System Modeling And Versatile Supervisor Design Based On States Control In Discrete Event Systems

For the convenience of system analysis and synthesis,a real-word system is usually ab-stracted based on some models.Discrete event systems(DES)are a kind of event-driven mathematical models made for some physical systems.With the development of informa-tion technologies,especially the manufacturing technologies of digital computers,more and more DES are stepping into peoples'daily life,such as communication networks,flexible manufacturing systems,logistic management,train or airplane scheduling systems,etc.The systems are typically event-driven,and the events are usually asynchronous,sequential,con-current,conflicting,non-deterministic,deadlocked,and mutually exclusive.These charac-teristics determine the complexity of these DES.To investigate the laws of system behavior and enforce some necessary control actions,it is necessary to study modeling techniques for DES,analysis methods of system behavior,and the corresponding control policies.This thesis employs Petri nets as the modeling tools and obtains the following results:1.A method is proposed to design a versatile supervisor with one monitor.Taking Petri net models as objects of the study,this method finds the sets of critical transitions and the corresponding dangerous markings by the analysis of a reachable graph.Systems are kept out of illegal state spaces by adding guard functions to the critical transitions based on the information obtained.The simulation results show that the method can effectively reduce the number of supervisors and also keeps the supervisor with the same structure.2.Based on the theory of regions,a method is proposed to reduce the cost of computing the sets of dangerous markings.Based on the definition of deadlock markings and P-invariants(a structure property of Petri nets),this method designs a Mixed Integer Programming(MIP)algorithm to find the sets of deadlock markings.Then the sets of dangerous markings and corresponding critical transitions are computed by state equations.Thus,the cost of com-puting dangerous markings is reduced since the bad markings and dangerous markings are deduced from deadlock markings instead of enumerating all reachable states.3.A model is made to solve Rubik's cube.Based on state control,this model imitates the corner-first solving process of human beings.Instead of using human beings' formulas,Macro-transitions are introduced to solve Rubik's cubes.The simulation results show that the model can solve Rubik's cube like human beings.4.A method is reported to design models of DES.By adding constraints during the process of modeling,this method aims to build models with lower behavior complexity,which makes it easy to analyze system behavior and give the corresponding control laws.This method is used to derive a supervisor for a special class of DES:the Missionaries and Cannibals problem.During the process of modeling,state-based control constraints are considered to forbid some boating combinations,which leads to a reasonable approach to across the lake without any manual intervention.This model explains that it is feasible to reduce system behavior complexity by adding state-based constraints during the process of modeling for DES.5.As confusions have impact on the multi-thread running and decision-making of flexible manufacturing systems(FMS),this thesis aims to present a novel algorithm to find all confu-sions of such an FMS,which can be considered as a foundation for further system modeling and control.An algebraic method is presented by analyzing the characteristics of confusion states,and build on the available detection schemes of subnet structure of Petri nets.The method implements state estimation of confusions by inputting a system static structure and an initial state instead of supervising system states.This result can be applied to conflict rationality analysis in the system modeling phase.Finally,an example of FMS is employed to simulate the proposed algorithm.The result shows that all confusion states of the target system can be estimated effectively by using the method.