| Low-cost inertial sensors are characterized by high noise and large uncertainties in the outputs such as bias, scale factor and non-orthogonality. Consequently, errors associated with a low-cost INS in terms of position, velocity and attitude grow rapidly in stand-alone mode. If good performance can be achieved with low-cost inertial measurement units (IMUs), cost in existing applications can be reduced and the development of new applications may be made feasible.; As most of the uncertainties exist in the sensor error behaviour, calibration would improve the accuracy significantly. Intensive calibration, however, would also increase the cost of using an IMU. Another way to improve the accuracy will be by augmenting the IMU with many other aiding sensors: for example, odometers or speedometers.; Choosing an appropriate estimation method is a key problem when developing an aided INS. Currently, there are three approaches: (i) traditional linearized Kalman filter (LKF) or the extended Kalman filter (EKF); (ii) sampling-based filtering such as the unscented Kalman filter (UKF) and particle filters; (iii) artificial intelligence (AI)-based estimation such as artificial neural networks (ANN) and adaptive neural fuzzy information systems (ANFIS). Of these approaches, the performance of the UKF is compared to that of other existing methods. The performance of the unscented Kalman smoother (UKS) is also compared with that of the Rauch-Tung-Striebel (RTS) smoother.; Tests have been conducted using micro-electro-mechanical systems (MEMS)-based IMUs. The most remarkable advantage of the UKF over existing methods is that it can handle large and small attitude errors seamlessly. Thus, the UKF is preferred in situations such as those where the heading error approaches the limit of what an EKF/LKF can deal with efficiently. The EKF/RTS approach would still be chosen if the attitude errors can be kept small such that the error dynamics are linear. The UKF can unify INS error models, although it cannot deal with complete uncertainties in attitude. Therefore, the development stage can be simplified by using the UKF. |
