Abstract
State estimate in general multimodal posteriors amounts to maximum-a-posteriori estimation, but estimating the error associated with it is non-trivial. Real-world problems that deal with significantly multimodal posteriors either maintain multiple hypotheses and prune branches after gathering sufficient evidence, or fuse estimates obtained from multiple sources to assess confidence in the estimation. Many applications require good estimate of error leading to faster convergence on state estimate. This paper makes two significant contributions for multisource Bayesian tracking problems. First, it derives a computationally light method for estimating the maximum-a-posteriori state, and second, it proposes a novel error estimator for faster convergence. The properties are demonstrated using application of terrain-aided navigation that fuses data from inertial navigation and altitude sensors.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call