Computationally Distributed Real-Time Dual Rate Kalman Filter

Many systems include sensors with large measurement delays that must be fused in a Kalman filter in real time. Often, the filter state must be propagated at a higher rate than the rate at which measurements are taken. This can lead to a significant amount of unused CPU time during the time steps in which no measurements are available. This paper presents a method of fusing delayed measurements for a restricted set of systems, which more efficiently uses processing resources at the expense of data availability. The new method splits the filter into a high-rate and a low-rate task running in parallel. The high-rate task propagates the whole states, and the low-rate task propagates and updates an error state filter, which can be distributed over several high-rate periods.