Resources — TechBrief

Choosing a Low-Cost GPS for Road Centerline Survey


GPS logging in a vehicle (experimental, not production)
Until recently, GPS data logging in a vehicle required a laptop and roof-mounted GPS/D-GPS antennae.  This usually necessitated some modification to the vehicle, with — at the very least — the passenger seat accommodating the laptop and other equipment. Today, Personal Digital Assistants (PDAs) or handheld computers are able to store significant data volumes, obviating the laptop arrangement at least for some purposes. Moreover, with the cessation of Selective Availability, it is now possible to log remarkably accurate data from consumer-grade GPS.  A large selection of affordable GPS units is available on the market.  How suitable are these for surveying road centerlines? In the future GPS will be further miniaturized and many high-end vehicles will gather data passively while driving.  But for the moment can we visualize a fleet of vehicles (including police cruisers, mail delivery vehicles, road striping trucks) using inexpensive GPS to log data while running everday duties, and compiling their coordinate logs into centerline maps?

The GIS aspects of processing the data are discussed elsewhere.  This article is about evaluating GPS receivers.  The objective is to identify units that can be used with a minimum of system preparation: portable, low-cost, rugged receivers that work with PDAs rather than laptop computers (the requirement for working with PDAs is not really a burden on the GPS; it is usually a matter of appropriate cabling, but vendors could help by offering the right accessories).  Of course the ultimate test is coordinate accuracy, which in the following article is discussed in 2 dimensions only.

This is not a comprehensive survey of market offerings in GPS hardware.  Many of the products reviewed here will be obsolete in a year or two. Although selected models are critically evaluated, the purpose of this article is not to review them comprehensively, but to illustrate how products might be evaluated and compared for the purpose of centerline survey, and to outline their most relevent capabilities and usability issues.
Note: Vehicles engaged in tests such as these have a high accident rate. We strongly advise against any active operation or even observation of GPS logging units while driving, regardless of driver experience.  A number of safety precautions were observed during these tests, that are not detailed here.

Models tested

The following models were considered appropriate at the time of purchase, May-June 2001.

Magellan GPS Companion for the Handspring Visor

Magellan GPS Companion
and Handspring Visor,
running CetusGPS
The $150 Companion is by far the most convenient, because it is designed to mate with specific PDA models — we used the model for the Handspring Visor Prism.  The combined unit can be suspended from a rear view mirror, using nothing more sophisticated than the wrist strap that accompanies the Magellan MAP330. Ideally it should be placed near the center of the windshield, to see as much sky as possible. Magellan's GPS units are well supported by the manufacturer, with customer-responsive firmware revisions and free, easily downloadable updates; however on the matter of mechanical minutiae, Magellan falls short.  The 12v power supply for the Companion does not feed the 6v Visor, so two separate power cords (and therefore a splitter for the vehicle lighter socket) are required.  The windshield mount (well built, and highly recommended because of its ability to stabilize the unit in position), while custom-contoured to fit the Visor, covers the power/serial port at the bottom of the Visor, and one has to cut out a suitable opening — a drill and cautiously wielded carpet knife are well suited to this ~30 minute project.

The drawback of the Companion arrangement is that, because it completely takes over the serial port of the Visor (although it physically uses the Springboard slot), it is not possible to feed in a real time differential correction signal.  In fairness it could be argued that the applications for which this hardware is designed do not require D-GPS accuracy.

Most GPS software products for the Palm/Visor receive and perhaps display NMEA inputs, but it is difficult to find one that stores the data.  An excellent product is CetusGPS, available free from Kjeld Jensen of Denmark.  CetusGPS compresses the data as it is received, storing essential elements of each NMEA record — time, position and altitude, speed, heading, HDOP — in a mere 28 bytes, permitting 50+ hours of logging in the 8Mb of Visor Prism memory (note that memory expansion is not an option because the Companion uses the Springboard slot).  Palm/Visor buttons are programmed to access the principal menu items, so navigating the menu in a vehicle is quick and easy. Aside from a compass rose, Cetus displays little information useful to a driver, so it does not pose much of a distraction.

Cetus is not specific to any GPS model, hence it does not switch on the GPS receiver — Magellan's software must be invoked to do that, and then Cetus must be launched before the receiver switches itself off.

Unfortunately the hardware combination has significant problems:

  1. Heat appears to be a factor in the repeated failure of equipment, possibly because it is left on while parked.  The temperature inside an unventilated vehicle can exceed 60°C, and metal parts in direct sunlight can reach even higher temperatures.  The Visor Prism's rechargeable battery understandably deteriorates in a matter of weeks under these conditions, and there can be frequent system failures and loss of data. The problem is that Handspring insists that it alone may replace the proprietary Li-ion battery, and while this service is free under the warranty period, it costs $125 out of warranty, and either way it puts the unit out of service for several days. Therefore, for no fault of the Magellan GPS Companion, the viability of this hardware combination for extended work is questionable.
  2. A relatively small problem: the Visor is not a multi-tasking unit.  Any process that interrupts GPS logging (e.g. low battery warning, screen brightness adjustment) requires Cetus to be re-started.

Magellan MAP 330

Magellan MAP330
The MAP330 has two principal advantages over the similar-looking and lower priced ($125) Magellan GPS315.  First, it incorporates a map of principal highways in the U.S. (and Canada, but the maps of Canada are often inaccurate and highways are mis-numbered).  Second, and more relevant to our purposes, it is WAAS-enabled, potentially yielding accuracy in the 3 m range.

The unit has 4 contact points on its back, which serve as (1) power supply, (2) NMEA output, (3) RTCM differential input.  A single data/power connector covers all four contacts.  Again the gap between good product design and poor support becomes apparent.  Magellan sells a connector cord that supplies power from a vehicle, and downloads data through a serial plug. But it is not possible to input the differential correction and download NMEA data at the same time.  It's not clear whether this is just a matter of the right cord not being made available (a bare-wire connector is), or a system design deficiency.  The power/data connector snaps neatly into the back of the windshield mount, so the GPS slips in and out easily without any explicit connector handling.  However, after a couple of hot days in the vehicle, the spring clip (see red arrow in photograph) is no longer able to hold the receiver in the mount, and it takes a bit of Velcro to secure the GPS in place thereafter — an unlikely juxtaposition of space age technologies.  The MAP330 itself is rated for operation up to 60°C, and our unit has shown no sensitivity to temperature.

For our tests we fed NMEA output from the MAP330 into a laptop computer, and logged the data using Hyperterminal.  Logging could probably have been achieved on the Visor, with special cabling.

The MAP330, like many other receivers in its class, can store a limited amount of tracking data, and Magellan offers MapSend software that transfers the data from the receiver to a computer.  But this is so poorly implemented that it has no practical value.  MapSend has little functionality of its own, yet it stores the data only in its proprietary format, making it unavailable for further use. Although the MAP330 is designed for recreational hikers rather than road surveyors, 3 simple changes would make it the data logger to beat:

Pharos iGPS-180

Pharos iGPS

The iGPS-180 generally has a good usable design.  The antenna and processor are combined into one light, small-footprint unit which can be placed on a dashboard and tethered to the windshield using the suction cup provided (a magnetic mount for exterior use would have been useful). The cable from the unit splits into two plugs, one DB-9 serial for data, the other a PS/2 keyboard plug that ingeniously draws power from the computer's external keyboard input port.  The Visor-specific cable, an extra, matches the Visor serial port and uses a cigarette-lighter plug for power.  Incredibly, the Visor cable supplies power only to the GPS, not the Visor, despite taking over the Visor power/serial port. This is a serious impediment to extended use, bearing in mind that Visor power consumption is much heavier than normal when the serial port is in operation.  The iGPS-Visor combination cannot match the convenient 1-piece integration of the Magellan GPS Companion, but it does allow the antenna-processor to be positioned independent of the Visor.

As in the case of the MAP330, NMEA data from the Pharos were logged using Hyperterminal on a laptop computer rather than the Visor.

A drawback of the minimalist design of the iGPS-180 is its lack of a display or any status indicators. Our iGPS-180 failed after a few months of use, perhaps due to heat in the vehicle (although it is rated to 80°C).  It now outputs NMEA fields with null data.

Trimble Placer GPS 400 (reference system)

Trimble Placer 400 antenna
CSI MBX3 beacon receiver
Our reference system is the Trimble Placer 400, a relatively old (c. 1995) but high quality unit designed for navigation.  D-GPS correction is supplied by a CSI MBX3 dedicated Coast Guard beacon receiver.  Output from the GPS unit is in Trimble's TAIP, not NMEA.  VITAL's custom software parses and logs this output. This unit certainly does not represent the state of the art in GPS, but it fills the role required of it in these tests.

Like the other GPS receivers described above, the Placer records GPS data at approximately 1 Hz.

GPS coordinate quality

4-lane test road
The receivers described here incorporate a number of practical consumer/application-oriented features, most notably: (a) they may extrapolate vehicle path when satellite signals are temporarily unavailable, and report a loss of signal only when severe; (b) when stationary, they report an average reading rather than the actual reading.  This makes it difficult or impossible to observe consistency of readings over a period of time in a stationary test.  To test repeatability, we drove the units together 20 times down each of 4 lanes of a straight section of road, late on a Sunday night (to minimize interference to and from traffic).  The Companion data were logged by the Visor; the other GPS receivers were connected to a laptop. The sequence in which lanes were driven was recorded.  The tracks are color-coded in the illustration below: blue and red are GPS readings in two southbound lanes; green and black are the two northbound lanes. The spread between tracks of a given color is a simple measure of repeatability.




Repeatability of four GPS units.  Differentially corrected Trimble (d) clearly shows the best consistency of readings.  Magellan's GPS Companion (a) appears to be the best of the low-end units, generally beating the WAAS-enabled MAP330 (b).

Predictably, the differentially corrected Trimble reference unit (d) produces the tightest readings.  The GPS Companion (a) is a surprising second, edging out the WAAS-enabled MAP330 (b).  The Pharos (c) is significantly inferior to the other models.  These results are consistent with stationary tests, where the Trimble showed the least drift, and the Pharos the most. Since we do not know whether or how readings are averaged internally by the receivers, the stationary tests by themselves are not good indicators of suitability for road survey.

What the results do not say

There is a danger of over-interpreting the above results.  GPS error varies gradually over time.  The illustrations above, showing dramatic lane separation with most test units, are based on data observed over 2 hours.  Lanes are not as easily distinguishable in data gathered over longer periods.  We are developing algorithms that “average” multiple tracks, on straight and curved roads.  Averaging pre-supposes that tracks are normally distributed about the true lane centerline, i.e. that low error is most frequent.  Surprisingly, preliminary results bear this out, though it could be partly a pattern of driving accuracy rather than GPS accuracy.

How much repeatability/accuracy is enough?  This is application-dependent.  For routine navigation, GPS position and centerline geometry can be within 20-25 m and be acceptable.  For emergency applications, where one needs to know whether an incident is in the north-bound or south-bound carriageway of a divided highway, 5-10 m accuracy is more appropriate.  For highway maintenance, assets such as guard rails are recorded with 10-15 m accuracy.

Highway operators typically reference locations in terms of linear measures. For coordinates to be compatible with linear referencing, centerline coordinates must be linearly consistent. Analysis of GPS tracks for this purpose is the subject of other UCSB research.  On most roads, linear measures derived from centerline coordinates from GPS units above are within 0.25% of DMI measurements, which is close to the limit of DMI calibration accuracy.


D-GPS Differential GPS.  In the U.S., differential signals are broadcast free by the Coast Guard in coastal areas, and a Nationwide Differential GPS broadcast is under development.
DMI Distance Measuring Instrument.  Odometer with ~1 m resolution, employed in highway operations.
HDOP Horizontal Dilution Of Precision.  Dilution of Precision is a function of satellite configuration, and is a rough indicator of the quality of GPS reading — a value of <2.0 is good.  When satellites are temporarily obscured, the statistic measures the effectiveness of the remaining configuration, which is usually not as good, so DOP is roughly correlated (inversely) with the number of visible satellites.  Generally speaking, DOP is directly proportional to error, but there are situations where a configuration may yield a poor statistic but good coordinate accuracy, and less frequently vice versa.
NMEA National Marine Electronics AssociationNMEA 0183 version 1.x and 2.x are standard formats by which GPS receivers communicate position and satellite configuration data.
PDA Personal Digital Assistant: handheld computer such as the Palm and Handspring Visor. Compaq/HP and other vendors offer equivalent models.
RTCM Radio Technical Commission for Maritime Services.  RTCM SC-104 is the standard format for feeding differential correction data to GPS units.
TAIP Trimble ASCII Interface Protocol
VITAL Vehicle Intelligence and Transportation Analysis Laboratory, UCSB
WAAS Wide Area Augmentation System, a Federal Aviation Administration sponsored GPS enhancement, originally intended to assist aircraft navigation in poor visibility.  WAAS-enabled equipment should be capable of about 3 m accuracy.  But signal reception is a problem.  Only two INMARSAT satellites currently transmit WAAS signals over North America, hence reception is spotty.  The NMEA format version 2.x (implemented in the above receivers) does not indicate whether WAAS is being received and processed.

All trademarks are the property of their respective owners.