VEHICLE INTEGRATED NAVIGATION SYSTEM Ian Humphery, Fibersense Technology Corporation Christopher Reynolds, Fibersense Technology Corporation
Biographies Ian P. Humphrey, Director of GPSI Engineering, Fibersense Technology, formally the project engineer on the ATFLIR IMU Development program as well as the manager of the electronic, systems and test groups at Fibersense. Mr. Humphrey has 20 years experience in the analysis, development and testing of inertial systems for both stand alone and aided navigation systems. Mr. Humphrey received a BS in Systems Engineering and an MS in Control Engineering from Teesside and Sunderland Polytechnics in England Christopher I. Reynolds is currently responsible for marketing and business development activities at Fibersense. He works closely with Fibersense's customers on their need for improved navigation within cities. Mr. Reynolds supported the founding of Fibersense helped guide the development of the companies IMU and gyro products for land vehicles, aircraft, missiles and space. Prior to Fibersense's spin-off from Northrop, Mr. Reynolds was the Program Manager during the development of IMUs for the Phoenix and AMRAAM Missiles. Previously he was a Project Engineer on a number of IMU development programs for spacecraft including the Apollo Lunar Lander, Viking Mars Lander, Space Shuttle, Titan and various satellites. Mr. Reynolds received a BSEE from the University of Salford, England. Abstract Fibersense Technology Corporation has developed an integrated navigation system for accurate navigation in urban environments. The first application of this system will be to improve the navigation accuracy of the New York City buses. The environment of New York City is particularly severe, and is probably the most challenging in the U.S.A. for GPS blockage and multi path. As part of the Americans with disabilities act the NYC buses must announce the location of the bus stop. Bus status is available at each bus stop (e.g. the next M2 bus will arrive in 6 minutes). During periods of good GPS data the system calibrates it’s sensors. This allows accurate navigation when the GPS data is unavailable or when the data is badly corrupted by multi-path signals. This also operates better
in situations when a small number of satellites are visible. This functionality can be added with very little additional cost compared with a system using only a GPS receiver. Introduction Use of GPS in cities such as New York is unreliable as a result of signal blockage and reflections from buildings. This environment has been named "Urban Canyon." For many consumer applications reasonable navigation data with breaks of up to 50% of the time may be considered acceptable. Other applications cannot tolerate GPS inaccuracy in the urban canyon environment. Obvious applications that will require robust navigation performance are: Emergency vehicles Transportation systems Some tracking systems The Americans With Disabilities Act requires that transit systems provide visual and audible annunciations to alert passengers of stop locations. The transit systems are finding reluctance on the part of their personnel to personally make such announcements. A number of transit authorities are implementing automated systems. This problem is particularly acute in urban areas were road congestion requires high level of attention from the driver. Unfortunately, in the very areas where an automated system is most desirable, GPS data is most unreliable. Fibersense has developed an intelligent Integrated Navigation System (I2NS) that uses both software algorithms and dead reckoning to overcome the urban canyon problem. Fibersense has delivered a significant number of systems for a New York City application. These systems show significant improvement over prior systems. Fibersense has used the urban canyon environments of both Boston and New York City in order to optimize its proprietary algorithms to enhance performance.
Pure GPS Performance GPS is a ranging system that requires line-of-sight to four satellites (three satellites on an occasional basis) in order to compute position. It is a well known process to include dead reckoning, usually using a heading sensor and an odometer input, to carry the navigation process through periods when satellite information is not available.
Figure 3, Chicago Urban Canyon A photo of the Chicago urban canyons is shown in Figure 3 Figure 1, An overpass causing brief GPS signal interruption Figure 1 shows an over pass that caused signal interruption and its effect plotted in Figure 2. This is a simple situation that is addressed by the typical GPS/DR system
New York City has some of the most extensive urban canyons in the world, Figure 4. On a typical Manhattan, street Figure 5 only 17% of the sky is visible, further, since the visible sky tends to be in a straight area above the road being traveled any visible satellites will tend to have very poor geometry leading to poor GDOP.
Figure 4, New York Urban Canyons
Satellite visibility and GDOP may momentarily improve as side streets are passed.
Figure 2, Effect of signal interruption by an overpass A more difficult situation is the urban canyon environment. Within the urban canyon line of sight visibility to satellites is greatly decreased. OnStar1 reported that in their testing within the Chicago urban canyon there were significant GPS outages. Their test consisted of eight runs each of two receivers. Their test results were three-dimensional fixes (four satellite) 75% to 80% of the time. The period of the outage varied from 1 second to 60 seconds. The mean outage (less than three satellites available, prohibiting a two dimension solution) was significantly different for the two receiver types tested. Reported worst-case accuracy in the OnStar testing appears to be approximately 350 meters (1,100 ft). For the OnStar, consumer application such breaks and occasional 350 m position error is considered acceptable.
Figure 5, Simulated fisheye photo showing limited NYC sky visability
Fibersense evaluated GPS receivers from three manufacturers, in New York City, for their resistance to drop out. The three receivers were tested simultaneously so that performance can be compared directly. Data from a test run is shown in Figure 6. Note the "Blue" receiver was selected for Fibersense's system.
Figure 8, Receiver reported ground track
Figure 6, Test of three candidate receivers In addition to blockage of satellite signal, causing the receiver to loose the ability to determine position there is a major problem resulting from reflected signals. A reflected signal, as shown in Figure 7 has a longer line of sight than a direct signal that will result in an inaccurate position estimate from the GPS receiver.
Figure 8 shows the reported ground tracks from the three receivers being evaluated. The "Red" receiver shows extremely high track errors, as expected due to its poor ability to reacquire after signal loss. The "Blue" receiver generated the best position data possibly because it was able to track an additional satellite over the "Green" receiver. It is evident from the data shown in Figure 8 that pure GPS is inadequate for applications requiring reliable, accurate position information. Intelligent Integrated Navigation System Fibersense's I2NS uses the selected GPS receiver (card) coupled with a MEMS gyro, vehicle odometer and proprietary software to overcome the urban canyon environment. Fibersense's proprietary software detects when the receiver's output is being corrupted by bad line-of-sight information (due to reflections or sources of interference i.e. jamming) and corrects the data. GPS receivers provide output performance indicators such as HDOP the "Horizontal Dilution Of Precision". These performance indicators are based on the geometry of the satellites being tracked. In the urban canyon environment the true line of sight is severely restricted and satellite range is corrupted. The GPS receiver does not know which satellites are valid and which have been corrupted due to extreme multipath from strong reflections off buildings etc.
Figure 7, Effect of Reflections on Line-of Sight
It is common for GPS receivers to be tracking 6 or 7 satellites and thus compute and provide data indicating a HDOP of 2, yet have a 1,000 ft position error. The I2NS eliminates these erroneous data.
Algorithm Development The keys to the success of an integrated navigation system for use in an urban canyon environment are the algorithms. Two versions of the algorithms have been developed, one version uses the generic NMEA interface the other one uses the more GPS receiver proprietary binary messages. The version using the NMEA interface has the advantage of being more easily adaptable to different GPS receivers. In this mode, however, the raw range/range rate measurements to the satellites are not available. The navigation solution from the receiver (position and velocity) must be used. This means that the data from and individual satellite cannot be rejected. Figure 9, I2NS MEMS Gyro Performance Very conservative simulation results are shown in the two data plots in Figure 10.
When the proprietary binary messages from the GPS are used the receiver's raw range/range rate date are available. The integrated navigation system has more information available to it than the GPS by itself. The navigation information available from the odometer and gyro inputs allows the algorithms to reject the data from individual satellites. This increases the accuracy of the system in the urban environment and also allows the system to degrade more gracefully as the number observable “good” satellites reduce.
As described earlier it is not unusual for a GPS receiver to compute an erroneous "valid solution" with a low HDOP or GDOP and yet be in error by 1,000 ft. A very important part of the development of the algorithms is the measurement rejection capability. The GPS receiver can not be trusted to supply reliable data in the urban canyon environment. Fibersense has spent a significant amount of development time characterizing the performance of GPS in very severe application of New York City. These intelligent measurement rejection criteria are the difference between the system working or not working. The performance of system is strongly related to the performance of the gyro. The low cost MEM’s used in the present system have been fully characterized, see Figure 9.
Cross-Track Position Error 200
Meters
100
0
100
200
0
20
200
400
600
800
1000
Heading Error Time - Sec
1200
1400
1600
10
Deg
A stand alone GPS receiver has to start making assumptions once the number of available satellites reduces below 4. Assumptions such as constant velocity etc are not required with the integrated navigation system. As the vehicle travels along inner city streets information from different satellites will become available. Even in heavily built up areas, cross streets are passed and for brief periods unobstructed data from other satellites is available. As the “inertial” solution can navigate accurately for short periods of time/distance traveled, very little information is lost even if all 4 satellites are not processed at the same time.
These simulation results assume a very conservative scenario of only calibrating for 100 seconds then no data for 100 seconds. A 100 second period encompasses design requirement of 60 seconds selected based upon typical maximum GPS outage times for a moving vehicle. The data set was evaluated using a sliding window (100 second calibration followed by 100 seconds of navigation) that was then moved forward in time 1 second steps. This gives a continuous estimate of the position error. It also assumes a speed of 30 mph. At 30 mph the system is unlikely to have no updates for 100 seconds (unless in a tunnel). The system would also have more than 100 seconds to calibrate the system.
0
10
20
Figure 10, Simulation Results
1800
2000
Test Results
Figure 12, I2NS and Pure GPS Ground Tracks Overlaid on NY City Street Map Fibersense's I2NS system includes dead reckoning and proprietary software in addition to GPS to overcome effects of degradation to GPS. Fibersense's I2NS's, shown in Figure 13, are installed in a number of buses in New York City, providing reliable and accurate navigation information. Typical accuracy is 40 meters.
Figure 11, I2NS and Pure GPS Performance in Manhattan The system has undergone extensive testing in New York City. The performance within New York City is shown in the Figure 11.
It can be seen from Figure 11 that the GPS data takes large jumps in position while the integrated solution is far more accurate. This data is also shown Figure 12 overlaid on a street map of New York City.
2
Figure 13, I NS with External Gyro module
Summary Performance of pure GPS systems within the city (urban canyon) environment is inadequate for applications requiring reliable navigation or position information. Navigation systems previously supplied for New York City buses proved to be too unreliable.
References 1
Daniel N. Aloi, Low-Cost Automotive Navigation Alternatives, ION 57th Annual Meeting/CIGTIF 20th Biennial Guidance Test Symposium
