| Abstract: | In low-cost micro-electro mechanical system (MEMS)-grade strap-down inertial navigation system (SINS), failure to compensate inertial sensors errors as well as un-modeled uncertainties in SINS could result in exponentially divergence in overall performance of low-cost SINSs. This study deals with the enhancement of low-cost SINS accuracy in combination of global navigation satellite system (GNSS). In this respect, a novel adaptive constrained integrated scheme for SINS/GNSS is developed based on type-2 fuzzy Hammerstein neural network (T2FHNN). To this aim, a gray-box Hammerstein neural network model are defined based on clear interpretation with the physical nature of the inertial sensors error. In addition a knowledge-based type-2 fuzzy programming extracted from inertial sensors data is also used for managing the learning rate of Hammerstein neural networks. Some vehicular real-world tests have been carried out in order to show the effectiveness and feasibility of the proposed integration scheme in the long-term performance and accuracy of the proposed navigation algorithm. The results indicate that the proposed integration algorithm improved the navigation accuracy, reliability and stability in the presence of state constraints of the stand-alone SINS during signal blockage of GNSS. |
