A Positon and Orientation Post-Processing Software Package for Land Applications - New Technology

Transcription

1 A Positon and Orientation Post-Processing Software Package for Land Applications - New Technology Tatyana Bourke, Applanix Corporation Abstract This paper describes a post-processing software package that uses tightly-coupled inertial/gps integration for land-vehicle mobile survey applications. The advantages of using this software, particularly in areas where GPS reception is problematic, are explained in detail. In addition, the paper describes the results of a performance test and corresponding data analysis to illustrate the accuracy improvements when compared to loosely-coupled integration, and GPS only post-processing methodology. Introduction In navigation applications where the accuracy of computed position and orientation data is a key requirement, postprocessing techniques should be applied. These techniques utilize processing in both forward and reverse directions to significantly increase the data accuracy beyond that of real-time performance. Because gyros and accelerometers do not measure acceleration and angular rate perfectly, the inertial solution will contain errors that increase over time. These errors can be corrected if data from other sensors, such as a DMI are available. The DMI will measure distance traveled and effectively correct the inertial solution when GPS data is not available. A GPS receiver can provide (to a specified accuracy) the position of the GPS antenna from satellite signals processed by the receiver. A solution with centimetre-level accuracy is achievable when differential GPS techniques are used and more than four satellites are visible. When this satellite visibility condition is not met, due to shading or obstacle interference, GPS There are several different paradigms for integrating inertial and GPS systems, and data. In this paper, we will consider looselycoupled and tightly-coupled inertial/gps integration. Loosely-coupled integration implies a Kalman filter processes the GPS computed position and velocity together with the inertial data. This provides a blended navigation solution that includes position and orientation. The GPS computed position and velocity included in the process is normally obtained using postprocessed differential GPS methodology. Tightly-coupled integration implies the Kalman filter processes raw GPS observables (pseudoranges, phases and In one pass, data can be acquired (viewed in real-time and/or collected for post-processing) and analyzed for precise results. POSPac software, developed by Applanix, applies this methodology to the combined measurements generated by the IMU (Inertial Measurement Unit), the DMI (Distance Measurement Indicator), and GPS. The result is an optimal blended position and orientation solution resistant to partial or complete GPS satellite signal loss. The inertial data is the key component in generating an accurate solution. Inertial measurement systems provide continuous position and orientation data by sensing vehicle acceleration and angular rate, using a combination of gyros and accelerometers. The inertial solution is always available regardless of the environment in which the vehicle is operating, such as in tunnels, under bridges and dense forest canopies, or in urban areas. cannot provide a three-dimensional position solution. The inability of a GPS-only system to provide a continuous solution, regardless of the environment, is a major limitation. By using a combination of inertial and GPS data, the limitations of both systems can be removed to provide continuous, accurate, and reliable solutions in all vehicle environments. In an integrated inertial/ GPS navigation system, when satellites are visible, the GPS data is the dominant source of information for position computations. In addition, the GPS data is also used for inertial sensor calibration. When visibility is poor, the calibrated inertial sensor data is the dominant source of information for position computations. Inertial/GPS Integration Techniques Doppler velocities) together with inertial data to provide the blended navigation solution. The Kalman filter also estimates floated double-differenced (DD) phase ambiguities that are used in an ambiguity resolution algorithm which fixes the floated estimates to integer values. There are significant benefits to using a tightly-coupled solution in a land environment when satellite visibility is blocked and signal interruption occurs, reducing satellite visibility to less than four. This technique is implemented in the POSPac software. For example, when signals from only two or three satellites are received, loosely-coupled processing ignores the data since no GPS navigation solution can be computed. Tightly-coupled processing effectively uses the data to correct the inertial navigation solution

2 Page 2 POS Data Inertial Data Smoothed GPS Data Solution Base Station Data (optional) Extraction DMI Data Conections Aided Inertial Output Output Position and Orienation Data Real-Time Data Figure 1 : Software Architecture and produce a more accurate position and orientation result through the partial signal outage. Post-processing Software Architecture The POSPac post-processing software developed for land applications computes an optimal smoothed blended navigation solution. It employs post-processing techniques to achieve the best possible position and orientation accuracy by employing a tightly-coupled integration approach. Inertial Data Inertial Navigator Error Controller Navigator Conection Blended Soultion Feed- Forward Error Controller Smoothed Errors, Covariance Smooth Solution The software architecture is shown in Figure 1. It includes the following elements: Extraction extracts and performs a quality check on the recorded inertial, GPS and DMI data Aided Inertial computes the optimal smoothed blended navigation solution from the inertial, GPS and DMI data Output transforms the smoothed navigation solution in to different coordinate systems and datum and provides various output formats GPS Observables DMI Data Base GPS Observables Error State Connection Floated Ambiguities Kalman Filter Estimated Errors Estimated Errors, Covariance Integer Ambiguities Smoother The functional architecture of the aided navigation is given in Figure 2. The software consists of the following components: Strapdown Inertial Navigator Kalman Filter Error Controller Ambiguity Resolution Module Smoother Feed-Forward Error Controller Ambiguity Filter Figure 2 : Aided Inertial Architecture

3 Page 3 The strapdown Inertial Navigator solves Newton s equations of motion on the rotating earth from accelerations and angular rates sensed by the strapdown IMU. In order to do this, the inertial navigator must first be initialized with a known position and velocity, and aligned with respect to true vertical and true north. Alignment with respect to true vertical is called levelling, and alignment with respect to true north is called heading alignment. The Kalman Filter processes measurements which relate the strapdown navigation solution to the data from the aiding sensors. This is done in order to obtain estimates of the errors in the inertial navigator, IMU, DMI and GPS receiver. It reflects the tightly-coupled integration. The tightly-coupled GPS measurements contain the observables directly, whereas a loosely-coupled integration employs the derived GPS position and velocity. The Kalman filter estimates gyro and accelerometer biases and scale factors, DMI scale factor and alignment errors, and floated DD phase ambiguities. Notably, the GPS observables measurements provide ongoing aiding to the inertial navigator when the number of visible satellites drops below the minimum of four. data from the available suite of navigation sensors. This is the most accurate navigation solution obtainable from the inertial and aiding sensor data. Comparison Testing A van test was conducted in order to evaluate the performance of the POSPac tightly-coupled post-processing software in a poor GPS environment. Comparisons were made with loosely-coupled and GPS-only post-processing solutions. Data was acquired along a road partially lined with mature pine trees. Satellite visibility was poor (Figure 3) and varied from zero to seven satellites (SVs). A POS LV system was used to collect inertial, GPS, DMI and GAMS (GPS Azimuth Technology) data. Base station data were recorded and used in post-processing. The baseline length was 3.5 kilometres. NUMBER OF SV S The Error Controller computes and applies corrections to the inertial navigator from the estimated navigation errors generated by the Kalman filter. It continuously aligns the inertial navigator and regulates its position and velocity to be consistent with the accuracy of the aiding sensors (GPS and DMI). The blended navigation solution exhibits the combined attributes of position solution continuity from the inertial navigator and position accuracy introduced by the error correction. The Ambiguity Resolution algorithm operates on the floated DD phase ambiguities and attempts to fix the ambiguities to integers. A single Kalman filter contains both inertial navigation error states and floated DD ambiguity states. Thus the observability of the floated ambiguities is enhanced when the inertial position error is sufficiently small. The Smoother computes the optimal estimates of the inertial navigator and IMU sensor errors by processing the data backwardsin-time and then combining it with the estimates from the forwardin-time Kalman filter. It performs a variance-weighted combination on the forward and backward estimated error and covariance data to produce the smoothed error estimates and covariance. The smoothed error estimates are based upon all available information from the past and the future, and hence is more accurate than a forward-only approach (i.e. real-time). The Feed-Forward Error Controller uses the optimal smoothed error estimates from the smoother to generate a correction which is applied to the integrated inertial navigation solution. This produces a smoothed blended navigation solution, called the Smoothed Best Estimate of Trajectory (SBET). The SBET at each time point is the optimal navigation solution based on all past, current and future Figure 3 : Number of visible satellites The van was driven through the area of poor GPS reception six times, following the same route to accumulate enough statistical data to allow for detailed analysis. The duration of poor GPS reception lasted between five and six minutes. The POSPac data generated was compared to a dataset produced using a high-accuracy inertial navigation system, utilizing ring laser gyro technology. The post-processed accuracy of the reference data was better than 5 centimetres over a five minute period of unaided data. The logged IMU, GPS, DMI and GAMS data were extracted and processed using POSPac tightly-coupled, POSPac loosely-coupled and GPS-only post-processing software. Smoothed solutions were obtained for all three cases. Performance results against the reference data are shown on Figures 4, 5, and 6. Figure 4 shows the position differences between POSPac looselycoupled smoothed solution and the reference solution. Figure 5 shows the same differences for POSPac tightly-coupled smoothed solution. The table in figure 6 summarizes the performance of

4 Page 4 POSPac tightly-coupled, POSPac loosely-coupled and GPS-only post-processing in situations of poor satellite visibility, and signal interruption. POSITION DIFFERENCE (M) LOOSELY - COUPLED Figure 4 : Loosely-coupled position measurements POSITION DIFFERENCE (M) TIGHTLY - COUPLED Figure 5 : Tightly-coupled position measurements Conclusion This paper describes the post-processing software (POSPac) which utilizes tightly-coupled inertial/gps integration, to generate accurate and continuous position and orientation data. This integration technique has specific advantages for land applications particularly in areas where GPS reception is subject to partial signal loss, a result of obstacles such as trees and buildings blocking the signal. A single Kalman filter processes inertial data, together with GPS observables and DMI measurements to estimate the sensor errors and floated DD phase ambiguities. GPS observables measurements provide ongoing aiding to the inertial navigator and are utilized even when the number of visible satellites drops below the minimum four needed to compute a GPS position. The results of the van test demonstrate a significant improvement in the performance of POSPac software with tightly-coupled inertial/gps integration, over the loosely-coupled alternative, particularly when satellite visibility is poor. After applying the smoother, the position error RMS was less than 20 centimetres (horizontal) and approximately 23 centimetres (vertical) in an area that was densely lined with trees. The tightly-coupled solution shows almost twice the accuracy of the loosely-coupled solution. The GPS-only post-processing results show the worst performance and robustness. For areas with four or more satellites the horizontal position error RMS was in the order of 60 centimetres. The vertical position accuracy was in the order of 2.6 metres. POSPac s tightly-coupled inertial/gps integration approach to data post-processing produces a highly-accurate position and orientation solution resistant to partial or complete blockage of GPS satellite signals. The software brings significant benefits to those organizations seeking continuous position and orientation accuracy while operating in a land environment. RESULTS TABLE RMS Error (M) GPS Only Loosely Coupled Tightly Coupled (POSPac LAND) NORTH EAST DOWN Figure 6 : Results of the comparison test The statistical analysis showed that POSPac tightly-coupled performance in a hostile GPS environment was consistently better than loosely-coupled for both horizontal and vertical directions. The GPS-only solution showed the worst performance, and to reiterate, did not provide any navigation solution when less than four satellites were visible.

5 Page 5 References Sherzinger, B., A Position and Orientation Post-Processing Software Package for Inertial/GPS Integration (POSProc). Proceedings of the International Symposium on Kinematic Systems in Geodesy, Geomatics and (KISS97), Banff, Canada, June 1997 Sherzinger, B., Precise Robust Positioning with Inertial/GPS RTK. Proceeding of ION-GPS-2000, Salt Lake City, UH, September 2000 Sherzinger, B., Robust Inertially-Aided RTK Position Measurements. Proceeding of KISS2001, Banff, Canada, June 2001 Sherzinger, B., Robust Positioning with Single Frequency Inertially Aided RTK. Proceeding of ION-GPS-2001, Salt Lake City, UH, September 2001