Session E1: LAND APPLICATIONS
Paper #1

VISION-GPS FUSION FOR GUIDANCE OF AN AUTONOMOUS VEHICLE IN ROW CROPS:
T. Bak, Aalborg University, Denmark

This paper presents real-time localization system for an autonomous vehicle based on fusion of measurements from a row guidance sensor and a GPS receiver. The system allows efficient and flexible autonomous operation of a vehicle in row crops without causing crop damage.

Most arable fields have inherent spatial variability in weed, soil, and crop properties, but conventional agricultural practice applies inputs such as herbicide and fertilizer at a constant rate. This is not only inefficient in terms of cost but also has an undesirable environmental impact. If this variability can be measured and mapped, then inputs can be varied according to a defined strategy. In order to develop reliable crop and weed maps, it is necessary to sense both the weed (identify individual weed plants) and crop population in the field. Repeated sensing allow high precision modeling in a decision support system for the farmer which may support the various field operations providing environmental and economic benefits for the farmer.

Mapping of the various field parameters has previously been done only at harvest or by hand. Cost efficient repeated sensing requires an autonomous guided vehicle with a vision system that allows monitoring of crops and weeds. Because such a vehicles can work continuously without operator, it is very efficient and able to do operations, which have been too expensive or impossible so far. Essential to the operation of such a vehicle are means for real-time infield localization and navigation. GPS allow precise absolute sensing of the position in the field, but it is not practical to guide the vehicle between crop rows on an absolute coordinate. The localization of the vehicle must also include information from sensors that sense the relative position of the crop rows directly in order not to damage the crop.

The objective of the work described here is to develop a localization system to enable inter row operation of a small autonomous vehicle. The aim is to pass between 0.25 m wide crop rows at 8 km/h without causing crop damage. The paper concentrates on the engineering aspects of the research and evaluation of an experimental system.

Field parameter sensing is undertaken using advanced vision technology that samples the field at defined locations. These locations may be dynamically redefined in order to allow e.g. gradients in the weed pressure to sample at a decreased spatial distance. The sampling points define a trajectory in the field, which is to be executed by the vehicle. The trajectory may be re-planned online in order to allow for crop row irregularities. The re-planning is based on a row guidance camera, which determines the localization of the vehicle relative to crop rows. The re-planning ensures accurate row operation at high work rates with minimal damage to the crop. A GPS system provides absolute position and is input to the trajectory execution closed loop control system. The trajectory generation is thus influenced by location relative to the local field while the actual trajectory execution is carried out in absolute coordinates GPS coordinates. The benefits are obvious outside the row crops where the row guidance camera provide no useful information. The absolute localization of the vehicle is, however, also beneficial for the field parameter sampling and field map generation.

The solution is a system that fuse data from a relative and an absolute measurement system. It allows autonomous operation and precise localization of the vehicle in the row crop, but also provide avoid having to switch from an relative to an absolute guidance in and out of the rows.

Work is underway spring 2001 at the Danish Institute of Agricultural Sciences to demonstrate the performance of the guidance system under a full range of agricultural conditions.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #2

GPS POSITIONING IN THE FAST TRACK: TRACK MODEL CONSTRAINT ENHANCEMENT FOR OEM4:
T. Ford, NovAtel Inc., Canada; K. Milnes, SportVision

The use of clock and position constraints is a standard part of GPS navigation. A 3 dimensional position constraint is normally used to facilitate the generation of differential corrections, and to provide a fixed location to which RTK vectors can be applied so a precise position may be computed. Furthermore, clock and height constraints can be used to improve the geometry provided by the satellite constellation and in some cases provide a degraded solution where none would be available. But such constraints are not particularly useful for navigation because in the first case the system is not moving, and in the second case the constraints are not accurate enough to strengthen the navigation solution significantly.

Recently, NovAtel Inc. was approached by SportVision with a request for a reasonably priced navigation system which could provide 1 metre positioning accuracy (2 sigma) on a racetrack where the satellite geometry would periodically be very poor due to restricted visibility. The SportVision problem arose because SportVision had a requirement to continuously provide accurate and near real time vehicle co-ordinates in various TV screen reference frames in order to allow production time TV annotation for all of the race vehicles in the NASCAR races. The positioning component is a significant element in this system. Initial attempts to use the NovAtel Inc. OEM4 GPS receiver with clock constraints, height constraints, a position velocity filter and a low cost inertial system did not meet the requirements. A novel approach using planar section constraints provides the required accuracy and in fact facilitates a dramatic improvement in accuracy and reliability. The method is applied to both pseudo range and carrier based positioning, and it is seen to cause a significant reduction in the time to ambiguity resolution as well as a significant increase in the ambiguity resolution reliability.

Briefly, the positioning method requires that a contiguous set of planar sections which represent the surface of the track on which the vehicle is driving is loaded into the OEM4 receiver. When a roving receiver is on the track, it searches for an appropriate planar section by projecting its approximate position onto a horizontal reference frame used by the model. Having found the appropriate planar section, the remote receiver constrains its position in the direction normal to the planar section. This constrain is used in both the pseudo range filter and in the carrier filter during the resolution stage. To be useful, the position of the defining planar sections corners must be known to high accuracy in the same reference frame as the base station co-ordinates. In this case the planar sections a produced photogrammetrically and have an accuracy of about 10 cm.

The use of such planar constraints has a dramatic impact on both pseudo range accuracy and RTK availability. This paper describes the application of the track model method to the television production as well as the positioning methodology used to constrain with planar section and will show test results which demonstrate the dramatic accuracy and reliability improvements achievable with this method.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #3

A LOW-COST PORTABLE MOBILE MAPPING SYSTEM INTEGRATING GPS, ATTITUDE SENSORS, AND A DIGITAL CAMERA:
C. Ellum, N. El-Sheimy, The University of Calgary, Canada; A. Tabsh, Premier GPS Inc.

A low-cost backpack mobile mapping system (MMS) is being developed in the department of Geomatics Engineering at the University of Calgary. MMS are multi-sensor systems that integrate various navigation and remote sensing technologies together on a common land-based or aerial platform. The systems capitalise on the strengths of the individual technologies in order to increase the efficiency of spatial data collection. Unfortunately, current mobile mapping systems are costly, large, and complex, and consequently their use has been restricted to large companies and governmental organizations. The backpack MMS overcomes the disadvantages of current MMS, and satisfies the demand for a mobile mapping system for less-specialised users.

The system integrates a dual-frequency GPS receiver, attitude sensors (a digital magnetic compass/inclinometer), and a consumer digital camera. The GPS provides estimates of the camera's position at the exposure stations and the digital magnetic compass/inclinometer provides estimates of the camera's attitude. These exterior orientation estimates are then used as weighted parameter observations in a photogrammetric bundle adjustment. The GPS is also used to precisely time-tag the attitude angles and the times of exposure. The navigation data is stored in a logging computer, while the digital images are stored in the camera itself.

The backpack MMS has numerous advantages over current methods of GIS data collection. GIS data acquisition systems of similar size and cost to the backpack system include point-wise GPS and traditional survey techniques utilising total stations. Both systems can only capture a single point at a time, resulting in extremely long field campaigns for even the most modest data acquisition campaign. The backpack system overcomes this by using digital cameras. Obviously many points and objects can be measured from a single image and thus both the time and costs required for data collection are reduced.

To test the developed system a target field of ground control points was established that simulated a "typical" urban environment in which the backpack MMS would be expected to operate. The target field had nearby vertical structures, pavement, and foliage - in short, somewhat of a worst-case environment for GPS. It also had nearby metal buildings and light standards that could influence the azimuth reported by the attitude sensors. The accuracies of points in the target field measured using the backpack MMS are examined at different camera-to-object point distances using different numbers of images and different numbers of image point measurements. With three images at a 20m object-to-camera distance, absolute accuracies less than 25 cm are achieved. This is comparable to current single-frequency point-wise GPS data acquisition systems. Furthermore, the internal agreement of points surveyed using the system was under 10 cm. The effect of including additional observations and different imaging configurations was also examined. Finally, investigations are made on the use of the photogrammetry to bridge the GPS signal outages.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #4

POS/LS - SECOND GENERATION INERTIAL/GPS FOR LAND SURVEY APPLICATIONS:
S. Baksh, Applanix Corporation

Summary
Applanix Corporation has extended the Position & Orientation System (POS) family with a new product for Land Survey applications termed POS/LS. The second-generation device uses the latest technology advances in Inertial Navigation Systems and RTK GPS to forge an integrated INS/GPS software architecture for improved performance. POS/LS delivers blended INS/GPS position, automatic Zero Velocity Updates (ZUPT), intelligent Inertial Traverse Adjustment in near real-time, and post processing, all within a smaller and lighter ergonomic backpack system. This paper will present the system concept and design, as well as an analysis of position data from the seismic survey applications to demonstrate a new level of portable Inertial Navigation performance.

Introduction
The seismic survey industry has a long-standing challenge to provide position in areas of extreme terrain and heavy vegetation. In these areas, GPS operation is limited or virtually excluded due to obstruction of satellite reception, and conventional optical instruments are slow and labor intensive.
Applanix has enhanced the surveyor's positioning capability by refining the navigation performance of inertial sensors integrated with GPS to address the challenge of topographic survey in difficult areas. The results of the POS/LS are compared to precise GPS coordinates and presented in this paper to demonstrate POS/LS position accuracy.

System
POS/LS has a proprietary software core that enables the integration of several popular IMU's with several aiding sensors. In this case, GPS and ZUPTS are used for aiding. The software design, shown in Figure 1, allowed ready integration with a modern strapdown IMU to take advantage of smaller size, lower weight, and lower power featured with the current generation of IMU. The high accuracy IMU features sensors with low noise and low drift characteristics required for precision inertial navigation. The Applanix proprietary computation process was fine-tuned to optimize the modeling of man motion and to maximize the ZUPT effect to extract maximum navigation accuracy. As a result, POS/LS is able to maintain position accuracy for a longer time than before, even when only aided with Zero
Velocity Updates.

GPS integration was done with an embedded RTK GPS receiver to provide a reliable, intelligent, and automatic inclusion of GPS into the position solution when GPS quality met acceptance criteria. The blended INS/GPS solution was complementary with the merge of self contained, time dependent INS performance and the external, time independent GPS performance influenced by satellite visibility and multipath.

Zero Velocity Updates were designed into the system as an automatic feature to maximize navigation accuracy. Whenever motionless, POS/LS automatically processes ZUPTS to limit error propagation. This feature has been exceptional in maintaining system accuracy, especially when GPS is unavailable or does not meet acceptance criteria.

Usage
The entire system is housed in a specially designed ergonomic backpack and the operator uses a Windows CE hand held computer for navigation and system commands. The backpack houses the IMU, POS computer system with embedded RTK GPS, and batteries. An accessory custom pole is provided for stabilization when doing standing ZUPTS.

The typical operation is a traverse between survey control points. Operation commences with an alignment at a control point to deduce orientation to North. Thereafter, position is continually computed for any motion sensed by the triad of accelerometers and gyros from the start point. Error drift is limited with ZUPTS at regular intervals and Position Updates from other control points along the traverse. The position coordinates recorded during a traverse are adjusted in near real-time at the closing control point, using stored Kalman Filter variables, to give an effective in-the-field re-processing function for improved accuracy.

Performance
The system performance was appraised by comparison with coordinates produced from more accurate static processed GPS data. Several scenarios were used to simulate the typical conditions for a user: no GPS and inertial navigation, intermittent GPS and inertial navigation, full GPS and inertial navigation. Whenever the system reverted to inertial navigation only, ZUPTS were the only aiding applied. This paper will describe and analyze the results from POS/LS as the second-generation application of portable inertial navigation.

Conclusions
From the empirical data, it is shown that POS/LS can reliably provide position accuracy, even when GPS is unavailable. With full GPS, the solution reflects the cm level accuracy of RTK GPS. Altogether, the system has addressed the problem of positioning in difficult areas with a device that provides acceptable position accuracy and simple operation for the user community.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #5

3D VIRTUAL EARTH MODELS BY GPS INTEGRATED WITH LASER RANGEFINDERS AND OBLIQUE CLOSE-IN DIGITAL PHOTOGRAPHY:
X. Xu, C.L.V. Aiken, Center For Lithospheric Studies, University of Texas at Dallas

GPS has been the key in capturing and integrating high accuracy and resolution models of the natural surfaces of the earth such as geologic outcrops. We either use reflectorless laser rangefinders to identify and "laser sketch" geologic features such as contacts and faults and terrain at centimeter to decimeter, or we combine that with close-in oblique digital photography which is then draped onto the terrain models, even with a single photograph. In the latter case we have captured almost all the information of interest and can take it back to the lab to produce the photorealistic outcrop for 3D measurement and analysis. In either case we can quantitatively analyze the geometries such as orientations and thicknesses, and integrate the data because it is positioned globally by GPS with other surface and subsurface data. We use ruggedized field computers so that using our own software we can integrate, visualize and analyze the results in real time in the field. We have installed and viewed our models in 3D virtual immersive environments at several sites. This method has been used to map several areas in Utah, Oklahoma, Wyoming, Nevada, Texas, California and Arkansas and Egypt. Although we have used post-processing GPS surveying, it is most efficient when we use RTK GPS or at least DGPS of course depending on the accuracy of the integration sought.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #6

PERFORMANCE TESTING OF A LOW COST SYSTEM FOR AUTOMATED DOCUMENTATION OF CRASH SCENES USING GPS (AUTODOCS-GPS):
C. Rodgers, OPTIMUS Corporation

OPTIMUS Corporation, under a contract to the National Highway Traffic Safety Administration (NHTSA), is developing a low cost (< $1.5k), automated, automotive crash scene, documentation system that utilizes the GPS for rapid, accurate measurements. NHTSA performs research on the causes and circumstances of automotive crashes in order to promote safer roadways and vehicles. An important sources of data for this research comes from the analysis of crash scenes. The position and orientation of vehicles relative to each other and roadway features are important to determining the crash dynamics and cause. Currently, crash scene measurements are done manually with wheel and tape measures. The data are recorded by hand and the analysis is done through manual calculation. The results of the measurements and calculations are captured on two-dimensional drawings developed by hand. The number of manual operations in the measurement and analysis are a source of inefficiency and errors. In addition, data and sketches recorded on paper are difficult for the researchers to access, correlate, manipulate, and analyze.

OPTIMUS is integrating a low cost system for a laptop computer that addresses these inefficiencies and inaccuracies by automating the entire process. We have developed a Kinematic Differential GPS subsystem to measure, record, and process the scene measurements as the user walks about the scene. The custom software links each measurement, as well as digital pictures, to scene elements, and can guide the user through the crash scene measurement process. Once the scene dimensions are determined, a Computer Aided Drafting (CAD) program is called to automatically construct a scene drawing using predefined scalable, graphical crash scene elements. An automated report generator then enters the scene data into a standardized text report complete with hyperlinks to the digital pictures. The report file, three-dimensional scene data, and CAD file can be stored in various formats, and transmitted, retrieved, manipulated, and analyzed as any other computer file.

The paper discusses the system architecture, Kalman filter design, and simulation and prototype test results. The performance of the new system is compared with current methods as performed by expert police officers on simulated scenes. The process and results of the two methods are evaluated to quantify the improvement in accuracy and efficiency provided by AutoDOCS-GPS. Results show that the AutoDOCS-GPS is three to five times faster than current methods, possibly significantly reducing traffic delays when used by local authorities. Each mile of traffic backup on a major highway costs the economy millions of dollars in lost productivity and fuel, and increased pollution, in addition to the psychological toll the frustration causes on the public. To date, the measurement errors of the new system's kinematic Kalman Filter have averaged 1.2 cm for scene element points, which is less than half of the resolution of the current method without even considering human measurement and recording mistakes. These accuracy and efficiency improvements combined with the electronic format of the data will greatly enhance NHTSA's capabilities and the capabilities of local authorities.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #7

GPS-BASED PROXIMITY WARNING SYSTEM FOR MINING AND CONSTRUCTION EQUIPMENT:
T. Ruff, National Institute for Occupational Safety and Health; T. Holden, J. Hamilton, A. Samaha, Trimble Navigation, Inc.; G. MacGougan, University of Calgary, Canada

Each year, an average of 6 fatalities occur in surface mines that involve a collision between a piece of mining equipment and a smaller vehicle or pedestrian worker, or involve a piece of mining equipment going over the edge of a dump point. Researchers at the National Institute for Occupational Safety and Health (NIOSH), Spokane Research Laboratory, in cooperation with Trimble Navigation Inc., are developing a proximity warning and edge detection system based on GPS technology and wireless network communications. A prototype system successfully demonstrated that vehicle location information can be transmitted between multiple mobile vehicles. Each vehicle's system consisted of a Trimble GPS antenna and receiver, a laptop computer for data collection and display, and Internet Protocol (IP) radio PC cards. This paper will describe the system components, the software interface, and tests conducted at Trimble and NIOSH. The prototype phase of this project has shown that GPS technology can provide an effective means of tracking the locations of nearby vehicles and providing a warning to the equipment operator if other vehicles are within dangerous proximity. Additional tests, completed this summer, verified system operation in a surface mine and showed additional problems that were encountered in this constantly changing environment. These problems include GPS satellite availability, location update rates, and operator interface issues. Other issues that also need to be addressed in future work include the protection of pedestrian workers and methods to insure system reliability.
[Back to Program]

Session E1: LAND APPLICATIONS
Paper #8

HIGH-ACCURACY GPS AND GRAVITY SURVEYS IN NORTH SLOPE, ALASKA:
T. Chen, J. Ferguson, C. Aiken, J. Siebert, J. Hare, J. Brady, The University of Texas at Dallas

Between 1994 and 2001, four high-accuracy GPS surveys have been carried out on the North Slope of Alaska supporting micro-gravity surveys. The goal has been to determine whether a surface gravity/GPS survey could be used to map density changes in a deep hydrocarbon reservoir due to water injection in the gas cap. A water injection program might last for decades and failure of the water to move in the subsurface as planned could result in a serious loss of oil production. Due to various environmental and logistical problems, which are peculiar to the Arctic, it is not practical to monitor the water flood by drilling wells. A time differenced or "4D" gravity survey was proposed as a cost effective alternative. Simulations of the water flood have indicated a requirement for sub 10 micro-Gal noise level in the gravity and 1 to 2 cm accuracy in elevation for about 500 stations.

In the Arctic environment it is impossible to establish any kind of permanent monument to facilitate resurveys (many of the stations are on seasonal sea ice!). GPS control is the only practical way to establish repeatable stations. Local monuments used for differential GPS measurements are re-established during each survey with respect to distant permanent (CORS network) control. Once a station is originally established the use of high-precision RTK GPS is necessary for its recovery. The latitude and the time of year of the survey also make the attainment of centimeter-level accuracy more challenging due to satellite geometry and ionospheric activity.

In the Arctic measurement of gravity at the micro-Gal level is also problematic. The fundamental resolution of modern gravity measurement is near 1 micro-Gal, but wind generated seismic noise, ice movement and other phenomena must be countered by refined procedures. Three fundamentally different gravity measuring devices, The LaCoste and Romberg Super-G meter, Scintrex CG-3M meter and Micro-g A-10 absolute meter were used in the winter at temperatures sometimes below -40 degrees F. The GPS antenna was mounted on top of the instruments for a constant known position relative to the gravity sensor.

About 30 stations were established in 1994 (and about 15 more in later years). Results after multiple occupations, over 1 to 3 year intervals, indicate that the required accuracy can be attained by this combination gravity/GPS survey techniques.
[Back to Program]

Session E1: LAND APPLICATIONS
Alternate #1

DATA PROCESSING AND MULTIPATH MITIGATION APPROACHES FOR GPS/ACCELEROMETER BASED HYBRID STRUCTURAL DEFLECTION MONITORING SYSTEM:
G.W. Roberts, X. Meng, A.H. Dodson, Institute of Engineering Surveying and Space Geodesy, University of Nottingham, UK

Research using kinematic GPS and triaxial accelerometers is conducted at the IESSG, University of Nottingham. In this paper, a dual reference station technique for monitoring a suspension footbridge with a known pedestrian loading is introduced. Detailed data processing methods are presented, which include moving averaging (MA) for 'long term' changes. These changes are mainly the result of multipath effects both at the reference stations and observation sites upon the bridge. Data decomposition for 'short term' fluctuation of real bridge movements and frequency identification using spectrum analysis to both GPS time series and accelerometer data is also investigated. In addition to supplying more robust and higher accurate resolution for the bridge movement, complete 3D frequency distribution are produced through the integration of GPS and accelerometer data sets, at rates typically of 300 Hz.

The algorithms used with dual reference stations for mitigating receiver site random noise and multipath are discussed. The adaptive filtering is used for reducing random noise and the multipath effects of the observation sites on the bridge. It is also found through these bridge trials that the position accuracy can be greatly improved for the site, even when using a lightweight antenna mounted by using adaptive method.
[Back to Program]

Session E1: LAND APPLICATIONS
Alternate #2

THE APPLICATION OF GPS TECNIQUES ON THE "RUINON" LANDSLIDE MONITORING:
R. De Stefano, M. Arciero, Telespazio

Space Geodetic technology, and in particular the GPS offers an efficient monitoring methodology for areas at high risk, such as landslide phenomena, subsidence and structural deformation. Telespazio S.p.A. makes possible the distribution of services related to territorial measurements and fleet navigation with high level of accuracy, by using a permanent network of GPS stations distributed over the Italian territory. One year ago, in collaboration with Regione Lombardia, Telespazio monitored a landslide called Ruinon located in Valtellina area (northern Italy) using GPS technology. The basis of the system was two GPS receivers placed on the landslide body, and a GPS reference station located at about ten kilometers from the landslide area. All GPS receivers remained in operation continuously and were equipped with telecommunications system based on GSM. The system is controlled by a Control Center located at Centro Servizi Telespazio in Rome. The Control Center also performs data acquisition and analysis using automatic procedures that allow updating of the movement maps. The aim of this work is to describe system architecture and to show results obtained during the monitoring of the 'Ruinon' landslide. The results highlight:

- A landslide characterised by an irregular movement in space and in time.
- Displacement, integrated over four months of monitoring, of almost 40 cm
- A different plane displacement of the two GPS receiver placed on the landslide body that show a broadening of landslide area
- Irregular height displacement characterised by plateau and great slope
[Back to Program]