A reliable multisensor fusion strategy for land vehicle positioning using low-cost sensors

Abstract

Many intelligent transportation system applications such as vehicle guidance, assistance driving, and safety alert require accurate, reliable, and continuous vehicle positioning whether in open environments or in Global-Positioning-System-denied environments. The way in which such positioning performance can be achieved using low-cost sensors is the main challenge for land vehicles. This paper proposes a reliable and cost-effective multisensor fusion strategy for real-time vehicle positioning. First, several key-data-preprocessing algorithms are developed to provide more accurate and richer information for subsequent global fusion. They include a sliding-mode observer for eliminating the noise from the wheel velocity, a recursive least-squares algorithm for vehicle attitude estimation and angle compensation, and a vehicle state estimator for estimating the unmeasured lateral vehicle state. Further, an interacting-multiple-model-based fuzzy-decentralized-information-filtering algorithm is developed to realize global fusion. In the implementation of the interacting-multiple-model-based fuzzy-decentralized-information-filtering algorithm, the decentralized-information-filtering algorithm is presented, which can conveniently isolate a Global Positioning System fault and fuse multiple observation sources with different sample rates, and the fuzzy inference module is designed and introduced to adjust model probabilities adaptively to make them more suitable for actual driving situations. Finally, the proposed strategy is evaluated through both simulations and experiments. The results validate the feasibility and the effectiveness of the proposed strategy.