Sensor Fusion in Case of Aliquant Sample Repetition Frequiencies of Measurements and Output Estimates

Abstract

The problem of measurement fusion from two sensors with no coincident pairs of measurement at any time is considered. Besides, the required moments of returning the result also lie between indication times. Several association algorithms based on Kalman filtering both with use of the centralized filter, and on the basis of two parallel local filters as a part of decentralized filtration approach are proposed. Filter parameter adjustment to operation at non-equidistant times is described. The options for combining the assessments of the process state inside the filtration loop and outside it are considered. Comparison of methods is carried out in terms of the relative mean square error value of result estimate. The case of uniformly precise measurements and a case when the noise level into indications of sensors considerably differs are separately analyzed. The operability of methods is investigated at various values of bandwidth determined by a ratio of process noise intensity and the measurement noise one.

Keywords

• kalman filter
• sensor fusion
• non-equally accurate measurement
• centralized filter
• local filters
• inertialess average
• process noise
• measurement noise