Hierarchical Hybrid Planning for a Self-Optimizing Active Suspension System

This paper presents a hierarchical hyprid planning approach developed for realizing self-optimizing mechatronic systems. The hybrid planning approach is based on a discrete plan and a forecast of continuous system behavior generated by simulation during runtime. This does not only allow an online adaptation of a previously generated plan with regard to the actual system state as well as the current environmental conditions, even the planning objectives may be adapted to newly arising needs during system operation. The planner exploits the hierarchy already present in the system model by considering hierarchical parameterizations of the system as a discrete dimension of choice. The parameterizations are proposed by a novel hierarchical multiobjective optimization, which calculates Pareto points in a bottom up fashion, taking into account the constraints imposed by lower levels of the system hierarchy on the higher ones. Evaluations of the hierarchical hybrid planner in the context of an innovative railbound transport system show that it outperforms its non-hierarchical predecessor.