Nonstandard problems of robust identification of regression models

Some new formulations of the problems of robust identification when the level of prior information is low, are given and, for some of them, procedures which guarantee perfect identification are constructed.