Font Size: a A A

Supervisory Control Of Discrete-event Systems Based On Symbolic Computation

Posted on:2020-03-16Degree:DoctorType:Dissertation
Country:ChinaCandidate:D G WangFull Text:PDF
GTID:1368330602963893Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
In the last three decades,discrete-event systems(DES)have been studied by researchers and engineers from different fields,with respect to modeling,scheduling,control,diagnosis,and opacity.When designing a controller for a DES,model-based approaches can conveniently describe a system's behavior.A well-known framework of such an approach is supervisory control theory(SCT),which provides a powerful and unified formalism for automatically synthesizing a supervisor to ensure that a system never violates the given specifications However,a main obstacle that SCT faces is the state explosion problem,i.e.,the state size increases exponentially with the number of components of a system.As known,an optimal nonblocking supervisory control problem for a DES is in general NP-hard.In order to man-age computational complexity,Ma and Wonham develop state-tree structures(STS),which have been shown to be computationally efficient for monolithic or centralized supervisor synthesis in the case of large-scale systems.A complete symbolic approach based on predi-cates and binary decision diagrams(BDD)for the optimal nonblocking supervisor design of DES is developed,and a neat control mechanism is implementedThis dissertation focuses on the supervisory control of two classical topics in the area of DES:partial observation and priority.Due to limitations on detection and communication,not all events are observable.In this case,the supervisory control and observation problem(SCOP)needs to be considered and investigated.To do that efficiently,a state-based for-malism that can to be carried forward to STS implementation directly is urgently needed The notion of priority exists in a wide range of DES,such as manufacturing systems and traffic systems,and priority plays an important role in affecting and improving the behavior of a DES.Therefore,a unified framework for efficiently solving various issues(industrial process,emergency handling,express delivery,etc)related to priority is urgently needed For the above mentioned topics,the symbolic approach is utilized for efficiently designing a nonblocking and optimal supervisor such that the notorious state explosion problem is alleviated.The main results are listed as follows1.A state-based approach for supervisor synthesis of DES under partial observation is proposed,which is based on predicates and predicate transformers.We focus on the normality property and provide an iterative algorithm for state-based normality syn-thesis.A condition is provided to simplify the algorithm.To bridge the gap between language-based normality synthesis and state-based normality synthesis,we prove that their synthesis results are mutually consistent.So far,there is no available result on supervisor synthesis of DES under partial observation based on STS.One of the rea-sons for that is the lack of a state-based approach that can be carried forward to the context of STS.Thus,it is important to bridge this gap to extend the applicability of STS.This work provides a theoretical guarantee for the correctness of non-blocking supervisor synthesis for partially-observed DES based on STS2.A conditional-preemption matrix is proposed to describe the preemption relations a-mong events,which can represent the priority intuitively.Practically,it is a hard task to describe priority-based specifications using formal languages or automata,especial-ly for beginners,while conditional-preemption matrices make such a task easier and more user-friendly.As known,the model construction is time-consuming and error-prone.By using the conditional-preemption matrices,the mentioned two issues can be well solved.3.In order to describe priority among events intuitively in a DES setting and,subse-quently,to efficiently compute a supervisor,a formalism,i.e.,state-tree structures with conditional-preemption matrices(STSM),is proposed,which aims to solve different kinds of issues related to priority.The optimal nonblocking supervisory control is de-ployed in the framework of STSM.Symbolic synthesis algorithms are implemented and integrated in a software tool STSLib that exploits binary decision diagrams as a basis for efficient computation.
Keywords/Search Tags:Discrete-Event Systems, Supervisory Control, Partial Observation, State-Tree Structures, Conditional-Preemption Matrices
PDF Full Text Request
Related items