PARALLELIZED SIGMA POINT AND PARTICLE FILTERS FOR NAVIGATION PROBLEMS

摘要

Advanced filters like the sigma point and particle filters are more accurate than the extendedKalman filter for nonlinear and non-Gaussian applications, but experience drawbacks such asbeing computationally expensive with a serial implementation. However, since the majorityof the computation can be carried out simultaneously, these filters are inherently well suitedfor parallel computing. This research leverages inexpensive and personal-level parallel computingarchitectures, such as the NVIDIA Graphics Processing Units (GPUs) and multi-coreCPUs to exploit such parallelism. In particular, parallelization of the Unscented Kalman filter(UKF) and the bootstrap Particle Filter (PF) applied to an orbit determination problemand a loosely coupled GPS/INS integration problem is the main objective of this work. Thiswork demonstrates that the UKF and the PF can be applied to the two navigation problemswith great benefits in the presence of nonlinearities and non-Gaussian noises. An 8-timespeedup is achieved for both filters with an 8-thread CPU, and up to two orders of magnitudespeedups are achieved using a M2090 GPU. The results show that the two UKF implementationscan be executed in real time without parallelization, but the two PF implementationscan be executed in real time only when parallelized on a GPU.