Reliable Positioning Domain Computation for Urban Navigation

Reliable positioning is a key issue for intelligent vehicle navigation. Interval-based positioning methods have shown to be capable of computing relevant confidence domains used for integrity monitoring in environments which are challenging for Global Positioning System (GPS). The approach presented in this paper consists in tightly coupling a GPS receiver with a 3D-map of the drivable area. Interval analysis is employed to solve the constraint positioning problem using contractions and bisections. Integrity is provided through the use of a robust set inversion scheme applied to a redundant measurement set. If the prior distribution of the measurement noise is known, it is possible to compute confidence domains that correspond to a given integrity risk, which is often set very low out of safety considerations. In this paper we examine a way of validating the proposed approach, using a real experimental dataset and a ground truth equipment. Different tunings of the method, corresponding to different risks, are assessed in terms of availability and integrity in order to compute statistical metrics. Results indicate that this methodology is relevant since the specified risk corresponds to experimental observations.