This paper addresses the problem of preserving stability of a pole placement direct adaptive controller in spite of output bounded disturbances, time varying plant model parameters and unmodeled dynamics. Two fundamental contributions are to be emphasized.
1) The class of the plants to be controlled allows for plant model parameter variations together with unmodeled dynamics to be small in the mean rather than at each time. More importantly the plant model controllability should hold only in the mean.
2) The controller parameter estimates are shown to track, in the mean, their true (time-varying) parameter values, (avoiding in particular, as well as known, the possibility of any chaotic phenomena). Such a convergence property is achieved using an adhoc, internally generated, excitation sequence.
This leads, besides robust global (BIBO) stability of the adaptive controller in question, to an improved understanding of the persistent excitation concept in non-ideal situations. Moreover new results are given, concerning the ideal case, where the considered plant is completely described by a disturbance free linear time invariant system of known structure.