J96_The Real-time Supervisory Control of an Experimental Manufacturing Cell

8/2/2019 J96_The Real-time Supervisory Control of an Experimental Manufacturing Cell

1/14

IEEE TRANSACTIONS ON ROBOTICS AN D AUTOMATION, VOL. 12, NO. 1, FEBRUARY 1996 1

The Real-Time Supervisory Control of anExperimental Manufacturing CellBertil A. Brandin, Member, ZEEE

Abstruct- In the last decade the manufacturing environmenthas changed dramatically. In the past, manufacturing systemswere sufficiently simple, in most cases, to permit the use ofintuitive and informal solutions in the developmentof supervisorycontrol systems. The increasing level of automation, integration,and flexibility encountered in automated manufacturing systemsrenders formal approaches to the supervisory control system de-velopment a necessity. The controlled-automata based approachto supervisory control development considered in this work isone such approach. It offers important advantages over other ap-proaches. It guarantees that: i) the resulting controlled behaviorsdo not contradict the behavioral specifications and are nonblock-ing and ii) the controlled behaviors are maximally permissivewithin the behavioral specifications. The work presented illus-trates the applicabilityof the method to the supervisory control ofindustrial automated manufacturing systems. The paper providesan overviewof the automated manufacturing field, and introducessupervisory-control-system development. The main concepts areillustrated through an exampleof a small manufacturing cell. Thedevelopment of a supervisory control system for an experimentalmanufacturing cell is presented subsequently. Both centralizedand modular supervision are considered. The automata modelsfor the complete cell are provided together with the correspondingsupervisors. The implementation of the resulting supervisorycontrol system only required the use of off-the-shelf industrialcontrol systems.

I. INTRODUCTIONA. Automated Manufacturing Systems

UTOMATED manufacturing systems are generally com-A osed of a number of interconnected material processingstations, a material transport system, a communication system,and a supervisory control system. The flexible manufacturingconcept advocates that manufacturing operations be carriedout within workcells, each workcell being responsible for theproduction of a specific part family. Cells are usually intercon-nected by a transport system. Although workcell configurationsmay vary, these typically incorporate the following systems:numerically controlled (NC) part processing machines, ma-terial handling devices, part inspectiodtesting devices, in-process part storage systems, and a supervisory control system.The latter performs the following three tasks: supervisorycontrol, communication, and housekeeping. Supervisory con-trol consists of i) the monitoring of the workcell behaviorvia sensory feedback; ii) control evaluation (determination)in accordance with a supervisory control law which maps

Manuscript received November 3, 1992; revised February 14, 1995.The author is with the Department of Electrical and Co mputer Engineering,Publisher Item Identifier S 1042-296X(96)00969-4.University of Toronto, Toronto, Ontario M5S 1A4 Canada.

the workcell behavior to corresponding controls; and iii)control enforcement via the downloading and execution of theappropriate device programs. Communication allows sensoryfeedback and control enforcement to be performed. House-keeping is the set of tasks related to supervisory control andcommunication which are necessary to their implementation,e.g., data-base management. Supervisory control system devel-opment consists in the procedure generating supervisors andcontrol laws which specify how the supervisory control systemis to react to the manufacturing system behavior in order tosatisfy given behavioral specifications (routing, sequencing,safety, etc.). To develop supervisory control systems, differenttechniques such as knowledge engineering 1121, [15]-[171,[21], [23]-[25], [29], [36], [42]-[44], [53], Petri nets [14],[18], [19], [28], [45], and controlled automata [l], [4]-[8],[27], [34], [38], [50], [51], [59] may be exploited. The workpresented is based on controlled automata, since these provideimportant advantages over other approaches, as is describedin Section III.B. The Supervisory Control of Discrete-Event Systems

Discrete event systems (DES) are dynamic systems thatevolve in accordance with the abrupt occurrence of events.They are generally asynchronous (not clock driven) and non-deterministic (some events may occur spontaneously). Suchsystems are encountered in a variety of fields, for exam-ple in manufacturing, robotics, computer and communicationnetworks, traffic, and logistics.The supervisory control of DES in accordance with behav-ioral specification is a new research area which is receivingincreasing recognition. The approach is based on informationfeedback on the occurrence of events and controlled automataconcepts. It offers two important advantages over other ap-proaches:

the resulting controlled behaviors do not contradict thebehavioral specifications and are nonblocking: the super-visors and control laws obtained are correct by construc-tion; andthe controlled behaviors are maximally permissive withinthe behavioral specifications: all events which do notcontradict the speciJcations are allowed to happen.

11. AUTOMATED ANUFACTURINGYSTEMSThe manufacturing environment has evolved from intensive

manual operation, where labor operated individual machines,to semiautomatic operation, where the machines were able to

1042-296W96$05.00 0 1996 IEEE


2/14

2 E E E TRANSACTTONS ON ROBOTICS AN D AUTOMATION,VOL. 12, NO . 1, FEBRUARY 1996

perform a few steps in automatic sequence, to a high degree ofautomation making extensive use of computers and automatedequipment [48]. The term automated manufacturing is usedhere when referi-ing to a manufacturing facility with a highdegree of programmable automation.An automated manufacturing system generally consists of anumber of interkonnected material processing stations capableof processing a wide variety of part types, a material transportsystem, a communication system for integrating all aspects ofmanufacturing, and a supervisory control system. Automatedmanufacturing systems generally exhibit a high degree of au-tomation, integration, and flexibility; the latter takes a numberof forms, including [48]:

volume flexibility, i.e., the ability to handle changes inthe production volume;* routing flexibility, i.e., the ability to route parts throughthe system in a dynamic fashion taking into account

machine breakdowns, required tooling, etc.; and* product flexibility, i.e., the ability to handle requests

for a wide variety of products, including the abilityto reconfigure the system to handle the production ofdifferent products.The latter may also involve various off-line supporttechnologies, including computer-aided design (CAD), andcomputer-aided process planning (CAPP), as well as on-linetechnologies such as robotics for processing and handlingmaterials, and computer aided testing (CAT).

A. Elements o Automated Manu facturing Systems1 ) Machining, Assembly, and Transportation Devices:

9 Numerically C ontroIledKomputerized Numerically Con-trolled (N C/C NC) Machines: The main types of CNCmachines are lathes and milling machines.

0 Industrial Robots: Robots are useful in a wide varietyof applications such as spray painting, spot welding, arcwelding, as well as material handling, assembly, andinspection.Material Transpo rt Systems (TransferSystems): Materialsneed to be transported bgtween manufacturing, assembly,and testing stations, etc. Typically, a variety of conveyorsystems and automated guided vehicles (AGV) are usedin addition to industrial robots.

2) Sensors and C ommunication Devices:* In most cases, sensors are an integral part of machines

(internal) but they may sometimes perform their functionindependently (external). Bar code readers and visionsystems are two such examples. Sensors, in generalexternal, play an important role in the supervisory controlcontext, since they provide the information required tocarry out closed-loop control.

* Communication systems are used in virtually every aspectof automated manufacturing. They provide, together withthe data storage and control systems introduced below,the basis for integration. Communication systems aregenerally constituted in the manufacturing kontext byinterconnected computer networks.

3) Data Storage and Control Systems: Data storage andcontrol systems provide, together with communicationsystems, the basis for integration. They are classified intothe following three categories:Microprocessor-Based Programm able Con trollers, whichare specialized devices used to control the operation ofmachines or processes by means of stored programs andfeedback from input/output devices.Microcomputer-Based and Minicomputer-Based Con-trollers, which are either embedded into the manufactur-ing devices, or may work as standalone units performingmonitoring and control functions. In automated man-ufacturing systems, production planning and controlare possible because of the availability of such highperformance microcomputers and minicomputers.Mainframe Computers, which are generally used to per-form high-level control tasks and to manage the variousdatabases for entire manufacturing facilities.

B. WorkcellsAutomated manufacturing systems are generally organizedinto workcells according to the flexible manufacturing concept

2241, 35] which advocates that manufacturing operations becarried out within workcells, each workcell being responsiblefor the production of a specific part family. Workcells are usu-ally interconnected by a transport system. Although workcellconfigurations vary, these typically incorporate the followingsystems:

* numerically controlled (NC) part processing machines;* material handling devices;0 part inspectiodtesting devices;* in-process part storage systems; anda workcell supervisory control system constituted bya computer system and corresponding communicationnetwork(s).Depending upon the decision-making ability and upon the

local memory capacity requirements, computer systems mayrange from small, microprocessor-based systems to sophisti-cated computers with powerful operating systems and pro-gramming languages. Depending on the workcell configurationrequirements, various types of communication networks maybe considered.Cell size and complexity may vary, with on the one handsmall workcells having very simple control systems incorpo-rating just a few components, and on the other hand, workcellssuch as that of the automated manufacturing research facilityat the National Bureau o f Standards, which considers anentire manufacturing facility to be a dynamic arrangement ofvirtual manufacturing workcells [34] assumed to be highlyautomated and flexible.

C. Su pervisory ControlSupervisory control systems carry out the following threetasks: supervisory control, communication, and housekeeping[4]. Supervisory control consists of


3/14

BRANDIN REAL-TIME SUPERVISORY CONTROL OF AN EXPERIMENT& MANUFACTURING CELL

Modelling

3

plant model++ pecification models + Synthesis

the monitoring of the system behavior via sensory feed-back;control evaluation in accordance with a supervisor andthe corresponding supervisory control law which mapsthe workcell behavior to corresponding controls; andcontrol enforcement via the downloading and executionof the appropriate device programs.

Communication allows sensory feedback and control en-forcement to be performed. Housekeeping is the set of tasksrelated to supervisory control and communication which arenecessary to their implementation, e.g., data-base manage-ment. Supervisory control system development Consists inthe procedure generating supervisors and control laws wliichspecify how the supervisory control system is to react tothe manufacturing system behavior in order to satisfy givenbehavioral specifications.

The advantage of considering a controlled automata basedapproach over other approaches lies mainly in the existence oftechniques [SI, [9], [5 11 which guarantee that the supervisorsand control laws obtained are correct by construction, ensuringthat the resulting controlled behaviors are guaranteed not toviolate the specifications considered; and that the supervisorsand control laws obtained are maximally permissive within thebehavioral specifications considered, allowing all occurrenceswhich do not contradict the specifications. Furthermore, con-trolled automata provide excellent modeling tools for systemssuch as manufacturing workcells. The models obtained arerealistic enough to capture the system features required forcontrol purposes, and yet are simple to visualize and use.Device integration and programming are important issues inthe supervisory control of automated manufacturing systems,becauseof the generally heterogeneous nature of such systems.The various machines and devices often support a varietyof programming languages, and the corresponding sensoryoutputs are usually expressed in incompatible terms, thusmaking device integration a difficult task. A successful andrecognized approach to this issue consists of

the development of high-level device programming lan-guages and related device customized programming prim-itives, which allow automated manufacturing system de-vices to be programmed on a uniform basis 121, 141;andthe translationof sensory outputs into correspondmg high-level language statements [4].111. SUPERVISORYONTROLYSTEM EVELOPMENT

The supervisory control of DES in accordance with behav-ioral specification is a new research area which is receivingincreasing recognition. The approach is based on informationfeedback on the occurrence of events (cf., Fig. 1) and con-trolled automata concepts. It offers two important advantagesover other design approaches:the supervisors and control laws obtained are correct byconstruction; andthe supervisors and control laws obtained are maximallypermissive within the behavioral specifications consid-ered.

Supervisory control system

EventsSupetvisor+control law

-Plant

Fig. 1. Feedback control.

4SUpervlsor(S)+control law@)upervisory control t- mplementation c-

Fig. 2. Supervisory control system development.

As shown in Fig. 2 , the development of a supervisoryThe Modeling of the Plant and Behavioral Spec$cationsto be Enforced: Both the plant physical behavior and thebehavioral specifications are translated in the form ofautomata.Supervisor and Control La w Synthesis: Taking into ac-count the supervisory control system architecture adopted(for example centralized or modular), the automata rep-resenting the physical behavior of the system to be con-trolled and the automata representing the correspondingspecifications are fed to a computer program which will:i) tell us whether it is possible for the system to behavewithin the specifications and ii) return the supervisor(s)embodying the maxiknally permissive controlled behaviorof the system within the latter specifications, and thecorresponding control law(s) yielding such behavior.Implementation: The supervisor(s) and the correspondingcontrol law(s) obtained are coded in a ProgrammableLogic Controller (PLC) or computer based control systemwhich will carry out feedback-based control; the supervi-sory control system reacts to event occurrences accordingto the supervisor(s) and corresponding control law(s).

Generally, specificationsencompass a set of nominal behav-ioral specifications, e.g., part routing, a set of error recoveryprocedures for modeled errors, e.g., a robot grasping error, anda recovery procedure for generally unmodeled occurrences,e.g., ihe communication breakdown between the manufac-turing system and the supervisory control system. The latter

control system is divided into three main steps:


4/14

4 IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION, VOL. 12, NO. 1, FEBRUARY 1996

w Devents

Machine is idle operation finishrepairand re to service

Fig. 3. A simple machine.

recovery procedure can take the form of a time-out requestingoperator assistance, for example, in case no expected eventoccurrence takes place during a specified time interval.A . ModelingThe theory for discrete event systems (DES) consideredin this work is based on controlled automata concepts. Thebehavior of DES, such as manufacturing systems, are modelednaturally by finite state automata. An automaton is a five-tupleG = (E, Q , S, Qm, qo) in which:

1) C represents a finite set of transitions or event labels;2) Q represents a finite set of states;3) 6 is a transition function describing state transitions;4) Qm represents a finite set of marker states (states witha particular meaning from a control perspective); and5) go represents the automatons initial state.Finite state automata are naturally described by directed

transition graphs. In order to represent an automaton G =(E, Q , S, Qm , qo), the set of states Q is identified with thenodes (represented by 0 ) of a directed graph whose edges arelabeled with transition labels in C (represented by o 0 ). Inthe following, (4 ) indicates the automatons initial state,and ( 0 ) indicates a marker state.For the purpose of supervisory control system development,plant behaviors and corresponding behavioral specificationsare modeled in the form of directed transition graphs. Fur-thermore, events are subdivided into controllable events, i.e.,the events which may be disabled through control, and un -controllable events, i.e., the events which may not be disabledthrough control.Consider the simple machine shown in Fig. 3. It has threepossible states idle ( I ) , orking (W) , nd down (D).Its initialstate is idle, and it is also a marker state. Four events may takethe machine from one state to another. For example, the eventoperation start ( s ) takes the machine from the state idle (I)to the state working (W).Typically, operation start ( s ) is acontrollable event, whereas breakdown (6) is an uncontrollableevent.B. Synthesis

We will limit the discussion to centralized and modularsupervisory control. Centralized supervision [50], [511 implies

that the overall supervisory task is carried out by one singlesupervisor embodying the closed-loop behavior of the plant.Furthermore this supervisor must be provided with completeinformation on the occurrences of events in the plant.Modular supervision [6], [7], [59] implies the division ofthe overall supervisory task into two or more subtasks. Eachof the latter tasks is solved using the results of centralizedsupervision and the resulting individual subsupervisors arerun concurrently to implement a solution of the originalcontrol problem. A modular supervisor is ideally more read-ily modified, updated, and maintained than its centralizedcounterpart. For example, if one subtask is changed, thenit should only be necessary to redesign the correspondingcomponent supervisor. Unfortunately, these advantages arenot always to be gained without a price. The fact that theindividual supervisory modules are simpler implies that theircontrol action must be based on an aggregated version of theglobal system state. Consequently, the different componentsupervisors, acting quasi-independently on the basis of partialinformation, may come into conflict and the overall systemmay fail to be nonblocking. Thus a fundamental issue thatalways arises in the presence of modularity is how to guaranteethe nonconflicting property of the various supervisors workingtogether. Shared resources are a common example of possi-ble source of conflict between supervisors: two numericallycontrolled machines (users) require the simultaneous use ofa tool-bit and a fixture (resources); two modular supervisorsare each responsible for the control of one resource [6].If these supervisors are such that at any one time eachuser is allowed to take possession of one resource by therespective supervisor, blocking results since effectively eitheruser prevents the other from finishing its task and freeing theresource it possesses.

Centralized supervisory control is seen to be suitable forfairly small problems (in the order of lo5 states, for plantand specifications combined synchronously [50]) since thecorresponding supervisory controls tend to be large in statesize; they are required to encompass the behavior of the wholeplant subject to all specifications. Modular supervisors jointlyachieve equivalent system behaviors but tend to be smallerin state size, even though they are generally computationallymore expensive. It must be noted, however, that the corre-sponding off-line computations may often be carried out inparallel, thus reducing the effective required computation time.Taking into account the supervisory control system archi-tecture adopted (e.g., centralized, modular, etc.), the transitiongraphs representing the physical behavior of the system tobe controlled and the transition graphs representing the corre-sponding specifications are fed to a computer program whichwill return, if these exist, the supervisor(s) embodying themaximally permissive controlled behavior of the system withinthe specifications, and the corresponding control law(s).The resulting supervisors which embody the controlledbehavior of the plant, are finite state automata (representedby directed transition graphs). Let S be a supervisor for aplant G. Let x be a state of S. The corresponding supervisorycontrol law V maps every state x of S to a set of transitionsV(x), termed control. Transitions in V(x) never contradict


5/14

BRANDIN: REAL-TIME SUPERVISORY CONTROL OF AN EXPERIMENTAL MANUFACTURING CELL 5

Fig. 5. Monitoring.plant inputdevices71

InterfaceFig. 4. General computer controlled system.

the specifications considered (and are therefore termed legal)and keep the system G within the specifications.

I readevent-data I

I current-sta;update Icontrol law

C. Supervisor an d Control Law ImplementationIn the context of discrete-event systems, the fundamentalobjective of a supervisory control system is the generation ofsuitable controls aimed at enforcing a correspondingcontrolleddiscrete-event system to behave in obedience to given speci-fications. In practice, a supervisory control system must alsopossess an architecture which will allow various plant struc-tures and control schemes to be adopted. Thus a supervisorycontrol system must exhibit a great degree of flexibility andmust be amenable to integration within the plant.Recall that a supervisory control system carries out mainlythree tasks: supervisory control, communication, and house-keeping. Supervisory control consists of: i) the monitoring ofthe plant behavior via sensory feedback; ii) control evalua-tion in accordance with a supervisory control law; and iii)

control enforcement via the downloading and execution ofthe appropriate device programs. Fig. 4illustrates the generalorganization of a computer controlled system. Although almostany digital computer can be used for real-time control andother related operations, hey are not all equally easily adaptedfor such work. A computer-based control system must com-municate both with plant and personnel: communication mustbe efficient and effective and the processor must be capableof rapid execution to allow real-time control action.I ) Monitoring [4]: The monitoring of the plant behaviorvia sensory feedback is essential in supervisory control. Mon-itoring requires that the supervisory control system possess thecapability to communicate with the various devices constitut-ing the plant, in order to gather the information provided by

the devices sensors.The sensory outputs of the various devices may not alwaysbe expressed in compatible terms. It may thus be necessaryto translate these into high level language statements whichcan be treated uniformly by the supervisory system or easilyunderstood by an operator.Assume in the following, that the output of the plantoutput-devices, referred to as event-data, consists of a networkaddress, and a bit pattern representing the occurred event.Then, monitoring can be carried out in one of two waysdepending on the data transfer technique used

store controlFig. 6. Control evaluation.

with polling, the supervisory control system checks peri-odically whether a device is ready for data transfer. In thatcase, the corresponding event-data is transmitted to andreceived by the supervisory control system and stored ina buffer; andusing interrupts, the occurrence of an event causes theoutput device to interrupt the supervisory control systemoperation. The corresponding event-data is transmitted toand received by the supervisory control system and storedin a buffer. Subsequently, he supervisory control systemreturns to its pre-interrupt (background) activity.Fig. 5illustrates, in the form of flow chart, the sequence of

operations carried out for polling.2) Control Evaluation 141: All controllable events are as-sumed to be disabled by default. Control evaluation determineswhich legal events must be enabled, in accordance withthe supervisory control law considered. Control evaluationrequires that the supervisory control system possess bothcomputing power and memory capacity, to carry out therequired decision making operations, and the capability tointerface with human operators if necessary.Let S be a supervisor embodying the closed loop behaviorof a plant G .Let x be the current state of S . Control evaluationis carried out in the following way (cf., Fig. 6 ) . Throughmonitoring, event-data has been stored. Subsequently, the datais read and translated in suitable symbolic form, e.g., theoperationJinish ( f ) event for the machine shown in Fig. 3,indicating the end of a production cycle. The current state xof S is thus updated to say x.Subsequently, he correspondingcontrol V(x) is evaluated and stored.3) Control Enforcement [4]: The enforcement of controls,when considered in detail, has further implications than justthe ability to communicate with the workcell devices andvice versa; as is the case with monitoring, device integrationbecomes an important issue because of the generally hetero-Recall that in the present context, legal has the following meaning: whichdoes not contradict any behavioral specification.


6/14

6

a___..s )...--

IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION, VOL. 12, NO 1, FEBRUARY 1996

a is energised and keptenergised until reset, ifpreceding logic sentence onrung is TRUE nce.

I readcontrol I

reset(a}

examine_on(a)

examine-off(u)

Not

a-----(RI----- a is de-energised if precedinglogic sentence on rung isTRUEreturns TRUE if status of a is-..--I I----- energised

a returns TRUE f status of U is..._-I de-energised

Fig. 7. Control enforcement.

geneous nature of manufacturing systems (due, for example,to different machine makes). A successful and generally rec-ognized approach to this issue consists of the development ofhigh-level device programming languages and related devicecustomized programming primitives [ 3 ] .This approach allowsall the devices of a workcell to be programmed on a uniformbasis. The various program statements are translated in real-time into device programs through the use of programmingprimitives in which the appropriate variables are instantiated.The downloading of these programs to the appropriate devicesand their subsequent execution results in the enforcement ofthe stipulated controls. Control enforcement is carried out asfollows. Control evaluation produces a list of events to beexecuted by corresponding devices. The latter list is read itemby item, and each item is appropriately translated for eachdevice into suitable control actions, e.g., device programsand corresponding execute statements, and thus transmittedto the corresponding devices, thus effectively enforcing thecorresponding events. Fig. 7 illustrates the related sequenceof operations.4 ) Supervisory Control 141: Fig. 8 illustrates the interac-tion between monitoring, control evaluation, and control en-forcement based on polling.D. PLC-Based Implementation [4/

Programmable (logic) controllers (PLC) are comparablein design to most computer-based control systems, but arespecialized in logic-based control [41], [47], [57]. They aredesigned to operate in industrial environments with wideranges of operating conditions (ambient temperature, humidity,noise, and vibrations). Their hardware and the correspond-ing software are designed for use by plant electricians andtechnicians. They are generally programmed in ladder logicor other comparable language, and repetitively execute asingle program, sequentially, from beginning to end. Pro-grammable logic controllers have built-in inputloutput (UO)modules linking them to the field devices constituting theplant; these are typically quite robust and easy to connect andreplace.The fast access memory of most programmable (logic)controllers is broadly divided in two: the first part containingthe user program and the second containing the related datatable. The user program generally accounts for most of thememory capacity, and consists of ladder logic or equivalent

I & ranslationcontrol lawevaluationI

ig.8. Supervisory control.

monitoring

evaluation

enforcement

TABLE IBASICPLC FUNCTIONS

a is energised ifprecedingis TRUEenmgise(a) logic sentence on rung

I

language statements relating inputs to corresponding internals(internals are user-program controlled flags) and outputs.The data table contains the information required to carry outthe user program; it includes information such as the sta-tus (energizedlde-energized)of inputs, outputs, and internals,timer and counter values, digital data from field devices inputs,etc.During each operating cycle, termed scan, the CPU se-quentially reads the status of the system inputs, evaluates the


7/14


Machine 1 Buffer (1)

Fig. 9. A small manufacturing system.

control logic embodied in the user program, and consequentlyupdates the status of the outputs by either energizing or de-energizing them. Data transfer is thus carried out using pollingwith periodic checks to determine whether the various fielddevices are ready for data transfer. Typically, scan times mayvary from 1 ms up to 100 ms 1471, [571.Ladder logic programming is based on contact symbology.The five basic functions [41], [57] available on most PLCprogramming languages are described in Table I.Where a refers to either an input, an output or an internal forthe e x a m i n e m and e x a m i n e 4 f f functions, to an output orinternal for the energise function, and to an internal for theset and r e s e t functions.For example, for a , b, and c referring to either an input, anoutput or an internal, a -+ c ( a implies c) is expressed withthe rung

a CI _ - - - - - - - 1 - - - - - - - _ ) _ - - _ - _ - Il a --f c [(not a ) implies c] is expressed with the rung

a CI_ - - - - - - - [ \ I - - - - - - - -( )- - - - - - -1 .We are now in a position to consider the PLC-based implemen-tation of discrete-event system supervisory control. Recall thatsupervisory control consists of three tasks: i) the monitoringof the plant behavior; ii) control evaluation;and iii) controlenforcement.

1 ) Monitoring: As stated earlier, monitoring is carried outusing polling: during every scan, the status of the variousinputs is checked.2) Control Evaluation: Let S be a supervisor for a plantG and let V be the corresponding supervisory control law.Control evaluation is carried out as follows. To every statex of S is associated an internal labeled accordingly. Onesuch internal is energized at any time and corresponds tothe supervisors current state. Such internals are set and resetupon the occurrence of events, i.e., upon the energizing ofcorresponding inputs. Whenever a state x of S is reached,outputs corresponding to V(x) are energized.

3) Control Enforcement: Control enforcement is carriedout by the programmable logic controller output modulesupon the energizing of given outputs.

IV. EXAMPLE: ENTRALIZEDUPERVISORYONTROLA . Problem Description

The manufacturing system considered consists of two ma-chines and a buffer of size l . Machine l and Machine 2are connected by the buffer (cf., Fig. 9). The machines are

Machine 2

IACHINE 1 IACHINE 2

Machine 1 s workingMachine 1 is downMachine 2 is idle

b2r2

Fig. 10. M a c h 1 and Machine 2.

Iventsoperation start on Machine 1operation finishonMachine 1breakdown ofMachine 1rep. +ret. to serv. of Mach. 1operation start on Machine 2operation finishonMachine 2breakdown ofMachine 2rep. + ret. to serv. of Mach. 2

either idle, working, or down. The buffer is either emptyor full. Initially the buffer is assumed to be empty and themachines are assumed to be idle. The transfer of workpieces isassumed to be part of the machines workcycle, during whichthe machines pick up a workpieces upstream, and transferworkpieces downstream. Centralized supervisory control isto be used. The machines operation and repair must becoordinated according to the following production and repairspecifications:the buffer must not overflow or underflow: Machine 1may not start operating while a workpiece is present inthe buffer,. and Machine 2 may not start operating unlessa workpiece is present in the buffer; andMachine 2 has repair and return to service priority overMachine 1: in case both machines are down, Machine 2must be repaired and returned to service first.

The operation of the machines is controlled via informationfeedback on the occurrence of events.B. Control System Development

Recall that the development of a supervisory control systemis divided into three main steps:1 ) Modeling of the Plant and Behavioral Specijications:The manufacturing system, that is Machine 1 and Machine2, and the behavioral specifications are modeled in the formof transition graphs as illustrated in Figs. 10-12. ConsiderMachine 1. Initially it is idle; its state is 11. Once it startsoperating, i.e., the event s l occurs, it changes state to W1.Depending on what event happens next it may either:

change its state to idle (Il ), if it successfully finishes itschange its state to down (Dl), if it breaks down (bl).work cycle (fl); or


8/14

8

states

E buffer s emptyF buffer is full

IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION, VOL. 12, NO. 1, FEBRUARY 1996

-eventsfls2 operation finish on Machine 1operation starton Machine 2

States

OK Machine 2 is fineKO Machine 2 is down

Fig. 11. Buffer specification.

eventsrlb2 breakdown of Machine2r2

~ p .ret. to sew.ofMachine 1rep. + ret. to sew. of Machine 2

r2OKn

Fig. 12. Breakdown specification.

If the machine is down (Dl), it is repaired and returned toservice (rl), bringing the machine back to its idle state (11).The same applies to Machine 2.The transfer of workpieces is assumed to be part of the ma-chines' workcycle, so that when (sl) and (s2) have occurred,the machines are assumed to have picked up a workpieceupstream, and when (fl) or (f2) have occurred, the machinesare assumed to have transferred a workpiece downstream.The events sl, s2, rl, and r2 are controllable: they can bedisabled and enabled by control action. The events bl, b2,fl, and f2 are uncontrollable: they cannot be disabled andmay occur spontaneously. The behavioral specifications aremodeled as shown in Figs. 11 and 12.The buffer specification transition graph means the follow-ing:* initially the buffer is empty;Machine 2 may not start operating prior to the buffer

being Idl, that is prior to the occurrence of an operationfinish on Machine 1 (fl);* the buffer may not be filled again (fl) before the occur-rence of an operation start on Machine 2 (s2); andno other event occurrence affects the buffer specification.The breakdown specification transition graph means the

initially Machine 2 is OK;following:

Machine 1 may be repaired and returned to service (rl)any time as long as Machine 2 is OK ;* in case Machine 2 breaks down (b2), Machine 2 must berepaired and returned to service before Machine 1 maybe repaired, that is the event rl may not occur at the stateKO; and* no other event occurrence affects the breakdown speci-fication.

2) Supervisor and Control Law Synthesis: In order to syn-thesize a control law which will enforce the specificationsconsidered, the above machine and specification transitiongraphs are fed to a computer program which will

11

Fig. 13.behavior. Supervisor S embodying the plant maximally pe m ss iv e controlled

tell us whether it is possible for the machines to behavewithin the specifications, and* return the supervisor embodying the maximally permis-sive controlled behavior of the machines within thesespecifications, and the corresponding control law.In this example the machines can behave within the bufferand breakdown specifications. The maximally permissive con-trolled behavior of the machines within these specifications isshown in Fig. 13.State 1 of the graph corresponds to Machine 1 being idle

(Il), Machine 2 being idle (I2), and the buffer being empty(E). htially Machine 1 starts working (sl) to fill the buffer.If Machine 1 finishes its cycle (fl ), Machine 2 can start work(s2), etc.Table 11provides further information on the meaning of thesupervisor states.From the above transition graph the control law (shown inTable m) is obtained by listing the controllable events exitingthe states of the controlled behavior transition graph.

Through the disablement of controllable events, the combi-nations of states which are not allowed by the specificationsare never reached by the machines and buffer. It is importantto note that, for example, the control law does not allow thecombination (Wl, 12, F, OK) to be reached from state 3 (11,12, F, OK ) by starting Machine 1 (sl): the combination (Wl,12, F, OK) could potentially lead to the buffer overflowing,that is having two parts in it, if Machine 1 finishes operating(fl) before Machine 2 starts operating (s2).3) Supervisor and Control Law Implementation: The su-pervisor and control law are coded in the supervisory controlsystem: a programmable logic controller or computer. Controlis carried out via information feedback on the occurrence ofevents.Initially, the machines are idle (11, I2), the buffer is empty(E), and the supervisor S is in the initial state, that is state1. According to the control law, the corresponding control is

sl: starting the operation of Machine 1. The control systemstarts Machine 1 and updates the supervisor state to state 2.Depending on what occurs next the control system will updatethe supervisor state and send out controls according to thecontrol law. For example if Machine 1 finishes its cycle (fl)the supervisor state is updated from statel2 to state 3, and thecorresponding control is s2: starting the operation of Machine2, etc.


9/14

BRANDIN: REAL-TIME SUPERVISORYCONTROL OF AN EXF'EFSMENTAL MANUFACTURING CELL

supervisorstate x

123456789101 112

9

controlV(X)

S Is2Sl

r2sl,r2r2r2r lrl

TABLE 11CONTROLLEDEHAVIORTATECORRESPONDENCE

supervisorstate

123456789101112

Machine 1stateI1w1I1I1w1I1I1I1w1D1D1D1

Machine 2stateI2I2I2w 2w 2w 2D2D2D2D2w 212

Using the approach presented, the contrc- system will iways react to any event occurrence in the most permissive waywithin the specifications considered. Initially, the machines areidle (11, I2), the buffer is empty (E), and the control systemis in the initial state of the controlled behavior graph, thatis state 1. According to the control law, the correspondingcontrol is sl: starting the operation of Machine 1.The controlsystem starts Machine 1 and updates its own state to state 2.Depending on what occurs next the control system will updateits state and send out controls according to the control law.For example if Machine 1 finishes its cycle (fl) the controllerstate is updated from state 2 to state 3, and the correspondingcontrol is s2: starting the operation of Machine 2, etc.V. EXPERIMENTALETUP

A. Problem DescriptionThe automated manufacturing cell considered consists ofa robot, a vision system, a conveyor belt, two simulatednumerically controlled machines, a mill and a drill, anda supervisory control system constituted by ProgrammableLogic Controllers. Two types of workpieces must be workedon by both the mill and the drill before being placed in the

bufferspecificationstateEEFEEFFEEEEE

repairspecificationstateOKOKOKOKOKOKKOKOKOKOOKOK

Superviso+Y control system

Conveyor belt U Uill~~Robot and VLSiOn system

0 utgoing pallet /' Pallet positioning system

Bin DrillFig. 14. The experimental workcell.

bin. Workpieces arrive in random order on the conveyor belt.Supervisory control may be carried out in either centralized ormodular fashion. The cell is shown in Fig. 14.B. Modeling

1 ) Robot and Vision System: A GMF S-100 robot with aKarel R-F controller, is used to transfer workpieces withinthe cell. Along with its grasping tool it is equipped witha camera for workpiece recognition, linked to the workcellvision system.The robot is programmed to carry out seven tasks. Theseare:1) go to conveyor and grasp, and go to mill and drop;2) go to drill and grasp, and go to mill and drop;3) go to mill and grasp, and go to drill and drop;4) go to conveyor and grasp, and go to drill and drop;5 ) go to mill and grasp, and go to bin and drop;6) go to drill and grasp, and go to bin and drop; and7) recognition.The robot is hardwired to the PLC with four output in-ternals, which are energized in patterns in order to activatecorresponding robot tasks.The vision system is characterized by object (workpiece)premarking, and active visual perception. Objects are mod-eled by a small number of their two-dimensional perspective


10/14


Robot + Vision SystemI recl.rec2

Recognitionop11,op12,op21,op22,exil,exi2

Transfer

Fig. 15. The robot and vision system.views, referred to as standard views, each associated with acorresponding standard view axis. The mobile camera attachedto the robot arm is used to acquire a standard view of theobject (workpiece) to be recognized. Local surface normals aredefined for each standard view by premarking the object withred circular markers (defining corresponding standard viewaxis).Initially the vision system attempts to detect a marker inthe field of view according to a preplanned search strategy.In case a marker is detected, the corresponding standard viewis estimated, and the surface normal and marker center arecomputed with respect to the camera frame. Subsequently,through the alignment of the cameras optical axis with the:estimated surface normal, a standard view of the object isacquired. Depending on the degree of similarity between theacquired standard view and the reference standard views ofall the possible objects considered, one or more candidates areselected. In case only one matching reference standard viewis found, successful recognition is assumed. In case severalmatching reference standard views are found, one additionalstandard view is acquired. In case no unique candidate or nocandidate is found, the object is assumed to be unrecognizable.Finally, having recognized an object, its orientation and posi-tion are determined. In this example, recognition will alwaysbe assumed to be successful, and the vision system will beable to distinguish between two types of workpieces: type 1and type 2.The robot and vision system are modeled as shown inFig. 15.All events apart from (recl) and (rec2) are control-lable. A complete list of events for the experimental setup isprovided in Table IV.

2) The Conveyor Positioning System: The conveyor usedis an Amatrol 83-129 loop conveyor with a 14 130 pallet po-sitioning system. Its purpose is the transportation of incomingworkpieces. The pallet positioning system, shown in Fig. 16,consists of

* a positioner limit switch, attached on the inner conveyorrail just upstream of the pallet positioner, is placed so thatthe incoming pallet will trip It, thus signaling the palletspresence;

0 two gage stops which are mechanical stops mounted oneither side of the conveyor belt immediately downstreamfrom the pallet positioner;* a restraint pin located upstream from the limit switchalong the inside rail of the conveyor, its purpose being toqueue pallets for the pallet positioning system;* a pallet positioner which consists of four lifting rodswhich straddle the belt in a rectangular formation; theirpurpose is to lift the pallet above the moving conveyorbelt; and

TABLE IVEVENTSOR FIGS. 5-21

inrereClreC2o p l lop12op21op22exilexi2gaupffadoPiUP

sens

pidoP U Ppod0stw2stw4wai2wai4

events (c: controllable, U:uncontrollable)initialise recognitionrecognition of a workpiece of type 1recognition of a workpiece of type 2go to conveyor and grasp, go to m ill and dropgo to drill and grasp, and go to mill and dropgo to mill and grasp, p d go to drill and dropgo to conveyor and g rasp, go to drill and dropgo to mill and grasp, and go to bin and dropgo to drill and grasp, and go to bin and droppositioner limit switch senses a palletgage tops is placed in the u p position,blocking a pallet in the positoner systemgage stops is placed in the down positionrestraint pin is placed in the u p position,queuing pallets upstreamrestraint pin is p laced in the down position@et positioner is placed in the up position,liftiig pallet from the conveyor beltpallet positioner is placed in the down positionstart time out of 2secondsstarttime out of4 econdstime ut of 2 secondstime out of4 econds

Pos. limit switchpido

Restraint pinpido

Timer2

StW2Fig. 16. The conveyor positioning system.

Pallet positioner

Timer4Aw4

-UUCCCCCCUC

CcCC

CCCUU-

tw o timers used to ensure that the pallets reach the desiredposition before acting on either the restraint pin or thepallet positioner (see corresponding specifications); thetimers areprogrammed in the supervisory control system.The various actuators of the pallet positioning system arepowered by compressed air and controlled with solenoids. Allevents are controllable apart from (wai2) and (wai4).3) The Simulated Numerically Controlled Machines: Thenumerically controlled machines considered are simulated andshown in Fig. 17. They are composed of a platform onwhich workpieces are positioned by the robot and mechanicalswitches which indicate their presence or otherwise. Theswitches are also hardwired to the PLC. Worktime is simulatedby time delays programmed on the PLC. Since for controlpurposes, only the end of operations by the mill and drill are ofinterest, the above models are simplified as shown in Fig. 18.

All events apart from (opll), (op12), (op21), and (op22) areuncontrollable.


11/14


Mill endl

Drill end2

Fig. 17. The numerically controlled machines.=pll, oplD

Fig. 18 . The numerically controlled machines (simplified).

4) The Supervisory Control System: The supervisory con-trol system is composed of two Allen-Bradley PLC-5 pro-grammable logic controllers linked by an Allen-Bradley DataHighway Plus network to a personal computer used for pro-gramming purposes.5 ) Behavioral Specifications: Four behavioral specifica-tions must be enforced through supervisory control. Theseare as follows.

1) Two types of workpieces arrive on pallets in the workcellin random order. Type 1 workpieces must be workedon first by the mill, then by the drill and subsequentlyevacuated to the bin. Type 2 workpieces must be workedon first by the drill, then by the mill and subsequentlyevacuated to the bin. The graphs shown in Fig. 19enforce the above specification. Type 1 workpieces areworked on first by the mill (opll, endl) , hen by thedrill (0~21,nd2) and are subsequently evacuated to thebin (exi2). Type 2 workpieces are worked on first by thedrill (op22, end2), then by the mill (op12, endl) and aresubsequently evacuated to the bin (exil).2) Both the mill and drill have a buffer of size one. Oneworkpiece only may be transferred to these machinesat any one time. This specification is implicit in thegraphs S4-S7 shown in Fig. 19: neither machine maystart operating on a new workpiece before the workpiececurrently on the machine has been transferred onwards.

3) The various actuators of the conveyor belt positioningsystem must be coordinated so that only one palletat time is halted and lifted from the conveyor, theremaining pallets being queued upstream. The graphshown in Fig. 21 enforces the above specification. Ini-tially the arrival of a pallet triggers the positioner limitswitch (sens). Subsequently, both the gage stops and therestraint pin are set in the up position (gaup, piup), thusblocking the pallet that has arrived, and queuing arrivingpallets. Furthermore, before setting the pallet positionerin the up position (poup), a time delay of two seconds(wai2) is introduced allowing the pallet to travel to thegage stop. The recognition and transfer of the workpiecemay now take place: the pallet will not move until a

Fig. I

4)

54

op22,op12,exi2,endl55 op12

Wopll,opZl,exil,endZexi26+:a

opll,op12,op22,end257 exi

,9. The routing specifications.

pallet advance (adpa) is requested. Then, the gage stopsand the pallet positioner are set in the down position(gado, podo). Before setting the restraint pin in the downposition (pido), to allow a new pallet arrival, a time delayof four seconds (wai4) is introduced letting the palletpresent in the positioning system to travel downstream ofthe gage stop. The events (wai2) and (wai4) are assumedinternal to the supervisory control system.Workpiece recognition may only be carried out uponarrival of a pallet at the conveyor belt presentationsystem. The graph shown in Fig. 20 enforces the abovespecification. The recognition procedure must be initial-ized (inre) and either a workpiece of type 1 or of type2 must be recognized before the robot is allowed totransfer it. Upon transfer of the workpiece, the palletmay be advanced (adpa), allowing the arrival of a newworkpiece.

C. Supervisor and Control La w SynthesisThe cell, constituted by the two machines and the palletpositioning system, is represented by a graph with 256 states

and 28 16 transitions. The specifications grouped together arerepresented by a graph with 540 states and 1977 transitions.With centralized supervision [5], [49], the correspondingsupervisor embodying the controlled behavior of the plantwithin the specifications is a graph with 540 states and 1329transitions.With modular supervision (under partial observations) [6],[50], it is possible to supervise the manufacturing cell, andobtain the same controlled behavior obtained with centralizedsupervision, with six supervisors functioning concurrently,namely the specification graphs S 4 4 9 shown in Figs. 19-21.


12/14


pidostw2

w a i lI wai21Fig. 20. The positioning system specification.

19i n r e

adpa~ rec2 J: r e c l i

adpa

Fig. 21. The pallet-advance specification.

This approach requires the coding in ladder logic of 29states (3 states each for specifications S4-S7, 12 states forspecification S8, and 5 states for specification S9) to track thesupervisors states, compared to 540 states when consideringcentralized supervision.D. upervisor and Control Law Implementation

The supervisors S4S7 and S9, responsible for the behaviorof the robot and the vision system and also the numeri-cally controlled machines, were coded on one programmablecontroller (PLCO) and supervisor S8, responsible for thebehavior of the pallet positioning system, was programmed ona second programmable controller (PLC1) (cf., Fig. 22). Theprogrammable controllers and corresponding supervisors im-plement the behavioral specifications concurrently and withoutconflict.

VI. DISCUSSIONND CONCLUSIONSA new approach to the real-time supervisory control ofautomated manufacturing systems was implemented and testedon an experimental manufacturing cell composed of a robotand vision system, a conveyor belt with pallet positioning

31

Mill

vision system

Ling sysrem

Fig. 22. The ceU and PLCs.TABLE VPLCs AND CORRESPONDINGUPERVISORS

PLC0PLCl I S8I S4, S5 , S6, S7, S9

system, and two simulated numerically controlled machines:a mill and a drill.The work illustrated the viability of automata based ap-

proaches, in particular modular supervision under partial ob-servation. Computational tools used were used to synthesizethe supesvisors and corresponding control laws, and/or toverify and guarantee the supervisors nonconflictitlg nature.The plant controlled behavior obtained was observed neverto violate the specifications considered, and was nonblockingas guaranteed by the theory: supervisors and control lawsobtained are correct by construction, and ensures that:

the resulting controlled behaviors do not to violate thespecifications considered; andULe supervisors and control laws obtained are maximallypermissive within the behavioral specifications consid-ered.

Off-the-shelf industrial programmable controllers fromAllen-Bradley were used in the experimental set-up illustratingthat the approach adopted can be implemented on industrialcontrol systems and does not require investment in specializedequipment.

The work carried out comprised the following steps:the integration of the industrial control system within theproblem scenario development;cell modeling:specification modeling;supervisor and control law synthesis;robot and vision system programming; and

* PLC programming.The work carried out took approximately 4 man-months topeople not familiar with PLCs.The results of this work are currently being applied to thesupervisory control of an automated assembly cell consisting

of 3 robotic assembly stations linked by five conveyor belts.Current results show that the supervisory control systemdevelopment approach presented is applicable to such rela-tively complex systems (over 400 inputs and outputs, and

manufacturing cell;


13/14

BRANDIN: REALT IME SUPERVISORY CONTROL OF AN EXPERIMENTAL, MANUFACTURING CELL 13

approximately 80007 states) using off-the-shelf industrialprogrammable controllers.

GLOSSARYplant: system to be controlled, e.g., a manufacturing cell.supervisory control system: physical system that performsthe following tasks: supervisory control, communicationand housekeeping.supervisory control: consists o f i) the monitoring of theplant behavior via sensory feedback; ii) control evalua-tion (determination) in accordance with a supervisor andsupervisory control law which map the plant behavior tocorresponding controls; and iii) control enforcement.communication: allows sensory feedback and control en-forcement to be performed.housekeeping: is the set of tasks related to supervisorycontrol and communication which are necessary to theirimplementation, e.g., data-base management.supervisory control development: consists of the formu-lation and synthesis o f supervisors and correspondingcontrol laws specifying how the supervisory control sys-tem is to react to the manufacturing system behavior inorder to satisfy given behavioral specijications.supervisor: consists of an automaton embodying the plantcontrolled behavior, in whole or in part.automaton: mathematical and modeling tool for discreteevent systems.discrete event systems: dynamic systems that evolve inaccordance with the abrupt occurrence of events.supervisory control law: maps the plant controlled behav-ior to corresponding controls.control: list of events.behavioral specijications: are classified into:

logic-based specifications such as safety, error recov-ery, the sequencing of operations, part routing, andproduction volume requirements;temporal production specifications: e.g., productiontimes; andutility optimality specifications: e.g., costs.

ACKNOWLEDGMENTThe author would like to thank Allen-Bradley (Canada)and Rockwell International (Canada) for their support ofthe Ontario/Rh8ne-Alpes Collaboration Project on The Su-pervisory Control of Automated Manufacturing Systems, and

in particular D. Pattengale, Integrated Solutions Consultant,Allen-Bradley (Canada). The author would also like to thankD. Anderson, Executive Vice-president University Relations,Manufacturing Research Corporation of Ontario (MRCO),for his help in establishing the first contacts with RockwellInternational and Allen-Bradley, and finally but not least,Technology Ontario and Dr. M. Walmsley, Director, PremiersCouncil Technology Fund.

REFERENCES[ l ] S. Balemi, Discrete-event systems control of a rapid thermal multipro-cessor, in INCOM 92, Toronto, Canada, May 1992, pp. 53-58.

H. S. Baird, E. G. Wells, and D. E. Britton, Coordination softwarefor rob otics workcells, in 1984 IEEE Int. Con$ Rob otics, Atlanta, GA,May 198 4, pp. 354-360.S . Bennett, Real-Time Computer Control: An Introduction. EnglewoodCliffs, NJ: Prentice-Hall, 1988.B. A. Brandin, The real-time supervisory control of automated manu-facturing systems, Ph.D. dissertation, Dep. Elect. Comput. Eng., Univ.Toronto, Canada, 1993. Also System s control group report 9302, Dep.Elect. Comput. Eng., Univ. Toronto, Canada, Feb. 1993.B. A. Brandin and W. M. Wonham, The supervisory control of timeddiscrete-event systems, IEEE Trans. Automat. Contr., vol. 39, no. 2,pp. 329-342, Feb. 1994.-, Modu lar supervisory con trol of timed discrete-event systems,in I993 IEEE Control and Decision Con$, San Antonio, TX, Dec. 1993,pp. 2230-2235.B. A. Brandin, W. M. W onham, and B. B enhabib, Manufacturing cellsupervisory control-a modular timed discre te-eve nt system app roach ,in 1993 IEEE Int. Con$ R obotics and Automation, Atlanta, GA, May1993, pp. 846-851.B. A. Brandin and W. M. Wo nham, The supervisory contr ol of timeddiscrete-event systems, in 1992 IEEE Control and Decision Con$,Tucson, AZ,Dec. 1992, pp. 3357-3362.B. A. Brandin, W. M. W onham, and B. Benh abib, Modular supervisorycontrol of timed discrete-event systems, in 1992 Allerton Commu-nication and Control Con$, Urbana-Champain, IL, Sept. 1992, pp.624-632.-, Manufacturing cell supervisory control-A timed discre te-event system approach, in 1992 IEEE Int. Con$ Robotics and Automa-tion, Nice, France, May 1992, pp. 531-536.-, Discrete event supervisory control applied to the manage-ment of manufacturing workcells, in 7th Int. Con$ Compu ter AidedManufacturing Engineering, 1991, pp. 527-536.B. A. Brandin, T. Chen, C. D erventzis, and G. Pang, The co ntroland superv ision of grou ps of lifts using the blackboard architectureapproach, Trans. Inst. M.C ., vol. 12, no . 5 , pp. 271-279, 1990.R. Cassin is, Hierarchical control of integra ted manufacturing system s,in Proc. 13th Int. Symp. Industrial Robots and Robotics, Chicago, IL,Apr. 1983, pp. 12-19.J. A. Cecil, K. Srihari, and C. R. Emerson, A review of Petrinet applicatio ns in manufacturing, Int. J. Advanced ManufacturingTechnol., vol. 7, no. 7, pp. 168-177, 1992.H. Chochon and R. Alami, NNS, a knowledge-based on-line systemfor an assembly workcell, in I98 6 IEEE Int. Con$ Robotics andAutomation, San Francisco, CA, Apr. 1986, pp. 6 0 3 4 0 9 .E. Coste-Manibre, B. Espiau, and D . Simon, Reactive objects in a tasklevel open controller, n 1992 IEEE Int. Con$ Robotics an d Automation,Nice, France, May 19 92, pp. 2732-2737.E. Coste-Manikre, B. Espiau, and E. Rutten, A task level robotprogramming language and its reactive execution, in 1992 IEEEInt. Con$ R obotics and Automation, Nice, France, May 19 92, pp.R. David and H. Alla, Autonomous and timed co ntinu ous Petri nets,presented at the 1 th Int. Conf. Application and Theory of Petri Nets,Paris, France, June 1990.R. David, Modeling of d ynamic systems by Petri nets, in ECC91European Control Con$, Grenoble, July 2-5, 1991, pp. 136-1 47.P. Freedman, Time Petri nets and robotics, IEEE Trans. Robot.Automat., vol. 7, no. 4, pp. 4174 33, Aug. 1991.B. Fromm herz, Rob ot action planning, Comput. Mech. Eng., vol. 6 , no. 3, pp. 30-36, Nov./Dec. 1987.C. A. Gauthier and M. Krieger, Developm entof a flexible assembly celltest-bed, in NRC Symp. Manufacturing Applica t. Languages, Winnipeg,Canada, June 1988, pp. 61-70.J. Gaylord, Factory Information Systems: Design and Implementationfor CIM Management and Control. New York Markcel Dekker, 1987.S . Y. Harmon, Steps toward autonomous manufacturing, in I987IEEE Int. Con$ R obotics and Automation, Raleigh, NC, Mar. 1987, pp.1896-1902.B. H ayes-Roth, A blackboard architecture for con trol,AI, vol. 26, no.3, pp. 251-321, 1985.W. B. Heginbotham, Ed., Programmable Assembly. Bedford, U.K.:Springer-Verlag, 1984.G. Hoffmann, C . Sch aper, and G. Franklin, Discrete-event control fora rapid thermal multiprocessor, in Proc. American Control Con$ 1991,Boston, MA, Ju ne 1991, pp. 2936-2938.E. Holloway and B. H . Krogh, Synthesis of feedback control logic fora class of controlled Petri nets, IEEE Trans. Automat. Contr., vol. 35,no . 5 , pp. 514-523, May 1990.

2751-2756.


14/14

14 EE E TRANS#ACTIONSON ROBOTICS AND AUTOMATION, VOL. 12, NO . 1, FEBRUARY 1996

[29] A. Jones and V. C. McLean, A proposed hierarchical control modelfor autom ated manufacturing systems, J. Manufacturing Syst., vol. 5,no. 1, pp. 15-25, 1986.[30] V. C. Jones, MAP/TOP Networking: A Foundation for Computer-Integrated Manufacturing.[31] J. S. Jun and C. R. McLean, Control of an automated machiningworkstation, IEEE Gntr. Syst. Mag., vol. 8 , no. 1, pp. 26-30, Feb.1988.[32] R. B. Kelley, Vertical integration for robot assembly cells, in IEEEInt. Con&Robotics and Automation, San Francisco, CA, Apr. 1 986, pp.

[33] D. T. Koenig, Computer Integrated Manufacturing Theory and Practice.New York Hemisphere, 1990.[34] P. K ozak, Supervisory control of discrete-event processes: A real timeextension, Inst. Inform. Theory Automat., Czechoslovak Acad. Sci.,Prague, Tech. Rep., p. 26, Mar. 1991.[35] A. Kusiak and A. Villa, Architectu res of expert systems for schedu lingflexible manufacturing systems, in I987 IEEE Int. Con$ Robotics andAutomation, Raleigh, NC, Mar. 1 987, pp. 113-117.[36] A. Kusiak, Ed., Flexible Manufacturing Systems: Methods and Studies.Amsterdam: Elsevier, 1986.[37] J. E. Lenz, Flexible Manufacturing: Benefits fo r the Low InventoryFactory.[38] F. Lin and W . M. Wonham, D ecentralized superv isory control ofdiscrete-event systems,Inform. Sei., vol. 44, no. 3, pp. 199-224, 1988.1391 0. Z. Maimon and S. B. G ershwin, Dynamic scheduling and rou tingfor flexible manufacturing systems that have unreliable machines, inIEEE Int. Con& Robotics and Automation, Raleigh, NC, Mar. 1987, pp.

[40] J. Martin, Design o Real Time Com puter Systems. Englew ood CLiffs,NJ: P rentice-Hall, 1976.[41] G. Michel, Programmable Logic Controllers and Architecture and Ap-plication. New York Wiley, 1990.[42] M. C. Moed and R. B. Kelley, An expert supervisor for a roboticworkcell, Intell. Robots Wsion Syst., vol. SPIE-848, pp. 502-506, NOV.1987.[43] -, An on-line supervisor for robotic tasks, in IEEE ComputerIntegrated Manufacturing Con$, Troy, NY, ay 1988, pp. 27&281.[44] P. W. Moir, ESJicient Beyond Imaginin g: CIM an d its Applications forTodays Industry. New York Halsted, 1989.[45] T. Murata, Petri nets: Properties, analysis, and applications, ZEEEProc., vol. 77, no. 4, Apr. 1989.[46] P. OGrady and K. Lee, A n intelligent cell control system for automatedmanufacturing, Int. J. Production Res., vol. 26, no. 5, pp. 845-861,1988.[47] F. D. Petruzella, Programmable Logic Controllers. New YorkMcGraw-Hill, 1989.1481 J. R. Pimentel, Communication Networks for Manufacturing. Engle-wood Cliffs, NJ: Prentice-Hall, 1990.[49] B. Prasad, Ed., Proc. 3rd Int. Con$ CAD/CAM Robotics and Factories

o the Future, South Field, MI, 1988.[50] P. J. Ramadge and W. M. Wonham, Supervisory control of a class ofdiscrete-eve nt systems, SIAM, . Contr. Optimization, vol. 25, no. 1, pp. 206-230, Jan. 1987.

New York McGraw-Hill, 1988.

585-590.

New York Marcel Dekker, 1989.

281-288.

[Sl] -, Th e control of discrete-event systems, IEEE Proc., vol. 77,no. 1, pp. 81-98, Jan. 1989.[52] U. Rembold, C. Blume, and R. Dillmam, Computer and IntegratedManufacturing Technology and Systems. New York Marcel Dekker,1985.[53] A. C. Sanders on, Parts entropy methods for robotic assembly systemdesign, in I984 IEEE Int. Con$ Robotics, Atlanta, GA, May 1984, pp.m o 8 .1541 K. P. Valavanis and P. H. Yuan, Hardware and software for intelligentrobotic systems, J. Intell. Robot. Syst., Theory Applicat., vol. 1,no. 4,

[55] R. Vijaykumar and M. A. Arbib, Problem decompo sition for assemblyplanning, in IEEE Int. Con& Robotics and Automation, Raleigh, NC,1561 I. G. Warnock, Programmable Controllers-Operation and A pplication.Englewood Cliffs, N J Prentice-Hall, 1988.[57] A. Weatherall, Computer Integrated Manufacturing: From Fundamen-

fals to Implementation.[58] C. C. White, Et, Final report, presented at the NSF Workshop on theDecision and Control of DED S, Airlie House, Airlie, VA, June 1988.[59] W. M. Wonham and P. J. Ramadge, Modular supervisory control ofdiscrete event systems, Maths. Contr., Signals Syst., vol. 1, no . 1, pp .13-30, 1988.

pp. 343-373, 19 89.

M a . 1987, pp. 1361-1366.

London and B oston: Butterworths, 1988.

Bertil A. Brandin (M95) received the Bachelorsdegree in mechanical engineering from the Uni-versity of New South Wales, Australia, in 1984.He received the Master of Applied Science andDoctorate degrees in electrical engineering from theUniversity of Toronto. Canada, in 19 89 and 199 3,respectively,From 1993 to 1995, he was Project Leader ina collaboration project on the supervisory controlof automated manufacturing systems between theProvince of Ontario. Canada. and R&ion RhBne-Alps, France. In 1994, he was an Invited Scientist at the Rockwell ScienceCenter, Thousand Oaks, CA. Prior to his Masters program, he workedfor three years at the Swiss Federal Institute of Technology, Lausanne,Switzerland. His research interests include the control of discrete-eventsystems from both theo retical and implem entation perspectiv es, artificialintelligence and how led ge engineering techniques, and their application insystems control. Automated m anufacturing systems have been the focus ofhis implementation efforts.

Documents

J96_The Real-time Supervisory Control of an Experimental Manufacturing Cell