Aided versus embedded-a comparison of two approaches to GPS/INS integration

The performance of two major design options for GPS/INS integration are studied in this paper. The first is the embedded approach, in which the receiver card is an integral part of the system and GPS-information at the measurement level is directly integrated into the navigation processor. The second is the aided approach, in which GPS information is processed independently and the resulting position and velocity is used to update the INS after appropriate time intervals. To study the two design options, the same mathematical model is used to operate two different filters. For the embedded approach, a centralized Kalman filter is applied. In this case the INS and GPS error models define the common state vector. INS input is used to define the reference trajectory which is then updated by GPS pseudo-ranges and range rates. For the aided approach, a decentralized filter design is chosen. In this case, the state vectors for INS and GPS are defined and operated independently. At specified time intervals the derived GPS position and velocity is used to update the INS position and velocity. The time intervals are determined from the error spectra of both systems. Results of simulations for a land vehicle application show that for a variety of scenarios the two design approaches give about the same accuracy, i.e. the tighter integration of the embedded approach does not result in superior performance. Reasons for this somewhat surprising result are discussed in the paper