Control of Pareto Points for Self-Optimizing Systems with Limited Objective Values

Abstract Self-optimization enables technical systems to adapt their behavior to varying environmental conditions and changing system settings. Objective functions serve as evaluation criteria for the system behavior. In this paper we propose a hierarchical control approach of an objective-based Pareto controller. We separate the processes of optimization and control. First, we use multiobjective optimization to compute a Pareto set of optimal system configurations offline. This set serves as a data base for the Pareto controller in an upper control loop, which is designed secondly. The goal of the Pareto controller is to drive the system toward a desired relative weighting of the objective values, despite unknown and varying environmental disturbances. Furthermore, the Pareto controller has to cope with limits of the objective values. For that, we propose a calculation of a reference value, which is based on an approximated Pareto front of the current situation. The Pareto controller selects suitable configurations out of the Pareto set and applies them to a lower control loop. A test rig of an active suspension system affected by unknown track excitations serves as application example. Finally, we give some results with the test rig that validate our approach and point out the advantages.