advantage that he does not have to employ another antenna or receiver to get the accuracy benefits of differential GPS. Just an operational note – WAAS is already running, giving the USA coverage and EGNOS is currently running via INMARSAT, but is officially a test service, expected to go live early in 2003. 

At this point, some commonly mistaken points with GPS are worth clarifying:

GPS requires line-of-sight view to the satellites. The frequency of the radio signal is 1575.42MHz, and is easily obstructed. If the signal is reflected, then the range determination of the satellite will be incorrect, distorting the position. Most modern GPS receivers will detect such reflections and reject such signals.

GPS provides the location of the antenna, not the receiver. This can be important with large vehicles and where differential calculations are employed, but can normally be ignored.

Comparative Accuracies

The Mobile Unit

GPS receivers suitable for OEM integration are widely available. Such devices, the Conexant Jupiter LP or Leadtek 9543 are small, easily integratable modules. We will return to these, but in order to communicate with a base station, the GPS unit must have a transceiver of some sort. The exception to this may be items such as hand-held data loggers which record position data in the field, and are returned to base to download data.

GSM Modems

For this paper, we shall consider the use of the GSM (Global System Mobile) cellular telephone system. Inherently digital, this system offers data connections at 9600bps in most countries and 14400bps in others. There are many dynamics in this market, one of which is the Wireless Application Protocol (WAP) which is receiving great press at this time. WAP for this application does not offer much at this time as the communication is still session-based, i.e. a full-time data call is made, there are not many opportunities for call cost reduction. With the General Packet Radio Service (GPRS) which is now available, the individual packets of data are able to be sent rather than dialling up, thereby providing the possibility of being charged for the data you send, rather than for the duration of a call. A simple GPS location is typically no larger than about 60 characters, so this may well be an advantage. Further away are W-CDMA or UMTS services, providing much faster data rates, but are not expected until 2004. These ‘3G’ services have further standards confusions at the time of writing, with CDMA-2000 being used in the US, and the W-CDMA which is licensed for Europe providing some interconnection problems.

Modules are available which are basically cut down mobile telephones. These modules (Siemens MC45 is shown as an example) provide data services, voice and SMS in very small compact devices. They require an antenna, and control via a serial interface. ‘AT’ commands are used, as employed my most modems in use today. These modules can be integrated into very small units, the module shown here is 53mm x 34 mm x 3mm in size.

Existing data services in the UK include 9600bps data, most UK networks apart from Orange do not support 14,400bps at this time, either as a standard or UDI data call. The latter connects to ISDN services, and avoids the use of modem training sequences which take about 20 seconds of a call – chargeable time in which you cannot transfer any data. UDI allows data to flow almost immediately. The modem training sequence happens because it may be the case that a user is calling the mobile using a normal analogue modem. In this case, the modem dials a gateway modem at the network operator, performs a normal modem handshake (the ping-ping noises at the beginning of a modem call), and then the data is entirely digital from there on.

Cost of Calls

GSM tariffing quickly becomes an important factor, we will consider this now. The table shows different amounts of data and the relative costs of communicating that data over the GSM networks. A number of assumptions are made (of course different networks will make different charges, these are constantly changing) :

1) An SMS costs 10p.
2) An SMS is sent to request data, and data is then returned, so one more SMS than needed to transfer the actual data is always sent.
3) Data calls cost 10p per minute (typical inter-network telephone costs).
4) A standard data call takes 20 seconds for the modem handshake.
5) An ISDN (or UDI) data call takes only 2 seconds before data can be transferred.

It will be noticed that in all cases, the UDI call is considerably cheaper than either SMS or standard data. This is due to the relatively high data rate, and the low latency before data can be transferred. It is commonly thought that SMS will be more cost efficient, but from the table it clearly is not, even for very low data consumption (for large amounts of data SMS is obviously not an option). Also the fact that if an SMS message is sent to request a position, and a position reply is sent, the whole process will take about 15 seconds, whereas using a data mode, once connected, response will be very fast.

SMS may well be considered for applications where the remote may temporarily go out of GSM coverage. In the UK however, this is reasonably rare, as over the last few years so much infrastructure has been added to the GSM networks that coverage is now very good, even for low power GSM products such as portable cellular telephones. SMS is a store and forward system, whereas either standard or UDI data will require the mobile to be within reach.

If the mobile is required to receive differential corrections, then these are straightforward to send when a two-way data link is established. Using SMS, it would be possible to send a batch of differential corrections in the position request message, but the time lag involved with SMS will compromise the position accuracy a little. WAAS/EGNOS capable GPS receivers remove the requirement to send differential corrections via GSM at all.

GPRS

GPRS provides for an ‘always-on’ facility which can allow the remote device to be contacted at any time. Contrary to some popular belief, GPRS does not necessarily offer a massive increase in speed over normal circuit-switched data calls. GPRS makes used of unused ‘timeslots’ in the GSM transmission. There are up to eight timeslots available for both transmit and receive in each RF channel. However, it is unlikely that the networks will either allow all eight timeslots to be used for one user, or that a device will be manufactured that will allow transmission using all eight timeslots. With this taken into account, and the error-correction schemes in use, typical data rates are currently about a maximum of 20000bps transmit and 40000bps receive.

However, GPRS does provide a direct Internet connection, and making use of such a facility allows data to be transferred irrespective of distance or national borders for a fixed cost.
Detailed technical information for GPRS is available at the GSM Association’s website. 

The GPS receiver

We have mentioned the Conexant Jupiter, a “12-channel” GPS receiver. The theoretical maximum number of satellites in view is 12 from an equatorial position (the satellites orbit in elliptical orbits, centred on the equator). With a constellation of 24 satellites, a more common maximum from the UK for example is 10, however with the current ‘overperforming’ constellation of 28, we have once seen 12 in view from Basingstoke. At lower latitudes, 12 visible satellites are very common. The twelve channels allow the receiver to allocate one RF channel for each satellite, allowing the fastest carrier lock once a satellite comes back into view after obscuration. Older, single channel receivers would receive a satellite at a time, and have to estimate the other satellites once the final satellite was received for the position solution.

Data is provided from the GPS receiver in two formats: NMEA (National Maritime Electronics Association) – a industry standard format which operates at 4,800bps, and provides positional output in ASCII. This can be the easiest method to use if a simple dumb-modem approach is taken. (see below). However, for more complex applications where some control over the receiver is required (time initialisation, satellite selection, antenna selection, satellite information etc.), NMEA does not provide a standard input message to set the receiver. The proprietary binary mode does provide such a set of functions, and although it sounds complicated, provides a full set of functionality without being very difficult to use. TDC can provide sample data in Visual Basic which demonstrates the use of binary mode. Conexant’s Serial Interface Guide also provides a comprehensive reference to both modes.

One of the important aspects of any GPS receiver is Time-To-First-Fix (TTFF). This is the amount of time taken to provide a position reference after initially being switched on. A number of factors affect this. Each channel of the GPS receiver has to perform complex signal matching algorithms to lock onto each satellite. The signals are received at very low RF levels, and are actually on the same carrier, so locking onto the codes transmitted from the satellites can be quite a process. For this reason, the receiver will hold an Almanac, which give information as to which satellites should be visible at any one time. Remember that the satellites are not geo-stationary, they are in view for about four hours at any one time. For the almanac to be of use, the receiver must know a reasonably accurate time, so a real-time clock is used. Over and above the almanac information, ephemeris is downloaded from each satellite giving more detailed timing information about that satellite.

Almanacs are downloaded for the whole system from every satellite, while ephemeris information has to be downloaded from the satellite itself.

Using this information, we get cold start TTFF of about 90 seconds for a receiver which does not know time, rough position of have any valid almanac or ephemeris. Once the receiver has operated, and has a real time clock in operation (this is normally battery backed), 45 seconds is achieved. If ephemeris information is valid, then a warm start of around 12 to 20 seconds can be expected.

Newer models of the Jupiter with a co-processor called Magna will reduce these start-up times dramatically. The co-processor is an ARM core and allows very fast carrier lock. TTFF with ephemeris is expected at around two seconds.

Newer still are the SiRF based receivers such as the Leadtek 9543 which offer similar performance to the Magna with WAAS and EGNOS with a smaller size and lower power consumption. These receivers, although quite new are finding rapid acceptance in the GPS marketplace.

Other developments include dead-reckoning (DR) solutions, using a gyro and wheel ticks from a vehicle’s speedometer to provide navigation without satellites for around two miles within the 10 metre accuracy limit set by SA. This is of use to allow position information to be valid while driving in tunnels or car parks where a valid GPS signal may not be available.

The Vehicle Tracking System

The simplest vehicle tracking system possible simply connects the GPS receiver directly to a GSM modem. The GSM modem is set to auto-answer and goes directly into data mode when called. The GPS receiver is in NMEA mode, so that data is pure ASCII, and a simple DCD blocking circuit prevents data from the GPS receiver reaching the modem until an acceptable data call is established.

An extension of this is shown in the diagram. A GPS receiver (Conexant have one called the Jupiter FLASH) has programmability included on the device. This again can be directly connected to the GSM transceiver, but this time, the data can be managed by a program which resides on the GPS receiver. This avoids the need for a separate controller, and now allows the mobile unit to function in either SMS or Data mode. TDC have demonstration code operating in SMS mode, an SMS request for a position is sent, and an SMS is returned giving latitude, longitude, speed, height etc., although of course this is modifiable for any particular application.
 

Other Considerations
Antenna position.

The best position for any antenna is generally as high as possible with the best unrestricted view. The latter is especially true for GPS antennas, but given the format, in that they are normally small ‘lumps’, and are not always easily integrated with an FM antenna, this positioning is not always possible or acceptable. Moreover, in covert applications, an obtrusive lump in the middle of the vehicle’s roof may not be desired!

Today, as the current GPS constellation includes 28 satellites, the coverage is very good, and limited obstruction may be acceptable allowing for installation in other positions on the vehicle, such as parcel shelf, under the surface of the dashboard, or even in a light cluster. In these positions however, the monitoring system should be able to tolerate some loss of service, as if a car is parked against a wall, the obstruction may deny the receiver the chance to see the four satellites required for navigation.

Mapping

At any control station, data is normally required to be viewed on a map. Maps allow a tangible, understandable view of a vehicles location, and will also allow operators to apply local knowledge. For example, I know that 51° 16’N,and 1° 3’W is Stroudley Road in Basingstoke, but if that was applied to a map, it would be obvious to anyone. Maps come in many different flavours, and scales, although a very appropriate street level map of the UK which can be customised for a particular monitoring need is about to be released by Isys Software in the UK. Other interesting mapping sources include MapInfo , who produce a package which can utilise almost any map, by ‘registering’ it with the package. This is an infinitely customisable program, interested parties should certainly contact MapInfo directly for more information. Microsoft have now released a newer version of MapPoint which integrates with the Office suite of packages and allows developers to add functionality to the package using COM components written in Visual Basic as well as integration with JET or Access database engines allowing locations to be overlaid on the map. MapPoint for Europe includes street level map data for all of western Europe, amazingly on one CD.

Development

As ideas of a final product can change at the concept stage, TDC developed a range of ‘Starter Packs’ which provide all the antennas, power supplies, software, and the product itself in a box ready to go. 

These packs are available for both the Jupiter GPS receivers, and the GSM/GPRS modems, and by combining the two starter packs a vehicle tracking system can be quickly built.

Child Tracking

Of topical interest at this time is ‘child tracking’, giving the possibility for children to wear or carry some sort of GPS/GSM tracker. Such a device could provide valuable information if a child is abducted or lost. Signals could be sent from the tracker every few minutes to a central site so that a record is kept of the child’s location, or just sent when a ‘panic button’ was pressed by the child.

Technology has developed to such a stage where these devices are just about feasible now, but key issues with such devices are many :

Battery life, batteries have developed well over the last 10 years, to the point where NiMH and Li-Ion batteries can provide a good battery life. With good power management, a standard mobile telephone battery (1000mAH) should be able to provide 2 to 3 days life for a GPS/GSM system.

System Considerations

GPS Antenna positioning, a critical consideration. The antenna needs to be as un-obscured as possible with a good view of the sky. Placing the GPS antenna on a belt will by implication only provide a 180 degree view of the sky, being obscured by the wearer. The best location would be in a hat, but there would need to be a connecting cable, or for the tracking device to be completely located in the hat (which is easily removed).

GSM antennas are a sensitive subject. Parents have been seen protesting at mobile telephone base stations being placed near schools. These base stations transmit about 8W of power, while a mobile telephone (often given by parents to their children) transmits up to 2W. However, the distance of a mobile mast (typically 15 metres or more), and the close proximity of a mobile handset actually means that the field strength from the mobile telephone is much higher than that of the mobile mast. Will parents be happy about a mobile telephone permanently worn by their child?

Fashion conscious children. If the device is to be worn by a child, one must assume that the child was happy to wear it. Different age groups have different ideas and different acceptabilites of fashion and brands. If the device were designed to fit in a shoe, it may be good to have the support of different brands of trainers to fulfil all tastes!

Removal of tracking unit by ubductee. Assuming that such a device is highly marketed and well known, the first thing a potential attacker would do is to remove the device and dispose of or possibly destroy it. In the former case, it would be useful for the device to automatically send out an alarm and notification. One way could be to have a proximity device which was worn somewhere else on the child. When the two devices were out of range, the tracker would automatically send a location to the control centre, giving at least the last known location of the child.

Central site services

Key to a complete system for child tracking is the central site service. As the GPS unit only knows it’s latitude and longitude, conversion to a map or address is fundamental to the success of the system. It would be envisaged that such a system would transmit its location to the central site, converting the position to an understandable address (or maybe even postcode). Such systems would probaby require a subscription and are almost certainly the financial justification for such a project, not simply the marketing/sales of the hardware itself.

Location Of Device

Two possible locations for such a tracking device spring to mind at this time.

Belt : Weight and bulk are much less of an issue on a belt, and the electronics could be spread along a large part of the length of the belt itself. However, the location of the antennas could cause at least the GPS antenna to be badly obscured.

Shoes : Trainers as commonly worn by children offer a large amount of space to contain the electronics and batteries required. A device could be designed in a modular format which could be moved from shoe to shoe (Parents would not be happy about having to purchase a new tracker unit every time the child grows out of a pair of trainers!). Making a module to fit into a shoe means that the module itself must be very robust, able to withstand the impact of normal use. Dimensionally, it is possible to make these devices to fit, with a battery, but some considerable development and system design will be involved. The GPS antenna can be very small, there are devices which will easily fit in the toe of a shoe, and in reasonable volumes would be very economical, able to be disposed of with the shoe. The GSM antenna is much more straightforward, and could easily be moulded into the shoes sole.

A GPS chipset can have software added to it in order to control the GSM engine, similar to the FLASH Jupiter described above. SiRF chipsets are very small in volume and conceivably, the 

complete system with a GSM engine, GPS chipset, ancillary components and battery could fit into a space of around 35cc, feasible for inclusion in the sole/heel of a shoe.

Other locations such as jackets and hats are possibly not so practical, as they will be easily removed (or lost!) by the child itself.

Other technologies could be employed in a shoe, such as a charging device which employs the pressure depressed on the heel or sole of a shoe to recharge the battery.

Conclusion
All of the concepts described here are very broad in that many readers will have specific variants of these ideas. Readers are invited to contact the author if they wish to discuss specific projects or ideas further.

Comments ?

Have any comments about this review ?  Post them here.