Nonlinear Filtering with IMM Algorithm for Ultra-Tight GPS/INS Integration

Abstract

This paper conducts a performance evaluation for the ultra‐tight integration of a Global positioning system (GPS) and an inertial navigation system (INS), using nonlinear filtering approaches with an interacting multiple model (IMM) algorithm. An ultra‐tight GPS/INS architecture involves the integration of in‐phase and quadrature components from the correlator of a GPS receiver with INS data. An unscented Kalman filter (UKF), which employs a set of sigma points by deterministic sampling, avoids the error caused by linearization as in an extended Kalman filter (EKF). Based on the filter structural adaptation for describing various dynamic behaviours, the IMM nonlinear filtering provides an alternative for designing the adaptive filter in the ultra‐tight GPS/INS integration. The use of IMM enables tuning of an appropriate value for the process of noise covariance so as to maintain good estimation accuracy and tracking capability. Two examples are provided to illustrate the effectiveness of the design and demonstrate the effective improvement in navigation estimation accuracy. A performance comparison among various filtering methods for ultra‐tight integration of GPS and INS is also presented. The IMM based nonlinear filtering approach demonstrates the effectiveness of the algorithm for improved positioning performance

Multi-Sensor Fusion with Interaction Multiple Model and Chi-Square Test Tolerant Filter

“…However, when measurement sub-systems return to work, the accuracy of the integrated navigation system returns promptly to the normal operation. In summary, the comparisons reported in this section are mainly with respect to [20,21] where a similar multi-sensor setup based on a SINS/GPS/BD2/CNS integrated navigation system is considered. The two scenarios considered in the paper are reported to achieve the following goals:

In Scenario 1, the goal is to evaluate the potential achievable performance improvements of the proposed methodology both in terms of the reduction in the false alarm rate and improvements in failure detection sensitivity.

…”

“…The two scenarios considered in the paper are reported to achieve the following goals:

In Scenario 1, the goal is to evaluate the potential achievable performance improvements of the proposed methodology both in terms of the reduction in the false alarm rate and improvements in failure detection sensitivity. In this regard, we present comparison results between the proposed algorithm and [20,21] to verify the validity of the proposed fault detection algorithm. It is observed that when we lower the accuracy of the inertial navigation unit, the algorithm proposed in [21] fails to work properly.

…”

“…In this regard, we present comparison results between the proposed algorithm and [20,21] to verify the validity of the proposed fault detection algorithm. It is observed that when we lower the accuracy of the inertial navigation unit, the algorithm proposed in [21] fails to work properly. This is mainly because of the absence of a precisely known model of the system in advance.…”

“…In this case, both algorithms from [20,21] fail to provide acceptable results, and even the computed estimates based on these algorithms diverge from the ground truth. We note that the shortcoming of these algorithms in handling this practical scenario is due to the presence of soft failures and high aircraft manoeuvres.…”

“…However, in scenarios where the accuracy of the inertial measurement units in the SINS is low, the aircraft manoeuvre will damage the accuracy of the SINS, thereby, reducing the failure detection sensitivity. The interacting multiple model (IMM) estimation algorithms [21,22,23] are, therefore, designed to overcome this difficulty by incorporating a set of candidate models to represent different uncertainties under realistic conditions. Research on the fault detection and isolation problem with the IMM algorithm has attracted a recent surge of interest [24,25,26,27,28,29].…”