METHOD FOR SENSOR FUSION AND FILTERING BASED ON CONSTRAINT MANIFOLD

摘要

A sensor fusion/filtering method for knowing a prediction value and a distribution value of each variable based on a constraint manifold is provided to apply to diverse constraints by applying a method for finding a joint probability on the constraint manifold defined on a unified fusion/filtering space, and apply to diverse models by directly handling the geometrically expressed constraint. After defining the sensor fusion/filtering space(S1), the constraint manifold including the diverse constraints is defined on a multidimensional space comprising the variables for a measured value(S2). Joint probability distribution on the constraint manifold is operated by using the priori probability distribution of the variables(S3). Marginal probability distribution of the variables is operated through back-propagation to each axis of the variables from the joint probability distribution(S4). The prediction value and the distribution value of the variables are obtained(S5).