A "Hybrid" Approach for Synthesizing Optimal Controllers of Hybrid Systems: A Case Study of the Oil Pump Industrial Example

TL;DR

This paper introduces a hybrid method combining quantifier elimination and numerical computation to synthesize optimal controllers for hybrid systems, demonstrated on an industrial oil pump example, achieving up to 7.5% improvement over previous methods.

Contribution

It presents a novel hybrid approach that reduces the complexity of controller synthesis for hybrid systems and improves controller optimality using industrial case studies.

Findings

Achieved up to 7.5% better results than previous methods.

Combined quantifier elimination with numerical methods for scalability.

Validated approach on real industrial oil pump system.

Abstract

In this paper, we propose an approach to reduce the optimal controller synthesis problem of hybrid systems to quantifier elimination; furthermore, we also show how to combine quantifier elimination with numerical computation in order to make it more scalable but at the same time, keep arising errors due to discretization manageable and within bounds. A major advantage of our approach is not only that it avoids errors due to numerical computation, but it also gives a better optimal controller. In order to illustrate our approach, we use the real industrial example of an oil pump provided by the German company HYDAC within the European project Quasimodo as a case study throughout this paper, and show that our method improves (up to 7.5%) the results reported in [3] based on game theory and model checking.