Wrong Association Risk Bounding Using Innovation Projections for Landmark-Based LiDAR/Inertial Localization

In this paper, we derive, analyze and test a new data association method for integrity risk bounding of LiDAR/inertial localization in ground transportation applications. Data association is the process of assigning currently-sensed features with ones that were previously observed. Most data association methods use a nearest-neighbor criterion based on the normalized innovation squared (NIS). They require complex algorithms to evaluate the risk of wrong associations (WA). In contrast, in this research, we derive a new DA criterion using projections of the extended Kalman filter’s innovation vector. The paper shows that innovation projections (IP) are signed quantities that not only capture the impact of WA in terms of their magnitudes, but also of their directions, which are predictable. The paper also shows that sample innovation vectors should be projected on a line joining the origin of the innovation space to the centroid of a convex polytope formed by predicted innovation vectors under a set of WA hypotheses. As compared to NIS, the IP method both reduces the actual risk of WA and improves our ability to bound the risk of WA. We first analyze the new IP method using simulated data. We then evaluate the inertial-aided LiDAR localization integrity performance improvement of IP over NIS using experimental data in an outdoor driving environment.