IMPROVING THE ACCURACY OF MEMS IMU/GPS POS SYSTEMS FOR LAND-BASED MOBILE MAPPING SYSTEM BY USING TIGHTLY COUPLED INTEGRATION AND AUXILIARY ODOMETER

Transcription

1 IMPROVING THE ACCURACY OF MEMS IMU/GPS POS SYSTEMS FOR LAND-BASED MOBILE MAPPING SYSTEM BY USING TIGHTLY COUPLED INTEGRATION AND AUXILIARY ODOMETER Thanh Trung DUONG *, Yun-Wen HUANG and Kai-wei CHIANG Department of Geomatics, National Cheng Kung University, No.1, Ta-Hsueh Road, Tainan 71, Taiwan; Tel: ; Fa: * KEY WORDS: INS, GPS, Tightly coupled, Kalman filter, Odometer. ABSTRACT: Moile mapping system (MMS) has developed rapidly over the past two decades. In that system, a direct geo-referencing system using Gloal Positioning System (GPS) and Inertial Navigation System (INS) is applied to determine the time-variale position and orientation parameters. The Kalman filter is used as an optimal estimation tool for real-time INS/GPS integrated kinematic position and orientation determination. The loosely coupled (LC) system is commonly used as an INS/GPS integration approach due to its simplicity to derive navigation information. However, in the uran area, since there is usually less than four GPS satellites can e tracked, the INS errors will grow as a function of time during GPS signal outages. In that case, tightly coupled (TC) integration should e eneficial in such poor GPS signal reception environments ecause the measurement updates can e provided even if there are less than four satellites in the sky. Additionally, vehicle ody velocities provided y the odometer are used to minimize the POS errors during GPS signal gaps. INTRODUCTION With the development of technology, the position and orientation of a moile mapping system now can e determined y a direct geo-referencing system instead of using a spatial control network as efore. Direct geo-referencing is the determination of time-variale position and orientation parameters for a moile mapping system in a certain reference frame (El-Sheimy, 1996). The most common technologies used for this purpose today are satellite positioning using the GPS and inertial navigation using an Inertial Measuring Unit (IMU). Although either technology used alone could in principle determine oth position and orientation, they are usually integrated in such a way that the INS is the main orientation sensor, while the GPS receiver is the main position sensor. Due to the cost and size, the high quality IMUs is restricted to use in the commercial MMS. The progress in Micro-Electro-Mechanical System (MEMS) technology enales complete inertial units on a chip, composed of multiple integrated MEMS accelerometers and gyroscopes. In addition to their compact and portale size, the price of MEMS-ased is far less than those of high quality IMUs as well, however, due to the lightweight and farication process, MEMS sensors have large ias, instaility and noise, which consequently affect the otained accuracy of MEMS-ased IMUs. To resolve the limitation of low cost MEMS-ased, the integration of INS/GPS is implemented. In such a system, some integration strategies may e employed. The most common mode is loosely-coupled (LC) in which the solutions from GPS such as position and velocity are used to aid the INS system. However, in LC, at least four satellites are required to help GPS can solve the solution and this requirement may not e satisfied in hostile environments like uran canyon areas. In that case, Tightly-coupled is a feasile solution. In

2 that integration mode, GPS provide the aid measurement to the INS at oservales level such as pseudo range or carrier phase; so that GPS can continuously provide measurement updates even if there are less than four satellites in the sky. In addition, an odometer mounted on the vehicle can provide velocity update, it also help to minimize the position error during GPS signal outage. INS MECHANIZATION The mechanization equation of INS is usually epressed in the navigation frame (n-frame) North-East-Down (NED), which is known as the local level frame(llf). The mechanization of INS can e descried in figure 1. In that mechanization, from the specific force (output of accelerometers) and angular rate in ody frame (output of Gyroscopes) after eing compensated for local gravity, the earth s rotation rate and the change in orientation of ody frame respect to earth s frame, position velocity and attitude of platform in LLF can e otained though a numer of mathematical procedures. INTEGRATION STRATEGIES Figure 1: the mechanization of INS There are some modes of INS/GPS integration. The most common integration scheme used today is loosely coupled (LC) integration. The typical LC integration architecture is shown as Figure 2. The position and velocity estimated y the GPS KF is processed in the navigation Kalman filter to aid the INS. This kind of integration has the enefit of a simpler architecture, which is easy to utilize in navigation systems. However, in the case of incomplete constellations, i.e. less than four satellites in view, the output of the GPS receiver has to e ignored completely, leaving the INS unaided. Figure 2. A loosely coupled INS/GPS integration architecture (Closed loop)

3 To overcome the issue of LC integration, the Tightly-coupled (TC) have een proposed. In the TC integration, the GPS pseudo-range and delta-range measurements are processed directly in the main Kalman filter (Hide and Moore, 25). The primary advantage of this integration is that raw GPS measurements can still e used to update the INS when less than four satellites are availale. This is of special enefit in a hostile environment such as downtown areas where the reception of the satellite signals is difficult due to ostruction. In addition, in the case when carrier phase GPS measurements are used, the IMU measurements will e used to aid the amiguity resolution algorithm. However, the TC integration is not commonly used to integrate GPS and INS simply ecause of its additional compleity over the loosely coupled approach. Pseudo-range and carrier phase computed Figure 3. A Tightly coupled INS/GPS integration architecture (Closed loop) Kalman filter and smoothing The integration of INS and GPS is necessary to derive the navigation solutions. The widely used method for such integration is the Etended Kalman filter (EKF) with simple mechanization equations in the local level frame, see Gel (1974) for details. The KF provides the optimal estimate of a state vector at epoch k using measurement updates, which are only availale up to epoch k. The Rauch-Tung-Strieel (RTS) ackward smoother is implemented in this study for post-processing mission to get more accuracy results. The RTS smoother consists of a forward sweep and a ackward sweep. The forward sweep is the common KF with all predicted and updated estimates and the corresponding covariance saved at each epoch of the whole mission. The ackward sweep egins at the end of the forward filter (i.e., at epoch N) ˆ ˆ with the initial conditions of s N, N = N, N P and s N, N = PN, N. The RTS algorithms can e found in details in Brown and Hwang (1992). Velocity update y odometer In this paper, an odometer is implemented to improve the navigation performance, especially during GPS signal lockage periods. It is the fact that unless the vehicle jumps off or slides on the ground, the velocity of the vehicle in the plane perpendicular to the forward direction is almost zero. This constrain can e considered as a velocity update (ZUPT) along two aes of the vehicle (right and down): v y, v z, and v v odo, where v y, v z and v are the velocity projections in the -frame along right, down and forward direction respectively. The measurement equation of odometer update can e epressed as:

4 δv δv δv y z Cnδv n δv δv εv εv εv y z (1) From this asic equation, measurement update from odometer will e integrated in EKF to aid the system for more accurately positioning mission. FIELD TEST To analyze the performance of tight integration comined with RTS smoothing for GPS and low cost INS for land ased MMS, a field test were conducted in Tainan city where two scenarios were considered, the open areas which GPS measurements were always availale and the area that high uildings esides the road that might ostacle some GPS signal. The typical views of the test environment in the Tainan city was shown as figure 4. The test platform was mounted on the top of a land vehicle. The test IMU includes SPAN-CPT of NovAtel, which was used as the IMU to provided reference trajectory, and a test IMUs, C-MIGITS III which comes from BEI Systron Donner. The test platform was shown in figure 5. Figure 4. Typical views in the field test in Tainan Figure 5. The eperiment platform Figure 6. The distant measuring instrument was mounted on a wheel The GPS raw data was collected y the duel-frequency receiver, OEM-V, which is emedded in the SPAN-CPT and the ProPak V3 receiver located on the Civil-NET station setup y the Century Instrument Company as a ase station. In addition, an odometer (OD), shown in figure 6, was mounted on one of the vehicle wheels to measure the vehicle-ody velocities. The raw GPS measurements were processed differentially using the GrafNavTM 8.1 software to provide measurement updates while processing IMUs with LC integration. The GPS updates were only used when the carrier phase amiguities are fied. The reference trajectories were processed y the raw measurement of SPAN-CPT with TC integration in smoothing algorithm. The test IMU data were processed y a low cost IMU, C-MIGITSTM III (~3k) with raw GPS measurement collected y emedded receiver of SPAN-CPT. In addition, all trajectories in LC or TC integrations were processed and smoothed y commercial IMU/GPS processing software,

5 Inertial Eplore 8.1. The position and attitude errors are then computed y the difference etween the test IMU trajectory and the reference IMU trajectory. TEST RESULTS AND ANALYSIS. The field test route of Tainan city was shown as figure 7. A five-minute static alignment process was applied at the eginning of the test to derive the initial attitude of sensors. The field test took aout 5 minutes. In this field test, the test vehicle was first going around in an open sky area for aout 3 minute then driving into the locks that GPS outages might occur in the remain test period. There were less than four satellites availale in the sky y 25% of the time. which means that position and velocity updates were not provided y GPS for LC integration during the last period of field test j y (λ = 22 o 58' ", φ = 12 o 9' ", h = ) ref. IMU/GPS DGPS Figure 7. Test trajectory Figure 8. Position errors of C-MIGITS III on the field test (Using odometer) The comparison etween LC and LC integration with odometer measurement updates of C- MIGITS III on position and heading errors were shown on figure 8 and 9, the comparison etween LC and TC integration of C-MIGITS III were shown on figure 1. As shown on figures, the POS has reliale accuracies in the open sky area. However, while passing the area that GPS signals were locked y uildings, the position accuracy degrades quickly since there are very few satellite signals availale in the sky. Hence, the measurement updates for LC are not provided so the position errors were estimated in prediction mode. In this case, etra updates provided y an odometer, which is ale to measure vehicle-ody velocities seamlessly, could e a update source of the Kalman filtering. Headind (deg) Roll (deg) Pitch (deg) Attitude Error of C-MIGIT III 1-1 LC LC+OD LC+DMI GPS Time (sec), t = Figure 9 attitude error of C-MIGITS III on the field test (Using odometer) CONCLUSIONS East (m) North (m) Height (m) Position Error of C-MIGIT III 2-4 LC -6 TC GPS Time (sec), t = Figure 1 position error of C-MIGITS III on the field test (TC and LC integration)

6 This article has shown the performance of the LC with odometer aiding and TC integration scheme with a post-mission smoothing algorithm in two different areas. The result shows that the TC integration and auiliary odometer can help to improve the accuracies of positioning and orientation of MMS significantly compared to LC, especially in case of during GPS outages and using MEMS IMU. The land-ased MMS requires cm-level position accuracy in order to get the coordinates of interested ojects correctly using direct geo-referencing techniques for precise mapping applications. The conventional LC integration requires at least four line-of-sight signals in the sky. However, the GPS signals are often ostructed in the uran area. In this case, the TC integration rings enefits since raw GPS measurements are used as updates directly. As the result shown in this article, the navigation solution computed y C-MIGITS III is suggested to provided position and orientation for land-ased MMS. However, the position and attitude accuracy of C-MIGITS III with TC integration in the down town area still do not meet the requirement of precise mapping with land-ased MMS. The future work of this study will focus on ridging the effect of GPS outages in the downtown area. Eternal sensors, such as odometer or arometer, will e considered as alternative update sources. Hence, improvements in accuracy are epected while a low cost MEMS IMU and GPS integrated with more sensors utilized TC integration scheme with a postmission smoothing algorithm for MMS applications. REFERENCE Chiang, K.W., 24. INS/GPS Integration Using Neural Networks for Land Vehicular Navigation Applications. Department of Geomatics Engineering, the University of Calgary, Calgary, Canada; UCGE Report 229. El-Sheimy, N., The Development of VISAT - A Moile Survey System for GIS Applications, Ph.D. thesis; UCGE Report No. 211, Department of Geomatics Engineering, the University of Calgary. Brown R.G.; Hwang P.Y.C., Introduction to random signals. John Wiley and Sons. New York. Gel, A., Applied Optimal Estimation. The Analytic Science Corporation. Hide, C.; Moore, T., 25; GPS and Low Cost INS Integration for Positioning in the Uran Environment, Proc. of ION GNSS 25; 25 pp Titterton, D. H.; Weston, J. L., 24. Strapdown Inertial Navigation Technology, Second edition, AIAA. Xiaoji Niu, Sameh Nassar, and Naser El-Sheimy., 27. An accurate Land-Vehicle MEMS IMU/GPS Navigation System Using 3D Auiliary Velocity updates, Journal of the Institute of Navigation, 54 (3), pp ACKNOWLEDGEMENT This work was supported y National Land Surveying and Mapping Center (NSLC 978), Taiwan and National Science Council, Taiwan (NSC E623-MY2).