A Survey on Automata-based Methods for Modelling and Simulation of Industrial Systems Vasileios Deligiannis University of Patras Electrical and Computer Engineering Dept. Division of Systems and Control vdeligiannis@ece.upatras.gr Stamatis Manesis University of Patras Electrical and Computer Engineering Dept. Division of Systems and Control stam.manesis@ece.upatras.gr Abstract The importance of modelling and simulation for the design of high quality industrial systems in a limited amount of time is generally acknowledged. Studying industrial systems by simulation enables the designer to study their dynamic behaviour and determine characteristics of the system. Due to the increasing complexity of industrial production systems, there exists a need for the development of formal approaches for their analysis and control. While there has been intensive research effort concerning theoretical aspects on control of discrete event systems, less work has been reported on practical implementation techniques with general acceptance. This paper focuses on Automata forms of modelling considered as the primary representation scheme of industrial plants. It also presents an overview of software tools with both educational and commercial orientation, necessary on the way to real applications in modern industry. 1. Introduction An industrial production line generally consists of various types of devices (e.g. robots, NC machines, actuators, sensors, etc.) under the control of either centralized supervisor or decentralized controllers capable of performing a specific set of operational functions. From a planning and control perspective, an industrial production line can be seen as a dynamic system whose states evolve according to the occurrence of abrupt physical events, thus exhibiting the characteristics of a discrete-event system (DES). In another point of view, systems of industrial relevance are governed by discrete state digital controllers, whose internal state transitions are triggered by the value of some measured continuous physical quantity (temperature, speed, fluid rate, etc.). Such systems, which are described by both discrete and continuous variables, are usually referred to as hybrid systems or more accurate as hybrid control systems [1, 2]. In the past, automated-manufacturing DESs have usually been sufficiently simple that intuitive or ad-hoc control solutions have been adequate [3]. However, the increasing complexity of these systems and the requirement of fast system response have created a need for formal approaches for their analysis, modelling and control [4, 5]. Formal methods allow a thorough analysis of the possible behaviours of a system and rigid proving of system properties in verification and validation. The modern system theory as developed by Kalman et al. in 1960s [6] made it clear that state space approach was directly related to automata theory. The next twenty years control theory met great success in many fields, especially by putting man on the moon, using continuous and more even “linear” methods. But, after the digital revolution and the new control technologies based on computers, control theorists were made to reconsider the relation between finite automata and dynamical systems. It was then that it came the need of introducing a new concept that is the discrete events dynamical systems. In the past two decades, advances in the control of automated manufacturing systems have evolved along two separate paths. Significant developments reported on the control of industrial production procedures have dealt primarily with the synthesis of supervisory-control laws, which generate correct control strategies within a formal theoretical framework. Control-theory-based modelling techniques have included Petri nets [7], real- time temporal logic [8, 9], and controlled-automata [10, 11, 12, 13]. Furthermore, in parallel, automation- hardware developers have brought to market PLC-type controllers with advanced device-interface capabilities and fast CPUs. In contrast with scientists’ expectations, Petri nets did not meet success when proposed for closed loop control, primarily because there was not a clear separation of plant and controller. On the contrary, automata [10, 11, 12, 13, 14] seem to be the right tool for solving this control problem. In general, an automaton is an intuitive and natural description of a DES, which reconstructs the system in a schematic way. Finite automata [13], are probably the simplest mathematical abstractions for modelling DESs. Finite automata model DESs with a finite number of states and transitions between those states. An extension of this classical notion, which models the coupled interaction of discrete events and continuous dynamical systems, are hybrid automata. Hybrid automata [13, 14, 15, 16, 17] have been proposed as a formal method for modelling hybrid systems and probably are the most well studied type of controlled automata. Another super-set of finite 1-4244-0826-1/07/$20.00 © 2007 IEEE 398