Low-cost INS/GPS integration: Concepts and testing

The high cost of inertial units is the main obstacle for their inclusion in precision navigation systems to support a variety of application areas. St andard inertial navigation systems (INS) use precise gyro and accelerometer sensors; however, newer inertial devices with compact, lower precision sen sors have become available in recent years. This group of instruments, call ed motion sensors, is six to eight times less costly than a standard INS. G iven their weak stand-alone accuracy and poor run-to-run stability, such de vices are not usable as sole navigation systems. Even the integration of a motion sensor into a navigation system as a supporting device requires the development of non-traditional approaches and algorithms. The objective of this paper is to assess the feasibility of using a motion sensor, specifica lly the MotionPak(R), integrated with DGPS and DGLONASS information, to pro vide accurate position and attitude information, and to assess its capabili ty to bridge satellite outages for up to 20 seconds. The motion sensor has three orthogonally mounted 'solid-state' micromachined quartz angular rate sensors, and three high performance linear servo accelerometers mounted in a compact, rugged package. Advanced algorithms are used to integrate the GP S and motion sensor data. These include INS error damping, calculated platf orm corrections using DGPS (or DGPS/DGLONASS) output, velocity correction, attitude correction and error model estimation for prediction. This multi-l oop algorithm structure is very robust, which guarantees a high level of so ftware reliability. Vehicular and aircraft test trials were conducted with the system and the results are discussed. Simulated outages in GPS availabi lity were made to assess the bridging accuracy of the system. Results show that a bridging accuracy of up to 3 m after 10 seconds in vehicular mode an d a corresponding accuracy of 6 m after 20 seconds in aircraft mode can be obtained, depending on vehicle dynamics and the specific MotionPak(R) unit used. The attitude accuracy was on the order of 22 to 25 arcmin for roll an d pitch, and about 44 arcmin for heading.