Feature ArticlePerformance of GNSS/INS System Using MEMS Sensors
By Joshua Houghton • Mike Mutschler
Portland test area.
For years, a more precise position, velocity and attitude solution has been sought by microelectromechanical systems (MEMS) inertial measurement unit (IMU) producers. With advances in technology and new manufacturing processes, this is becoming a reality. NovAtel Inc. (Calgary, Canada) has integrated IMUs with global navigation satellite system (GNSS) receivers for nearly a decade through the SPAN (Synchronized Position Attitude Navigation) product line and Waypoint post-processing software. Historically, SPAN products have used tactical-grade IMUs, such as fiber-optic gyroscope (FOG) or ring-laser gyroscope (RLG) technology. MEMS technology has long held the promise of replacing these expensive, and often bulky, sensors with a smaller, lighter system. Recently, some of these sensors became viable for position and attitude determination. This article compares the use and accuracy of a SPAN tightly coupled GNSS/INS system with a tactical-grade FOG IMU, as well as a leading MEMS IMU in a hydrographic test case using a multibeam sonar.
Hydrographic survey benefits from GNSS positioning augmented by inertial technology for several reasons. Adding an IMU allows for a complete attitude solution, which cannot be achieved with GNSS alone. Additionally, environments, such as areas around bridges and high canal walls, present challenges for GNSS positioning as the satellite signals are subject to blockage, attenuation and reflection. Maintaining a continuous and accurate solution is critical to the resultant seafloor model because the position and attitude errors propagate through to the resultant bathymetry.
With position and attitude errors playing a key role in the final product, the error characteristics of the IMU must be considered. The difficulty with MEMS sensors is their relatively large error characteristics when compared to their higher-grade counterparts. Key error sources include bias, bias instability, scale factor and angular or velocity random walk errors. With the exception of random walk, even large errors can be estimated and mitigated if they are stable.
The MEMS IMU evaluated in this article is the STIM300 from Sensonor AS (Horten, Norway). This IMU measures 38.6 by 44.8 by 21.5 millimeters, which is approximately the size of a box of matches. The error characteristics, particularly gyro bias stability, are also very attractive when compared to other commercially available MEMS IMUs.
To investigate the performance of the STIM300 sensor, real-world data were collected in Portland, Oregon, on the Columbia River in February 2013. With the support of SEAHORSE Geomatics Inc. (Portland, Oregon), multiple patch tests were performed and data collected in a variety of GNSS conditions. As reference, a high-grade, tactical IMU was used alongside the STIM300.
The iMAR Navigation GmbH (Sankt Ingbert, Germany) FSAS is a 0.75-degree-per-hour gyro bias IMU with a history of excellent performance. The IMUs were coupled with NovAtelís OEM628 GNSS receivers. Both systems utilized NovAtelís ALIGN GNSS dual-antenna heading functionality. SPAN provided real-time position, attitude and heave solutions, and the raw data collected were post-processed using NovAtelís Inertial Explorer. Post-processing allows for a better solution as data can be processed forward and backward. The majority of errors incurred in real time can also be filtered and smoothed. All results presented in this article are from the post-processed solutions.
In this test, both systems were initialized while the vessel was docked. When the vessel departed, a few simple S-turn maneuvers were executed to observe and estimate IMU errors. After the brief initialization, survey lines were collected on the Columbia River. Two areas of interest were surveyed, including one of open sky and one close to a high embankment.
The IMUs were installed near the center of rotation aboard the 9-meter SEAHORSE Geomatics training and development vessel. Both sensors were carefully aligned to the keel, and offsets were measured to each GNSS antenna and a preproduction Norbit Group AS (Trondheim, Norway) Wide-Band Multibeam Sonar (WBMS). Conditions on the river were generally very calm on the day of testing. Results from this test represent a typical inland bathymetric multibeam survey. To continue this article please click here.
Joshua Houghton has been a geomatics engineer at NovAtel Inc. since he graduated with a bachelorís degree in geomatics engineering from the University of Calgary in 2010. He has worked in the GNSS+ group since joining NovAtel. †
Mike Mutschler cofounded SEAHORSE Geomatics Inc. to enhance users hydrographic capability with sensor specifications, systems commissioning and practical training. With 17 years in the industry, he works closely with private business, government, academia and sensor manufacturers. Mike is chair of The Hydrographic Society of America, West Coast Chapter.