Robust penalty adaptive model predictive control (PAMPC) of constrained, underdamped, noncollocated systems

Abstract

This paper investigates the control challenges posed by noncollocated mechatronic systems and motivates the need for a model-based control technique towards such systems. A novel way of online constraint handling by penalty adaptation (PAMPC) is proposed and shown to be of particular relevance towards robust control of underdamped, noncollocated systems by exploiting the structure of such systems. Further, a new tunneling approach is proposed for PAMPC to maintain feasibility under uncertainty. The PAMPC is shown to be optimal for control of a benchmark mass–spring–damper system, which poses all the mentioned challenges.