A maneuvering multi-sensor information fusion algorithm for enhancing localization reliability in ADAS testing

A novel algorithm, the maneuvering first-order generalized pseudo-Bayesian error-state Kalman filter (MGPB1-ESKF), is proposed in this study for localization in advanced driver assistance system (ADAS) testing platform vehicles. The proposed algorithm substantially enhances the reliability and accuracy of multi-sensor localization in ADAS testing. This improvement is particularly significant in challenging non-line-of-sight (NLOS) environments, which typically degrade the performance of global navigation satellite system (GNSS). By adaptively identifying and isolating faulty signal sources, the MGPB1-ESKF provides a novel approach to achieving robust and reliable localization in the presence of transient sensor failures. Rigorous simulations and experimental results demonstrate that the proposed algorithm effectively mitigates the impact of faulty sensors in challenging environments, outperforming conventional multiple model (MM) algorithm and leading to improved localization accuracy.