GPSi Dead Reckoning White Paper: Product Overview

Transcription

1 GPSi Dead Reckoning White Paper: Product Overview A Neve Technologies White Paper February 2001

2 GPSi DEAD RECKONING Vehicle Positioning in the Urban Environment Chris Wood and Owen Mace, Neve Technologies Pty Ltd. INTRODUCTION Policeman Hall and his partner are on patrol in a dangerous part of town when they receive a call for help. When the officers arrive, all seems quiet and so they enter the apartment building to investigate. On the way they are attacked and disabled. Luckily, Officer Hall pressed his duress alarm to call for help. Hall is confident that help will soon arrive as his equipment functioned as he expects. Unfortunately for Officer Hall, the position transmitted from the GPS receiver in his radio is in error by a couple of hundred meters; the receiver has been unable to receive clear signals in the street or in the apartment building. The dispatch center is doing its best, but he wasn t at the location sent in the transmission... In this example, and in many others where positions are needed in urban environments, inaccurate positions from GPS receivers limit the usefulness of GPS as a positioning device. The reasons for this are well known: Reflections of GPS signals from buildings (multipath); Poor signal propagation in cities, under dense vegetation or in deep valleys (shadowing); No signal in tunnels, undercover car parking, covered roadways; Extended time to first fix (TTFF) following a signal loss, especially in areas of poor signal quality; and Dynamic limitations, such as maximum jerk, and other receiver limitations. For someone relying on GPS for position, there are two outcomes: the GPS receiver may not provide any position (or it just repeats what it thought was the last good solution) or the position accuracy degrades to much worse than the advertised accuracy, perhaps even greater than 500 m. This paper aims to give an overview of the shortcomings of many GPS receivers using the standard positioning service and to show how a position augmentation system can correct and compensate for these. BACKGROUND The authors have used many Automatic Vehicle Location (AVL) receivers in their home town of Adelaide, South Australia (approximately 138º East and 35º South) which is not a difficult city for GPS. Adelaide has some urban canyons in the Central Business District. They have also tested systems in Hong Kong (approximately 114º East and 22º North) which is one of the worst places in the world for GPS reception. Adelaide, the capital of the state of South Australia, has a population of about one million people and, in the Australian way, sprawls over hundreds of square kilometers. In the Central Business District (CBD) there are many five to ten storey commercial buildings and a small number of taller buildings, up to 30 stories, see the photograph. The streets typically have four lanes and there are many narrow lanes between the buildings. The CBD is surrounded by North, South, East and West terraces and parklands. It is a very pleasant town. Even though Adelaide is not a big city, journeys into the CBD can result in unreliable and inaccurate positions, as we see in Figure 1. Notice how, as the vehicle drives down Grenfell Street, the receiver reports positions that 2

3 are hundreds of meters in error. Next, the vehicle drove into Wyatt Street, a narrow lane between multi-storey buildings. To test the receiver, it stopped for a minute in Wyatt Street. During this period, the positions reported by the 12-channel receiver jumped about randomly. Figure 1. Urban Canyon Journey What is going on here? The GPS manufacturer claims that their receiver is accurate to 25 m and yet positions up to 500 m in error were reported. In this case, the accuracy was degraded by a condition known as multipath. Multipath is a condition where signals from the GPS satellites are reflected off large objects (in this case the surrounding buildings) causing the receiver to make incorrect pseudorange measurements. If this can happen in Adelaide, imagine the sort of results that could be expected in the "urban canyons" of Hong Kong, New York or Tokyo where the satellites are shadowed by buildings for most of the time. OTHER GPS LIMITATIONS Manufacturers of GPS receivers for vehicles recognize that signals in cities can be badly degraded and so they incorporate sophisticated algorithms that attempt to continue navigating even when the number of satellites falls short of the minimum required for accurate navigation. For example, they may assume that the altitude remains constant or that the vehicle keeps traveling in the same direction. Sometimes these assumptions are reasonable, but often they are not - and the position solution suffers. The receivers themselves are not perfect; look at Figure 2 where the vehicle has driven rapidly 360º around a roundabout. Notice how the GPS receiver has totally lost the plot (Literally!). What has happened? The answer is that the receiver's maximum jerk specification has been exceeded, which, for this receiver is 2 ms -3. Jerk is the rate of change of acceleration. Although not shown, the receiver recovered after some time. 3

4 GPSi solution continues GPS receiver has lost signal lock here Figure 2. Dynamic Limits Exceeded In the previous figures where position solutions were poor, the receivers had been operating in non-differential mode. Many people think that by using DGPS (Differential GPS), the situation would be improved. In many cases this is true, DGPS typically improves accuracy to about 5 m for most AVL receivers, depending on various factors. However, the authors have seen examples in CBDs where differential GPS solutions are actually worse than non-differential GPS! ERROR ESTIMATES GPS receivers don't often fail, but like any system, they have their limitations. The informed user understands those limitations and makes allowances for them. However, many receivers do not provide the user with adequate information about the accuracy of the position solution so that the user can make an informed judgment. An estimate of the quality of the position derived by the GPS receiver is needed. Aviation users recognize the need for measures of the integrity of their positions and aviation GPS receivers incorporate Receiver Augmentation and Integrity Monitoring (RAIM) software. Some may say that the various Dilutions of Precision (DOP), such as Horizontal Dilution of Precision (HDOP), are such quality measures, but they are not. HDOP is an indication of how much the geometric arrangement of the satellites in view affect the quality of the solution. DOPs account for shadowing but they do not account for multipath signals. What is needed is an estimate of the error of the position solution that the receiver calculates. One way for a receiver to estimate error is for it to calculate the root mean square of the differences between its measured pseudoranges and the distance from its position solution to the GPS satellites. Which ever way a receiver calculates its estimated position error, at least the user has an idea of the quality of the position solution that the receiver is providing. Different receivers calculate different measures of the quality of their position solution, such as Estimated Horizontal Position Error (EHPE) and Horizontal Figure of Merit (HFOM). But remember, none of the DOPs truly give an indication of quality of the position solution. Those receivers that do go to the trouble of providing a position quality estimate cannot deliver it if they use the industry standard NMEA 0183 message interface. To get position quality information, users must delve into the manufacturer s proprietary 4

5 binary message interface, which is often complex to decode and provides more information than required by most applications. AUGMENTING GPS When an application requires a position even when GPS is not accurate enough or not available, there is no choice but to augment the GPS receiver. There are several ways of doing this: in-car navigation systems (electronic street directories) use an electronic map stored on CD-ROM. Sensors in the vehicle inform the system of a rough position, the distance traveled and turns taken, from which a very accurate vehicle position can be determined by matching the data to geographical street information stored in electronic maps. Such systems direct drivers around cities and can navigate a vehicle to an individual street address. Instructions are displayed to the driver and spoken out loud. This type of car navigation system generally works very well and is in widespread use, particularly in Japan, Europe and the USA. The major drawback however is its price, typically in the vicinity of US $1,500. Another way of augmenting GPS is with a dead reckoning system. DEAD RECKONING Dead Reckoning (DR) is the system that mariners use to navigate. Their compass measures direction and the ship's log measures speed (or distance traveled). By combining the two measurements, the vessel's track from its starting point can be determined. The important point is that the dead reckoning positioning is relative to a known starting point. The starting point could be based on a well known location such as the tower at the Greenwich Observatory, or more conveniently, measured using a GPS receiver. A vehicle can use the same system, provided it has a means of determining speed and direction. Speed is easy. Modern vehicles have an electronic speed sensor that accurately measures speed and the distance traveled. But an accurate, low cost measure of direction is not so easy. DISTANCE AND DIRECTIONS SENSORS The odometer in a modern automobile measures rotations of an axle or drive shaft that is directly related to distance traveled. Unfortunately, the relationship changes with tyre wear, temperature of the tyre and wheel slip. There is even a change when the vehicle is driven up a slope. The result is that there can be as much as a ten percent change in the distance traveled per wheel revolution. An electronic magnetic compass might be used to measure the vehicle s direction from the Earth's magnetic field but this scheme suffers from a number of major shortcomings. The Earth's magnetic field changes from place to place and it changes slowly with time, so that compasses must be continually calibrated. Furthermore, the earth's magnetic field is altered by the currents flowing in vehicles, such as the car's air conditioner or hi-fi system as well as by external influences such as bridges, tunnels or buildings. Also, changes in the slope, or tilt, of the car can change the measured direction markedly. In Adelaide, a tilt of the car of one degree results in an apparent change of direction of up to three degrees. Using the Earth's magnetic field, despite its apparent attraction, is a poor way to measure direction. Aviators have solved the problem of measuring direction with gyros. (Remember how a spinning top tends to maintain its initial direction - upright - despite gravity trying to make it fall?) High quality gyros allow aircraft to navigate within a few kilometers after a flight across an ocean. Unfortunately, high quality gyros come at a high quality price. Nevertheless, solid state gyros are available for a cost comparable with an AVL receiver. However, with low cost comes relatively low performance and so designers of navigation systems using solid state gyros must take exceptional care to understand and account for the limitations of their gyros. (Recognizing the need for better performing, low cost gyros, many companies and departments of defense are trying to improve low cost gyros.) Neve Technologies has developed a DR system for vehicles, called GPSi - the "i" is for inertial. The system uses the vehicle's odometer to measure speed and its reverse light to indicate when it is moving backwards. A 5

6 miniature vibrating beam or tuning fork gyroscope is used to measure rate of turn. These gyroscopes operate from a 5 V DC supply and typically have a sensitivity of approximately 22 millivolts per degree/sec. Zero degrees per second (called the gyro bias voltage, or just bias) is approximately 2.5 V. To determine the vehicle s heading, the gyro bias must be subtracted from the gyro output and the result integrated to yield the change in direction relative to a known initial heading. This process sounds straightforward enough, especially since modern microprocessors make the calculations easy to perform, but there are numerous difficult problems to overcome. For example, the gyro bias voltage must be measured very accurately to ensure that bias errors do not translate into large navigation errors due to the mathematical integration process. To make things worse, the gyro bias changes with ambient temperature and is affected by electronic and mechanically induced noise. Also, as mentioned previously, the vehicle odometer is susceptible to various sources of error that must be take these into account. If a DR system is to be useful in the most critical time of need (i.e. when GPS is not providing a solution), then the various sensor errors must be managed very carefully. SOURCES OF ERROR In fact, all errors affecting the position estimate must be estimated and accounted for. Low cost GPS receivers used for vehicle navigation exhibit random errors in the order of 25 m. The predominant sources of error are in the ephemeris, which manifests itself as uncertainty in the satellite position; as aberrations in the delay due to variations of electron content in the ionosphere and, in some receivers, as phase noise in the local oscillator. Frequently, positioning errors as low as 2 m are reported, though seldom without the necessary additional information describing the state of the ionosphere etc. And of course, errors can be considerably larger, as we have already seen. Many GPS receivers fail to provide meaningful estimates of position error and consequently GPSi estimates position errors for those receivers. This is done using proprietary algorithms which give very good results under most conditions. However, there are limitations to this approach and GPSi performs better with GPS receivers that provide their own position error estimates. To compensate for the sources of odometer error discussed previously, GPSi continually calibrates the odometer. Calibration is achieved by generating an error signal between the distance measured by the odometer and the distance estimated by the GPS receiver when confidence in the GPS solution is high.. Using this approach, odometer calibration can be maintained at better than 1% of distance traveled. As for the gyro, first order errors manifest themselves as inaccurate estimates of gyro bias. Consequently, every measurement of gyro output includes a small error in angular rate. This error is then integrated along with the true signal. To understand the effect this has on the heading calculation, consider a system where the gyro output is measured with a 10-bit analog to digital converter (ADC). Let s assume that the gyro bias voltage is free of noise. Then in this case, the measurement error is equal to the quantization error and is up to half a bit or about 2.5 mv, which equates to about 1/10 of a degree/second for the gyro used in GPSi. Now let this error integrate for 5 minutes which is not an unusual time to loose GPS signals on the roads of Hong Kong. At the end of 5 minutes, the bias error in the gyro output will have produced an erroneous estimate in direction of 0.1 deg/sec 300 seconds = 30 degrees, which is quite unacceptable. A simple mathematical analysis shows that the position error due to gyro bias increases as the square of the distance traveled with an uncorrected gyro bias. Of course, the gyro output voltage does contain noise which includes noise from its internal electronics and noise induced by mechanical vibration (engine and road surface). This noise effectively dithers the ADC quantization error which has the effect of reducing the overall measurement error. So in the previous example, 30 degrees of heading error is actually quite an exaggeration. Second order gyro errors manifest themselves as nonlinearities and other perturbations in the scale factor. Compared to the bias errors, these are typically very small and are not significant for most AVL applications. This of course assumes the gyro scale factor is accurately measured via a calibration process. There is one further source of error worth mentioning that can enter the gyro system and it is a function of the type of terrain the vehicle travels over. Gyros have an axis of rotational sensitivity. This axis should be perpendicular to the coordinate plane the position solution is referenced to. If the axis is not perpendicular, the scale factor sensitivity must be reduced by the cosine of the angle of incline. For an incline of 10 degrees (which is a very steep slope for a vehicle) the effect will add about 1.5% of error. 6

7 GPS-INERTIAL DEAD RECKONING Through careful signal processing and sensor management, reliable measurements of vehicle heading and distance over ground are derived. Using these measurements and standard navigation equations, an inertial position solution is calculated; this is the i part of the GPSi. However, the story does not end here. No matter how expensive the sensors are or how good the signal processing algorithms are, the residual errors that have not been nulled out, will accumulate and eventually render the inertial solution useless. This is where the GPS part of the GPSi comes in to play. Figure 3 shows the (simplified) strategy employed for blending the two solutions. This process is used to correct the inertial position and heading estimates as well as the gyro bias estimate. GPS Inertial Subsystem True + GPS errors - True + INS errors Σ Quality Navigation Filter Figure 3. Blending GPS and Inertial solutions. Σ ε = INS error estimate GPSi Solution The Inertial Subsystem is comprised of the gyro, vehicle odometer, the software that manages these and the navigation equations that transform the heading and distance over ground to a position solution expressed in the same coordinate system used by the GPS receiver (in most cases these are geodetic coordinates referenced to the WGS84 datum). The GPS box is the GPS receiver that provides position solutions and vehicle heading at regular intervals. Some GPS receivers may also output solution quality information often called the Figure of Merit (FOM) that is a measure of how good the receiver thinks its solution is. Some receivers produce FOMs for horizontal and vertical positions, horizontal velocity and even time. If the receiver does not provide this (and many don t) GPSi estimates it as previously mentioned. The first step in the blending process is to create an error signal which is the difference between the GPS variables and the inertial variables. In the ideal case this difference would be zero because the inertial solution would perfectly track the GPS solution. However, there are may reasons why the error is non-zero and, in fact, it will always diverge over time. Also, it is worth noting that the sources of error in both the systems display quite different properties. GPS errors are absolute and are less than 22.5 m for 95% of the time (Standard Positioning Service without selective availability). In contrast, inertial errors are cumulative and increase without bound at a rate determined by the quality of the sensors and the signal processing algorithms. The error signal (which is the GPS and inertial errors combined) and the quality factor (FOM) are then passed into the navigation filter. In case of GPSi, the navigation filter is a Kalman filter, which is a statistically optimal digital filter. The job of the navigation filter is to estimate the value of the inertial error variable from the combined error input signal. The resulting inertial error estimate ε is then subtracted from the inertial solution to produce the GPSi corrected position solution. The FOM is used by the navigation filter to determine an estimate for the GPS measurement error. The reader will have observed that the GPSi output is actually the corrected inertial solution. This is significant because almost all low cost AVL receivers can supply a position solution at a maximum rate of only 1 Hz. However, the inertial solution can be calculated much faster and with a typical low cost 16-bit microprocessor, GPSi can actually output solutions at 10 Hz (using the NMEA 0183 RMC message and Baud rate of 19.2 Kbps) and even faster with more powerful devices. This can be very useful in some applications. - 7

8 AN IMPLEMENTATION OF GPSI Integrating inertial and GPS solutions is a well understood process and has been done successfully for many years. However, producing reliable results using low cost gyroscopes and GPS receivers that do not supply a solution quality factor is still somewhat of a challenge. The design goals of the GPSi project have been to deliver a dead reckoning technology with a small firmware footprint, excellent performance over a wide range of operating conditions and is in a price bracket that make it a feasible addition to all AVL GPS receivers at the point of manufacture. We estimate that in large quantities (> 100k), the hardware and software cost of adding GPSi technology to a GPS chipset should add no more than $4 (excluding the gyroscope and temperature sensors). During the development of GPSi, particular attention was paid to minimizing the processing load on the CPU. The most compute intensive processes are the navigation filter, coordinate translation algorithms, statistical estimators and the GPS FOM estimator. Novel solutions to these problems were engineered and as a consequence, GPSi is able to run comfortably on a commonly available 16-bit microcontroller (without a floating point unit) operating at 10 MHz. As explained earlier, accurately estimating the gyro bias voltage is critical. There are many factors that affect the bias, the most important of which are: electronic noise, mechanically induced noise, temperature, power supply fluctuations and ADC resolution. A dynamic 2-stage bias estimator was developed that is able to produce excellent results both whilst the vehicle is stationary and when it is moving. The estimator is able to cope with rapid temperature fluctuations that may occur in some parts of the world (such as Adelaide where it is not uncommon for a morning starting temperature of 15 C to rise over 45 C within the space of a short journey over which the vehicle may seldom stop). AVL receivers are notorious for producing wildly unreliable heading measurements in urban environments. Apart from altitude, heading measurements are probably the most inaccurate data provided and the most susceptible to urban canyon effects. Much effort was expended in the development of algorithms able to discriminate between reliable and unreliable heading measurements. This information is imperative in determining when the heading solution can be reliably updated. GPSi takes a pragmatic approach to the effects of gyro scale factor sensitivity to angle of incline. Since the vast majority of urban roads have a gradient of less than 10 degrees (1:6), the standard software does not compensate for these effects as the induced error is small. However, the hardware provides an extra input to the ADC that can be used for an inclinometer, accelerometer or other application specific sensor and the capability can be easily added to the software. GPSi-M12 is one physical incarnation of GPSi technology. It is a PCB that is the same size and format as a popular commercial AVL GPS receiver board. Its design concept allows it to be installed to existing OEM equipment by removing the GPS receiver, replacing it with the GPSi-M12 board and then plugging the GPS receiver into the GPSi-M12 board (see Figure 4). The gyro, temperature sensor, odometer and reverse indicator are connected either directly to the GPSi-M12 board through a 90º connector (not shown), or for new designs via a straight through connector to the OEM board. The GPSi-M12 board is in-circuit programmable by the OEM equipment so that new or customized versions of the firmware can be downloaded whilst units are in the field. The firmware code size is approximately 38 Kbytes; a size that can easily be transmitted wireless to field units. Figure 5 shows a hardware block diagram of the GPSi-M12 board. 8

9 A GPSi-M12 board and a quarter coin for scale. The board has been designed to be exactly the same size as the popular M12-Oncore receiver (40 x 60 mm, 1.57 x 2.36 inches). The CPU section consists of a 16-bit microcontroller which provides 3 serial ports, 128 Kbytes of flash memory, 4 Kbytes of SRAM, timers and various other peripherals. The microcontroller is augmented with an external 128 Kbytes SRAM chip. The analog section consists of antialiasing filters for the gyro, temperature sensor and inclinometer signals, a high precision ADC, opto couplers to interface with the vehicle reverse and speed sensors and power conditioning. The connector section shows the 10-pin OEM connector on the right and the DR connector on the left. This view shows a GPSi-M12 mated with its M12- Oncore GPS receiver. The M12-Oncore s OEM connector is brought through by the GPSi-M12. The GPSi-M12 is available with a DR connector option that brings these signals out via a 90º connector. This allows it to mate with existing OEM equipment without modification. Figure 4. An implementation of GPSi. Gyro, temperature and tilt sensors Antialiasing filters Analog to digital converter Speed & reverse sensors Optical isolation Microcontroller and Memory GPS receiver Analog power GPSi Firmware OEM equipment Power conditioning Digital power Program mode Figure 5. GPSi-M12 hardware block diagram. Future GPSi implementations will integrate other receivers in the same way. In large scale applications, we expect GPSi will be integrated into the GPS receiver firmware or other in-vehicle electronics. 9

10 GPSi RESULTS To test the navigation system, GPSi was compared with plain GPS for various types of journeys. Results have been collected over a period of about 18 months at different times of the day (and night) and under varying weather conditions. Different gyro and GPS receiver combinations were used. Some of the more interesting journeys are presented here. Some of these results were taken while Selective Availability was present and all are without the assistance of differential GPS. The equipment used to record these tracks included a number of 12 channel AVL receivers from different manufacturers each augmented with GPSi dead reckoning. The GPSi software was configured to log the GPS and GPSi solutions simultaneously. 10

11 Suburban Journey This route is typical of a short suburban journey. The environment presents no particular difficulty for a GPS receiver. Most of the area is residential single storey dwellings where the only obstructions to GPS signals are the canopy of leaves from the Ash trees on First Avenue. The map scale is 1.3 km horizontally by 1.3 km vertically. As expected, the results are very good for this journey. The GPSi track (Figure 6-A) plots right over the streets. Overlaying the GPS and GPSi solutions (Figure 6-B) reveals how well the navigation filter is doing its job. GPSi is designed to allow the navigation filter to run at a different rate from the GPS measurement rate. This provides some interesting opportunities. For one, if a very low cost solution is desired, the navigation filter may be run only Figure 6-A. GPSi Track occasionally to save CPU bandwidth. Using this approach GPSi can be run on a two dollar microprocessor and still keep the inertial errors within specified limits (say 20 m). If on the other hand, the error budget is tighter, the navigation filter can be run faster (requiring more horsepower from the microprocessor) and keep the solutions tracking within a tighter range. Figure 6-B. GPS and GPSi Tracks Overlaid In the system used to generate these data, the navigation filter is only applying corrections every 10 seconds. 11

12 Urban Canyon Journey This route takes the vehicle down some very narrow lanes flanked by tall buildings on either side. The lanes are not marked on the map. Typically there are zero to one satellites visible in the lanes and three or four visible elsewhere. The map scale is 1.4 km horizontally by 1.1 km vertically. The GPSi solution (Figure 7-A) accurately reflects the true path of the vehicle even in the streets where no satellites are visible. Notice that the GPSi solution does not wander off the roadways into the buildings on either side. The GPSi solution exhibits sharp right angle turns and straight segments without jitter and wander. Figure 7-A. GPSi Track In contrast, the GPS solution (Figure 7-B) does not accurately reflect the true path taken by the vehicle. The segments through the streets are completely missing which is to be expected, since no satellites are visible. Notice how the path wanders and jitters as the GPS signals are affected by the tall buildings. Some segments show the vehicle off the road and traveling through the city buildings. Figure 8-B. GPS Track 12

13 Hills & Tunnel Journey The map scale is 2.2 km horizontally by 4.7 km vertically. Adelaide is flanked to the west by the Gulf of St. Vincent and to the east by the Mt. Lofty Ranges. This journey takes the vehicle through a steep ascent cut into the foothills and through the newly completed Crafers tunnel. The cuttings leading up to the tunnel have steep sides that are 100 to 150 m high. The tunnel is cut through the rock and is approximately 750 m in length, and ascends for its full length. Figure 9-A. GPSi Track. In the tunnel there is zero satellite visibility. Notice how the GPSi solution (Figure 9-A) has continued to report position in the tunnel and has kept the vehicle on the road. Also note that there are no abrupt discontinuities as signal was lost on entry and re-acquired on exit. In contrast, the GPS solution (Figure 9-B) begins to repeat the same position after the signal is lost on entry to the tunnel. After the vehicle has exited, the GPS receiver takes some time to re-acquire and produce a solution. Notice how initially there is a large amount of error and it takes the receiver some time to get its solution back to within normal error limits. Figure 9-B. GPS Track. 13

14 Undercover Car Park For some applications (e.g. road user charging), accurately measuring distance overground is of prime importance. Calculating distance over ground from GPS measurements does not produce reliable results. There are a number of reasons for this: The GPS measurement rate (for AVL typically 1 to 10 sec) is such that significant loss of information can occur. For instance, if the GPS receiver was being sampled at 10 second intervals, and the vehicle drove 360º around a roundabout in that time interval, the system would calculate that zero distance had been covered; In areas of GPS signal degradation (e.g. CBDs or undercover areas) position measurement errors in the order of hundreds of meters can be reported. Calculating the distance between such points gives silly results. Figure 10 shows an every day occurrence for some of us; a trip to the shopping mall. No trip to the mall is complete without driving around and around searching for the best (undercover) parking space. In this case we tried for an undercover space but without luck. In the end we settled for an outside space. The interesting thing to note is the behavior of the GPS receiver while the vehicle was under cover. The GPSi solution depicts what the vehicle actually did. The GPS solution continues for a brief time undercover and then begins to produce some pretty strange measurements. If we were to calculate distance over-ground from these measurements we would be in error by hundreds of percent. In contrast, GPSi calculates distance over ground from the vehicle speed sensor which it continually re-calibrates from GPS so that accuracy is consistent over time. Figure 10. The effect of Signal Obstruction 14

15 CONCLUSIONS The positions provided by AVL GPS receivers have limitations due to the disruption of clear signals in an urban environment. The informed GPS user understands the limitations of their AVL receiver and demands either: A receiver that reports the quality of its solution; or If positions are required continuously even when the GPS signals are compromised, a second system that augments the AVL receiver with another navigation system. But whatever the application, users should insist that their GPS receiver reports the quality of the position solutions that it provides. Only by knowing the quality of the solution, can the results be interpreted intelligently. That is exactly why aviation users have RAIM when they rely on GPS for navigating their aircraft. An excellent means of augmenting GPS receivers for automobiles is with a dead reckoning system based on inertial sensors such as the one described here. Dead reckoning opens up a whole new world of applications not previously possible for low cost AVL receivers. If an application requires continuously available position, heading and speed, accurate distance over ground, an estimate of solution quality and immunity to common GPS problems such as shadowing and multipath, then GPS augmented with dead reckoning is an excellent solution. BIOGRAPHIES Chris Wood (chris.wood@neve.com.au) is a senior project engineer with Neve Technologies Pty Limited working on the GPSi project. Previously he was the project manager in charge of development of the software component of the Electronic Road Pricing (VPS) trials held by the Hong Kong Government. He has an honors degree in Electrical and Computer Systems Engineering from Monash University, Melbourne Australia. Dr. Owen Mace (owen.mace@neve.com.au) is chief scientist with Neve Technologies Pty Limited. He holds a Bachelor of Electrical Engineering (Electronics) and a Doctorate of Philosophy in physics from the University of Melbourne. After many years as a Senior Lecturer, he left academia to join industry. He has a passion for GPS technology despite of his ability to navigate at sea by dead reckoning. 15

16 GPSi Dead Reckoning White Paper Product Overview February 2001 This document is provided for informational purposes only and the information herein is subject to change without notice. Please report any error herein to Neve Technologies Pty Limited. Neve does not provide any warranties covering and specifically disclaims any liability in connection with this document. GPSi is a trademark of Neve Technologies Pty Limited. All other company and product names are used for identification purposes only and may be trademarks of their respective owners. Neve Technologies Pty Limited 83 First Avenue St. Peters, SA 5069 Australia Worldwide Inquires: Voice Fax info@neve.com.au Copyright Neve Technologies Pty Limited 2000, 2001 All Rights Reserved. Printed in Australia 16