Vehicular Relative Positioning With Measurement Outliers and GNSS Outages

Relative positioning among vehicles finds many applications regarding location-based services in intelligent transportation systems (ITSs). This article introduces a hybrid relative positioning strategy integrating global navigation satellite system (GNSS), inertial navigation system (INS), and ultrawideband (UWB) observations, which addresses measurement outliers and GNSS outages simultaneously. An improved extended Kalman filter (IEKF) based on tracking an appropriately weighted window of past innovations is developed to deal with measurement outliers. To perform relative positioning during GNSS outages, a nonlinear autoregressive network with exogenous inputs (NARX) is developed to predict GNSS measurements increments, with the least-squares support vector machine (LSSVM) algorithm identifying the NARX model. The NARX-LSSVM module learns the relationship between GNSS pseudorange and Doppler shift increments of a target vehicle and its local and neighbors' dynamics. Whereas during GNSS outages, the increments of GNSS measurements are generated by this prediction algorithm. The feasibility of the proposed methodology is evaluated on empirical road data, and improved positioning accuracy is achieved in cooperative positioning (CP) systems.