Model-Based Engineering of Supervisory Controllers using CIF

Abstract

In the Model-Based Engineering (MBE) paradigm, models are the core elements in the design process of a system from its requirements to the actual implementation of the system. By means of Supervisory Control Theory (SCT), supervisory controllers (supervi- sors) can be synthesized instead of designing them manually. In this paper, a framework based on the Compositional Interchange Format for hybrid systems (CIF) has been developed that integrates the MBE and the SCT paradigms. To illustrate the framework, an industrial-size case study has been performed: synthesis of a supervisory controller for the patient support system of an MRI scanner. In this case study, we address 1) modelling of the components and the control requirements; 2) synthesis of the supervisor; 3) simulation of the synthesized supervisor and a hybrid model of the plant; and 4) real-time, simulation based control of the supervisor and the actual patient support system of the MRI scanner. Complex manufacturing machines consist of physical components (hardware) and control systems. The physical components, typically sensors, actuators and main structure, provide the means of the machine. The interactions between the physical components result in the so-called uncontrolled behavior of the machine. The control systems interact with the sensors and actuators to employ the means of the machine, which results in the controlled behavior of the machine. The controlled behavior should be such that the machine fulfills its functions, i.e. meets its pre-defined requirements. The control systems can be divided into five functional subsystems, see (PFC89): 1) Regulative control (also known as direct or feedback control) that assures that the actuators reach the desired position in the desired way. 2) Error- handling control (also known as fault detection and isolation or exception handling) that detects erroneous behavior, determines the cause, and acts to recover the machine control system. 3) Supervisory control (also known as logic control) that coordinates the control of the individual machine components. This includes planning, scheduling and dispatching functions. 4) The data processing subsystem that stores and manipulates gathered data. 5) The user interface subsystem that allows the user to interact with the machine control system. In this paper, we focus on the development process of supervisory controllers (supervisors). The current practice of developing supervisory controllers is to code them manually, based on (possibly informal) control requirements. Creating and changing requirements, a design and/or an implementation can be time consuming and error-prone. An other possibility is to use the Model-Based Engineering (MBE) paradigm, see (Ogr00, Bra08), in order to design the supervisory controller. In this case, (pos- sibly formal) executable models for the supervisory controller are developed (by hand). Using analysis † This work was partially done as part of the Darwin project under the responsibility of the Embedded Systems Institute, partially supported by the Netherlands Ministry of Economic Affairs under the BSIK program, as part of the ITEA project Twins 05004, and as part of the Collaborative Project MULTIFORM, contract number FP7-ICT-224249.