Modeling Quantitative Effects for the Reconfiguration of Hybrid Systems

Reconfiguration is the task of recovering a valid system state after an error has occurred, which led to an invalid system state. Especially for hybrid systems, identifying the necessary changes to restore valid system functioning is challenging: Hybrid systems contain continuous and discrete variables, leading to an infinite search space which, in addition, suffers from combinatorial explosion. Existing approaches to the reconfiguration problem mostly require a pre-definition of faults and a large amount of expert knowledge and thus, enable the system to adapt only to known faults. This paper presents a novel approach which does not need a pre-definition of faults such that the system is enabled to adapt even to unknown faults. It works on an representation of the reconfiguration problem in a logical calculus. Therefore, the hybrid system is modeled in first-order logic. To integrate continuous variables, which have infinite domains, they are discretized using intervals. The approach is shown to reconfigure faults on simulated systems from process engineering. This way, the reconfiguration problem of hybrid systems can be modeled and solved efficiently.