An intelligent and autonomous MEMS IMU/GPS integration scheme for low cost land navigation applications

Abstract

An intelligent scheme to integrate inertial navigation system/global positioning system (GPS) is proposed using a constructive neural network (CNN) to overcome the limitations of current schemes, namely Kalman filtering (KF). The proposed CNN technique does not require prior knowledge or empirical trials to implement the proposed architecture since it is able to construct its architecture “on the fly,” based on the complexity of the vehicle dynamic variations. The proposed scheme is implemented and tested using Micro-electro-mechanical systems inertial measurement unit data collected in a land-vehicle environment. The performance of the proposed scheme is then compared with the multi-layer feed-forward neural networks (MFNN) and KF- based schemes in terms of positioning accuracy during GPS signal outages. The results are then analyzed and discussed in terms of positioning accuracy and learning time. The preliminary results presented in this article indicate that the positioning accuracy were improved by more than 55% when the MFNN and CNN-based schemes were implemented. In addition, the proposed CNN was able to construct the topology by itself autonomously on the fly and achieve similar prediction performance with less hidden neurons compared to MFNN-based schemes.