Abstract
This paper presents some tools based on interval analysis for guaranteed non-linear parameter and state estimation in a bounded-error context. These tools make it possible to compute outer (and sometimes inner) approximations of the set of all parameter or state vectors that are consistent with the model structure, measurements and noise bounds.
Notes
1Such a situation may even occur in the linear case, when the perturbation or measurement-noise sets are not convex or consist of disconnected parts.