Integration of MEMS INS with GPS Carrier Phase Derived Velocity: A new approach

In view of the fast development of MEMS technology, the quick reduction of MEMS sensor cost, and the continuous improvements in MEMS sensor accuracy, MEMS based INS have become an increasingly important geospatial sensor in navigation, positioning and control applications. It is believed that MEMS INS could eventually replace the conventional INS sensors in middle to lower end applications. However to fully exploit its potential, there are still many challenging issues to be solved. Compared with the tactical grade conventional strapdown INS, the performance of MEMS INS is still suffering from large errors including significant scale factor nonlinearities, misalignment, high noise, and temperature varying biases. Precisely modeling and compensating for these errors present significant challenges from several research aspects. Together with this trend, integrated geo-referencing systems for mapping applications are gradually turning from the post processing mode to the on-line processing mode. Accuracy is not the only bench mark that users are looking for, but also real-time delivery ability, for examples in UAVs and some emergency services. Also a standalone working mode is more favorable for lower end users in terms of cost reduction, and flexibility. To that end, this paper investigates an integration scenario where a standalone GPS receiver has been integrated with a MEMS INS, the Crossbow IMU400. To improve the performance, additional velocity aiding and a new integration structure have been adopted in this implementation. Processing of field test data shows that velocity determination accuracy could reach the one centimeter level when time-differenced carrier phase measurements from one single GPS receiver is processed using the proposed velocity algorithm. MEMS sensor errors have been limited to such a level that MEMS INS can be used to generate stable attitude and heading references under low dynamic conditions. These positive test results have validated the effectiveness of the proposed integration approach. INTRODUCTION Integration of GPS and INS is a widely accepted solution in providing position, velocity and attitude in modern navigation, and geo-referencing applications. The combination of these two sensors have the advantages of long term positioning accuracy, high update rates, robustness to GPS signal jitters and interference sources, and continuous calibration of INS errors during operations so that the inertial navigation performance could be improved during GPS outages. Without considering military applications, this kind of integrated systems is conventionally targeted at higher end applications, such as mobile mapping and remote sensing mainly due to the high cost of the sensors. Therefore the main technical concerns of such systems are on data post processing to achieve surveying level position, velocity and attitude accuracies. However, with more INS sensors being manufactured with MEMS technology, the situation has changed dramatically. MEMS INS provides a more affordable solution for middle to lower end applications like car navigation, UAV, micro-robots and other location based services. With the advance of the MEMS technology, and more smaller-size and more accurate MEMS sensors becoming available, there is an increasing trend to use MEMS INS as a substitute for more expensive conventional INS sensors in real-time oriented applications. At present, MEMS INS sensors are still not performing as well as conventional high-grade sensors in view of significant scale factor nonlinearities, misalignment, high noise and temperature varying biases. Regular calibration using external aiding sources is essential to limit rapidly growing errors during operations. Theoretically, GPS is capable of providing precise positioning information and is ideal for being utilized as an INS aiding source. With GPS Selective Availability (SA) turned off, a stand-alone GPS receiver can reach positional accuracies as low as several meters. Differential GPS (DGPS) can provide a positioning accuracy of sub-meter level for real-time applications. Using RTK technology, even centimeter level accuracy is achievable in real-time. However, these high accuracies are not always readily achievable due to several limitations: • The GPS signal can become very weak under some operating environments, and is sensitive to signal blockage and attenuation. • To carry out high accuracy positioning, differencing of two GPS receivers’ measurements are required in order to eliminate uncertain errors. Hence, infrastructure support is necessary such as continuously operating reference stations (CORS) and other communication facilities between reference and rover stations. • Resolving the carrier phase integer ambiguities is critical for achieving centimeter level accuracy, which is challenging and not always reliable especially in the RTK mode. Under most situations of low-cost real-time applications, the usage of standalone GPS is dominant due to cost considerations, or infrastructure limitations. So using the measurements from standalone GPS to precisely calibrate MEMS INS errors and to improve the integration performance to the level that a moderate-cost conventional INS would provide is a very attractive research subject topic. This article describes the structure, implementation and experimental results of an integrated system in which MEMS INS errors have been calibrated using carrier phase derived velocity information of a stand-alone GPS receiver. The overall structure of this implementation includes two calibration steps. The inner loop estimates and corrects IMU errors using delta positions between GPS epochs. The outer feed forward loop bounds the integrated position and attitude error growth using GPS pseudorange measurements so as to eliminating the bias of the inner filtering loop. The algorithms described here are suitable for real-time applications.