《系统工程》课程教学资源（英文文献）Logistic control for fully automated large scale freight transport systems

Logistic control for fully automated large scalefreight transport systems; Event based control forthe Underground Logistic System SchipholGuide Words:Logistic control; Evaluation of the logisticAbstract: Freight transport systems of tomorrow will differ largely from the systems that are usedtoday. They will be large, and they might be fully automated. These new freight transport systems askfor a new logistic control approach. This paper provides new concepts for logistic control of highlyautomated transport systems.Further-more, some novel characteristics of logistic control aredescribed.The concepts are illustrated by examples from a large research project on a highlyautomated transport system, the Underground Logistic System Schiphol (OLS).The most importantaspects of the OLs logistic control are decentralization, distribution, and an event-based informationexchange between hierarchical control layers. The logistic control developed for the OLS was testedand evaluated by using a combination of simulation and real (scale) models and prototypes of theequipment that will be used within the OLS. It proves to work very well. In current studies the controlconcepts of the OLS are introduced for other complex transport systems as well. The first results areverypositiveI.INTRODUCTIONFreight transport systems of the early decades of the 21st century will differ largely from the onesused in the past decades (Rijsenbrij 1999). The freight flows will have larger scales, since the freightflows between‘rich'countries continue to grow strongly.The customers are becoming increasinglymore demanding. Customers demand more reliability,faster throughput times, higher service levels.and more flexibility (Versteegt 1999; Evers 1999; Evers 2000). These changes put a heavy burden onthetransport systems.On the other hand new possibilities have arisen. The equipment that will be used in futuretransport systems is highly automated, like Automatic Guided Vehicles (AGV) and automatedtransshipment facilities. These high levels of automation makefaster throughput times possible, whilesimultaneously reducing labor costs.Furthermore, the use of new information and communicationtechnology (ICT), like wireless networking, improve mobile communication. This makes it possible to

Logistic control for fully automated large scale freight transport systems; Event based control for the Underground Logistic System Schiphol Guide Words：Logistic control; Evaluation of the logistic. Abstract：Freight transport systems of tomorrow will differ largely from the systems that are used today. They will be large, and they might be fully automated. These new freight transport systems ask for a new logistic control approach. This paper provides new concepts for logistic control of highly automated transport systems. Further-more, some novel characteristics of logistic control are described. The concepts are illustrated by examples from a large research project on a highly automated transport system, the Underground Logistic System Schiphol (OLS). The most important aspects of the OLS logistic control are decentralization, distribution. and an event-based information exchange between hierarchical control layers. The logistic control developed for the OLS was tested and evaluated by using a combination of simulation and real (scale) models and prototypes of the equipment that will be used within the OLS. It proves to work very well. In current studies the control concepts of the OLS are introduced for other complex transport systems as well. The first results are very positive. I. INTRODUCTION Freight transport systems of the early decades of the 21st century will differ largely from the ones used in the past decades (Rijsenbrij 1999). The freight flows will have larger scales, since the freight flows between ‘rich’ countries continue to grow strongly. The customers are becoming increasingly more demanding. Customers demand more reliability, faster throughput times, higher service levels, and more flexibility (Versteegt 1999; Evers 1999; Evers 2000). These changes put a heavy burden on the transport systems. On the other hand new possibilities have arisen. The equipment that will be used in future transport systems is highly automated, like Automatic Guided Vehicles (AGV) and automated transshipment facilities. These high levels of automation make faster throughput times possible, while simultaneously reducing labor costs. Furthermore, the use of new information and communication technology (ICT), like wireless networking, improve mobile communication. This makes it possible to

control automated and moving systems on a real-time basis. These possibilities are the enablers ofnew transport systems.These increased demands on transport systems and new possibilities ask for a different approachfor the logistical control of these transport systems. The research described in this paper focuses onthe development of logistic control for a new highly automated transport system, the UndergroundLogistic System Schiphol(OLS).Afterthisintroduction,thenext sectionwill providea shortoverviewof logisticcontrol.Thethirdsection will introduce the case study for which the logistic control was designed, the UndergroundLogistic System Schiphol.Section 4discusses the implementation of new types of control concepts inthe case study. The evaluation and tests of the logistic control are covered in section 5. This paperends with conclusions and some thoughts on future research on logistic control.II.LOGISTICCONTROLControl is defined here as a set of mechanisms used to regulate or guide the operation of amachine, apparatus or system.A logistic control system is responsible for controlling the flow ofentities (goods, passengers, or even information) through the entire system, fulfilling demands of the(final)customersas muchas possible."A logistic control system has two kind of control activities (Euwe 1999). A logistic control actionis a request for an activity in a logistic system on a certain moment in time. A logistic coordinationaction is a specification of a dependency at a certain moment in time between different controlsystems.The design of logistic control has become an important part within the field of logistics. In thenear future, when the transport systems become more complex and higher performance is needed, itwill play an even more important role. Several authors support this development (Roderique 1999,Kulick and Sawyer 1999, and Ryan 1998).III.CASE:UNDERGROUNDLOGISTICSYSTEM SCHIPHOLIn the Netherlands around Amsterdam Airport Schiphol and the Flower Auction Aalsmeer theroads are heavily congested. This leads to long throughput times and unreliable delivery rates of thetransport of time-critical and expensive airfreight (e.g. flowers, computer parts, newspapers) betweenAmsterdam Airport Schiphol, logistics centers near Schiphol, the Flower Auction Aalsmeer, and a(future)RailTerminalnearSchiphol.Tosolvethisproblemanundergroundlogisticsystemhasbeendesigned for the transport of the expensive and time-critical air-cargo. Because of the separation of

control automated and moving systems on a real-time basis. These possibilities are the enablers of new transport systems. These increased demands on transport systems and new possibilities ask for a different approach for the logistical control of these transport systems. The research described in this paper focuses on the development of logistic control for a new highly automated transport system, the Underground Logistic System Schiphol(OLS). After this introduction, the next section will provide a short overview of logistic control. The third section will introduce the case study for which the logistic control was designed, the Underground Logistic System Schiphol. Section 4 discusses the implementation of new types of control concepts in the case study. The evaluation and tests of the logistic control are covered in section 5. This paper ends with conclusions and some thoughts on future research on logistic control. II. LOGISTIC CONTROL Control is defined here as a set of mechanisms used to regulate or guide the operation of a machine, apparatus or system. A logistic control system is responsible for controlling the flow of entities (goods, passengers, or even information) through the entire system, fulfilling demands of the (final) customers as much as possible.” A logistic control system has two kind of control activities (Euwe 1999). A logistic control action is a request for an activity in a logistic system on a certain moment in time. A logistic coordination action is a specification of a dependency at a certain moment in time between different control systems. The design of logistic control has become an important part within the field of logistics. In the near future, when the transport systems become more complex and higher performance is needed, it will play an even more important role. Several authors support this development (Roderique 1999, Kulick and Sawyer 1999, and Ryan 1998). III. CASE: UNDERGROUND LOGISTIC SYSTEM SCHIPHOL In the Netherlands around Amsterdam Airport Schiphol and the Flower Auction Aalsmeer the roads are heavily congested. This leads to long throughput times and unreliable delivery rates of the transport of time-critical and expensive airfreight (e. g. flowers, computer parts, newspapers) between Amsterdam Airport Schiphol, logistics centers near Schiphol, the Flower Auction Aalsmeer, and a (future) Rail Terminal near Schiphol. To solve this problem an underground logistic system has been designed for the transport of the expensive and time-critical air-cargo. Because of the separation of

other traffic the transport can be carried out congestion-freetnal (amerailtmin=mini-terminalatSchipholJouble one-waytube (one tubeforeach direction)singleone-waytubewo-way tube (one tubeforbothdirections)Figure I: Overview of the OLS Schiphol.The Underground Logistic System Schiphol (OLS) is highly automated, as it will use AutomaticGuided Vehicles (AGVs) and fully automated transshipment facilities. The OLS will use around 400eight meter long AGVs weighing 10 ton each when it is fully operational.The AGVs are autonomousi. e. they have control responsibility of their own activities. In the early design stages, it was decidedbecauseof scalabilityto maketheAGVsunableto communicate witheachother.Theycanonlycommunicatewithlocal controllersusingawirelessTCP/IPnetwork.Thecontrolofthetransshipmentequipment and management of the facilities are also fully automated.Although there clearly is an economical and environmental need for such an underground logisticsystem, the implementation phase has not yet been started. One of the main obstacles is the largeamountofuncertaintiessurroundingtheimplementationofsuchasystem.Thetechnologythatwillbeused is new, there is almost no experience with automated transport systems of such a large scale, andlittle is known about fully automated control of transport systems of this size.The OLs research project, of which this paper is a result, aims at solving some of theuncertainties for the logistic control of fully automated, large scale transport systems.IV.LOGISTICCONTROLFORTHEUNDERGROUNDThis section discusses some of the choices that were made in the design of the logistic control forthe Underground Logistic System Schiphol. These choices were derived from the demands that thepotential future customers and operators put on the system. Some of the most important choices thathavebeenmadeforthelogistic controlare:ODe-central anddistributedapproachFunctional decomposition

other traffic the transport can be carried out congestion-free. Figure 1: Overview of the OLS Schiphol. The Underground Logistic System Schiphol (OLS) is highly automated, as it will use Automatic Guided Vehicles (AGVs) and fully automated transshipment facilities. The OLS will use around 400 eight meter long AGVs weighing 10 ton each when it is fully operational. The AGVs are autonomous, i. e. they have control responsibility of their own activities. In the early design stages, it was decided because of scalability to make the AGVs unable to communicate with each other. They can only communicate with local controllers using a wireless TCP/IP network. The control of the transshipment equipment and management of the facilities are also fully automated. Although there clearly is an economical and environmental need for such an underground logistic system, the implementation phase has not yet been started. One of the main obstacles is the large amount of uncertainties surrounding the implementation of such a system. The technology that will be used is new, there is almost no experience with automated transport systems of such a large scale, and little is known about fully automated control of transport systems of this size. The OLS research project, of which this paper is a result, aims at solving some of the uncertainties for the logistic control of fully automated, large scale transport systems. IV. LOGISTIC CONTROL FOR THE UNDERGROUND This section discusses some of the choices that were made in the design of the logistic control for the Underground Logistic System Schiphol. These choices were derived from the demands that the potential future customers and operators put on the system. Some of the most important choices that have been made for the logistic control are: ●De-central and distributed approach. ●Functional decomposition

OLayeredhierarchical approachInformationaspects.EachofthechoiceswillbediscussedbelowinmoredetailThe control responsibilities are decentralized and distributed.There is no central controller thatcontrols the entire system. Each component of the system has its own local control responsibilities.Each terminal, AGV, and dock is responsible for the control of its own activities, i.e. there is a highlevelofautonomyorself-control(VanAken1978).Whenthereisapossibleconflictbetweenmoreresources, a higher level control may be needed to avoid dead-locks or conflicts. The higher levelcontrol is responsibleforcoordinatingthecriticalactivitiesof a number of decentralizedcontrollersAnexampleisabidirectional tunnel,wherevehiclesshould notbeallowedtoenterfrombothsidesatonce. Each vehicle has its own vehicle control, but the bi-directional tunnel control can intervenewhen entering the tunnel would lead to deadlocks.The decentralized approach was needed because a number of reasons. The airport authorities donot allow any other party to control vehicles on the airport terrain for safety reasons.The airportauthorities want to control the part of the OLS that is on the airport terrain themselves. Furthermore, adecentralized approach makes the control system very scalable.The system can be scaled-up (or down)to any size, without a substantial loss of efficiency of the logistic control. When more centralizedcontrol systems are scaled up the complexity of the logistic control follows a high-order-polynomialor even an exponential curve (Evers et al 2000). Distributed control is concerned with the execution ofcontrol on geographically distributed computers interconnected via a local area and/or wide areanetwork (Fujimoto 1998; Fujimoto 1999). Each control component has its own hardware and software.Distributed control has a number of advantages. By subdividing the control activities intosub-activities that can be executed concurrently, the execution time of the control can be reduced up toa factor equal to the number of the processors that are used. Furthermore, a lot of information withinlogistic and transport systems is only locally available and thus of a distributed nature. A distributedcontrol system closely represents this and it can take local decisions based on the information that islocally available. If the protocols for information exchange are chosen carefully (e.g. CORBA), thecomputers and software that are used the sub-activities can be made by different manufacturers. Adistributed setting allows the systems to easily replace, remove or add individual controllers, ratherthan creating a new central system after each change that occurs in the overall systemReduction of communication bandwidth was another reason to strive for a high level of autonomy

●Layered hierarchical approach. ●Information aspects. Each of the choices will be discussed below in more detail. The control responsibilities are decentralized and distributed. There is no central controller that controls the entire system. Each component of the system has its own local control responsibilities. Each terminal, AGV, and dock is responsible for the control of its own activities, i.e. there is a high level of autonomy or self-control (Van Aken 1978). When there is a possible conflict between more resources, a higher level control may be needed to avoid dead-locks or conflicts. The higher level control is responsible for coordinating the critical activities of a number of decentralized controllers. An example is a bidirectional tunnel, where vehicles should not be allowed to enter from both sides at once. Each vehicle has its own vehicle control, but the bi-directional tunnel control can intervene when entering the tunnel would lead to deadlocks. The decentralized approach was needed because a number of reasons. The airport authorities do not allow any other party to control vehicles on the airport terrain for safety reasons. The airport authorities want to control the part of the OLS that is on the airport terrain themselves. Furthermore, a decentralized approach makes the control system very scalable. The system can be scaled-up (or down) to any size, without a substantial loss of efficiency of the logistic control. When more centralized control systems are scaled up the complexity of the logistic control follows a high-order-polynomial or even an exponential curve (Evers et al 2000). Distributed control is concerned with the execution of control on geographically distributed computers interconnected via a local area and/or wide area network (Fujimoto 1998; Fujimoto 1999). Each control component has its own hardware and software. Distributed control has a number of advantages. By subdividing the control activities into sub-activities that can be executed concurrently, the execution time of the control can be reduced up to a factor equal to the number of the processors that are used. Furthermore, a lot of information within logistic and transport systems is only locally available and thus of a distributed nature. A distributed control system closely represents this and it can take local decisions based on the information that is locally available. If the protocols for information exchange are chosen carefully (e.g. CORBA), the computers and software that are used the sub-activities can be made by different manufacturers. A distributed setting allows the systems to easily replace, remove or add individual controllers, rather than creating a new central system after each change that occurs in the overall system. Reduction of communication bandwidth was another reason to strive for a high level of autonomy

or self-control, especially in underground transport systems of a large scale. The high level ofautonomy reduces the amount of communication between equipment and controllers. The AGVs onlyhave to communicate with local controllers, like controllers for a safe passing of crossings and dockcontrollers. Within a tunnel communication with the AGVs is difficult. Reducing the need forcommunication in the tunnels reduces the need for special and expensive communication equipment.The high level of autonomy makes the logistic control also very leap.Most activities are notcontrolled by separate logistic controllers, but by the equipment that carries out the activity. This isimportant when in future the OLs is extended to a larger area or even an (underground) logisticsystem for the entire Netherlands. This means that in future several thousand vehicles can drivedistances of 1: unereds of kilometers without major changes in the control systems used. The highlevel of autonomy makes it possible that the vehicles drive these distances entirely by themselves.Aminimum of external control is needed. A further advantage is that when a part of the control systemcommunication system, or physical system brakes down, the other parts of the system can continue tooperate in a normal manner and the autonomous equipment can finish its current activities. Acentralized system will break down (almost) immediately after the controller fails (Rogers andBrennan 1997).The logistic control for the OLS was decomposed in a functional way.Different parts of thesystem that have to be controlled all have their own representation in the logistic control. Some of thecontrol components are: AGV control, transshipment equipment control, parking control, and orderhandling control. This makes the control transparent and extendable. New components can be addedwithoutchangingtheexistingparts.Of course it is necessary to exchange information between different sub-systems now and thenWhen an AGV wants to transship a load, the dock needs to be available at the right time. Instead ofhaving different subsystems communicate directly with each other, a layered hierarchical approachwas used for the logistic control (see figure 2). The decisions that have to be taken by the controlsystem are divided into a number of sub-decisions that are less complex that the 'overal' decision.Furthermore, by using several layers of control it is possible to make changes to single layers of thecontrol system without changing the other layers of the system. Hierarchy divides the power ofdecision among the layers. The top layers have the highest level of power, the lower lavers have lesspower

or self-control, especially in underground transport systems of a large scale. The high level of autonomy reduces the amount of communication between equipment and controllers. The AGVs only have to communicate with local controllers, like controllers for a safe passing of crossings and dock controllers. Within a tunnel communication with the AGVs is difficult. Reducing the need for communication in the tunnels reduces the need for special and expensive communication equipment. The high level of autonomy makes the logistic control also very leap. Most activities are not controlled by separate logistic controllers, but by the equipment that carries out the activity. This is important when in future the OLS is extended to a larger area or even an (underground) logistic system for the entire Netherlands. This means that in future several thousand vehicles can drive distances of 1: unereds of kilometers without major changes in the control systems used. The high level of autonomy makes it possible that the vehicles drive these distances entirely by themselves. A minimum of external control is needed. A further advantage is that when a part of the control system, communication system, or physical system brakes down, the other parts of the system can continue to operate in a normal manner and the autonomous equipment can finish its current activities. A centralized system will break down (almost) immediately after the controller fails (Rogers and Brennan 1997). The logistic control for the OLS was decomposed in a functional way. Different parts of the system that have to be controlled all have their own representation in the logistic control. Some of the control components are: AGV control, transshipment equipment control, parking control, and order handling control. This makes the control transparent and extendable. New components can be added without changing the existing parts. Of course it is necessary to exchange information between different sub-systems now and then. When an AGV wants to transship a load, the dock needs to be available at the right time. Instead of having different subsystems communicate directly with each other, a layered hierarchical approach was used for the logistic control (see figure 2). The decisions that have to be taken by the control system are divided into a number of sub-decisions that are less complex that the 'overall' decision. Furthermore, by using several layers of control it is possible to make changes to single layers of the control system without changing the other layers of the system. Hierarchy divides the power of decision among the layers. The top layers have the highest level of power, the lower lavers have less power

Manager(1:n)Order assignmentScriptdispatchescriptEvent (s)do(pi...Pn)E(pi..Pn)Control(1:1)EventHandlerScript interpreterCommandEvent (s)do(p.p.)E(pl..p,)Equipment(physicalresource)Command handleEventgeneratorFigure2:Hierarchical layered control withasynchronous communicationbetween thelayersThe high-level logistic control is kept as generic 9s possible. The clear advantage of this is thatindependently of the kind of equipment that is used, the high-level logistic control will stay the same.All types of resources such as AGVs and transshipment facilities are controlled in a similar way. Onlythe equipment layer is specific for each kind of equipment. Because of the layered and genericapproach, equipment of several manufacturers can work besides each other without any changes in thehigh-level logistic control. This makes the OLS independent of single equipment manufacturers.The clear interfaces between the different layers (see figure 2) are nothing more than a descriptionof what kind of messages are exchanged between the different controls components.For the logisticcontrol it doesnotmatterhow componentsfunction internally,aslongasthey complytothedefinedinterfaces. The layered approach also makes it possible for different researchers to work:simultaneouslyonthecontrol.Withinthedesignphaseofthelogistic controlalotof attentionhasbeen paid to information exchange.The information is distributed, minimized, event-drivenasynchronous, and sometimes redundantly available. Since the OLS covers quite a large geographicalarea andis almost entirely underground,most information is onlylocally available.Thelogisticcontrol is built in such a way that it only needs the information that is really locally available. Thecommunication is event-driven,Communication only occurs when the control system or theequipment has reached a certain stage in the execution of activities. This minimizes the amount ofcommunication that isneeded,and no processes are activelywaiting after issuinga command. The

Figure 2: Hierarchical layered control with asynchronous communication between the layers The high-level logistic control is kept as generic 9s possible. The clear advantage of this is that independently of the kind of equipment that is used, the high-level logistic control will stay the same. All types of resources such as AGVs and transshipment facilities are controlled in a similar way. Only the equipment layer is specific for each kind of equipment. Because of the layered and generic approach, equipment of several manufacturers can work besides each other without any changes in the high-level logistic control. This makes the OLS independent of single equipment manufacturers. The clear interfaces between the different layers (see figure 2) are nothing more than a description of what kind of messages are exchanged between the different controls components. For the logistic control it does not matter how components function internally, as long as they comply to the defined interfaces. The layered approach also makes it possible for different researchers to work: simultaneously on the control. Within the design phase of the logistic control a lot of attention has been paid to information exchange. The information is distributed, minimized, event-driven, asynchronous, and sometimes redundantly available. Since the OLS covers quite a large geographical area and is almost entirely underground, most information is only locally available. The logistic control is built in such a way that it only needs the information that is really locally available. The communication is event-driven. Communication only occurs when the control system or the equipment has reached a certain stage in the execution of activities. This minimizes the amount of communication that is needed, and no processes are actively waiting after issuing a command. The

event based system also makes it easier to trigger time-outs by independent monitoring processes.Whenever necessary for recovery after a break-down of any part of the system, redundancy ininformation storage was introduced. Of course, this has to be done in such a way that recovery isindeed possible for themost occurring situations.On the other hand this should not lead to much extracommunication.V.EVALUATIONOFTHE LOGISTICCONTROLSYSTEMIn order to review and validate the potential success of transport systems and to test and tune theircontrol systems, computer simulation is increasingly used as a key tool (Kia et al. 20o0; Kulick andSawyer 1999, Ryan 1998).However, not all technical aspects can be fully tested by simulationExamples are the physical interactions between different types of equipment, and communicationbetween subsystems.Tests with real equipment or models of the equipment are also needed. Auipgeret al. (1999) suggest to use a combination of simulation and reality to test complex control systemsIn the OLS project, several simulation models were made of the logistic control and the physicalcomponents of the OLS to evaluate the effectiveness of the hierarchical logistic control system(Saanen et al. 2000,Verbraeck et al. 2000). The logistic control system could be tested in a fullysimulated environment (seefigure3).In order to test the control system architecture in the simulations, it was important to keep thesimulated control system components and layers as close as possible to the real control systemcomponents. Therefore, a one-to-one translation of the control system parts was made in the objectoriented simulation language Simple++(Tecnomatix, 1999).All interfaces between the sub-systemswere implemented with CORBA similarly to how they would be realized in the final system. Thesimulation implementation therefore offered the possibility to distribute the control responsibilitiesand simulated physical system parts over more computers. One problem, however, was that it provedto be difficult to synchronize the clocks of the distributed simulation models (Fujimoto 1999), becausethe Simple++ language offers no possibilities for clock synchronization in distributed simulations.Therefore, we decided to keep the model of the physical system in one simulation model with an eventcontroller, and have the models of the control sub-systems function as timeless state machines. Theinteraction between the models was implemented using CORBA according to the asynchronouscommand / event mechanism as depicted in figure 2. After the tests with the simulated resources inSimple++, more advanced emulations of the equipment were built in C++ which could easily becoupledtothesimulationsviaCORBA

event based system also makes it easier to trigger time-outs by independent monitoring processes. Whenever necessary for recovery after a break-down of any part of the system, redundancy in information storage was introduced. Of course, this has to be done in such a way that recovery is indeed possible for the most occurring situations. On the other hand this should not lead to much extra communication. V. EVALUATION OF THE LOGISTIC CONTROL SYSTEM In order to review and validate the potential success of transport systems and to test and tune their control systems, computer simulation is increasingly used as a key tool (Kia et al. 2000; Kulick and Sawyer 1999; Ryan 1998). However, not all technical aspects can be fully tested by simulation. Examples are the physical interactions between different types of equipment, and communication between subsystems. Tests with real equipment or models of the equipment are also needed. Auipger et al. (1999) suggest to use a combination of simulation and reality to test complex control systems. In the OLS project, several simulation models were made of the logistic control and the physical components of the OLS to evaluate the effectiveness of the hierarchical logistic control system (Saanen et al. 2000,Verbraeck et al. 2000). The logistic control system could be tested in a fully simulated environment (see figure 3). In order to test the control system architecture in the simulations, it was important to keep the simulated control system components and layers as close as possible to the real control system components. Therefore, a one-to-one translation of the control system parts was made in the object oriented simulation language Simple++(Tecnomatix, 1999).All interfaces between the sub-systems were implemented with CORBA similarly to how they would be realized in the final system. The simulation implementation therefore offered the possibility to distribute the control responsibilities and simulated physical system parts over more computers. One problem, however, was that it proved to be difficult to synchronize the clocks of the distributed simulation models (Fujimoto 1999), because the Simple++ language offers no possibilities for clock synchronization in distributed simulations. Therefore, we decided to keep the model of the physical system in one simulation model with an event controller, and have the models of the control sub-systems function as timeless state machines. The interaction between the models was implemented using CORBA according to the asynchronous command / event mechanism as depicted in figure 2. After the tests with the simulated resources in Simple++, more advanced emulations of the equipment were built in C++ which could easily be coupled to the simulations via CORBA

Figure3:Simulation1animation screenofoneof thesimulationmodelsBesides the simulation models a laboratory, called the TestSite, was constructed at DelftUniversity of Technology.The TestSite is a special area of 1600 m equipped with scale models (1:3)of the equipment (AGVs and docks) that will be used for the OLS(see figure 4). Furthermore,prototypes (scale 1:1) of the AGVs and docks are also available in the laboratory.Figure 4:Control systemtests at theTestSitewith1:3scale models of theAGVsAgain, because of the clear and lean interfaces between the different control components andbetween the control and equipment, it proved to be easy to control the real equipment at the TestSitewith the simulation models of the control system. The control system models can exchange theirasynchronous CORBA commands and events with any type of simulated or real equipment. TheAGVs and docks at the TestSite were given CORBA communication or they were equipped withCORBA wrappers written in C++. It even proved possible to control both the simulated equipmentand the scale models of the equipment simultaneously (Verbraeck and Versteegt, 2000). Thereby,simulated and real vehicles could 'drive' in the same space, and experiments with more vehicles thancurrentlypresentattheTestSitebecamepossible

Figure 3: Simulation 1 animation screen of one of the simulation models Besides the simulation models a laboratory, called the TestSite, was constructed at Delft University of Technology. The TestSite is a special area of 1600 m 2 equipped with scale models (1:3) of the equipment (AGVs and docks) that will be used for the OLS(see figure 4). Furthermore, prototypes (scale 1:1) of the AGVs and docks are also available in the laboratory. Figure 4 : Control system tests at the TestSite with 1:3 scale models of the AGVs Again, because of the clear and lean interfaces between the different control components and between the control and equipment, it proved to be easy to control the real equipment at the TestSite with the simulation models of the control system. The control system models can exchange their asynchronous CORBA commands and events with any type of simulated or real equipment. The AGVs and docks at the TestSite were given CORBA communication or they were equipped with CORBA wrappers written in C++. It even proved possible to control both the simulated equipment and the scale models of the equipment simultaneously (Verbraeck and Versteegt, 2000). Thereby, simulated and real vehicles could ‘drive’ in the same space, and experiments with more vehicles than currently present at the TestSite became possible

The tests at the Test Site showed that the logistic control for the OLS functioned as expected. Ofcourse there turned out to be differences between the simulated equipment and the real equipment,such as the driving precision and computation speed, but these did not impactthe main principles behind the automatic transport control systemVI.CONCLUSIONSANDFUTURERESEARCHThe decentralized and distributed approach works very well for the logistic control of highlyautomated, large-scale transport systems. Furthermore, decentralization and distribution lead to aflexibleand scalablelogisticcontrol.The combination of simulation models and a laboratory with scaled and full size equipmentproves to be a strong tool to evaluate and test logistic control for highly automated transport systems.Further tests for the OLS will focus on real operating conditions with disturbances, order interfaces tothereal world,and a largenumber of real equipment shouldultimatelyprovethefeasibility and theadvantages of the logistic control. The logistic control will be used for other research projects oncomplex transport systems as well. Examples are: underground city distribution systems, automatedpeople movers, and automated container terminals. At this moment tests are being carried out to testhowwell theconcepts workfor othertransport systems.Thepreliminaryresults seemverypromisingHopefully, the designed logistic control principles can be generalized for many types of automatictransportsystems

The tests at the Test Site showed that the logistic control for the OLS functioned as expected. Of course there turned out to be differences between the simulated equipment and the real equipment, such as the driving precision and computation speed, but these did not impact the main principles behind the automatic transport control system. VI.CONCLUSIONS AND FUTURE RESEARCH The decentralized and distributed approach works very well for the logistic control of highly automated, large-scale transport systems. Furthermore, decentralization and distribution lead to a flexible and scalable logistic control. The combination of simulation models and a laboratory with scaled and full size equipment proves to be a strong tool to evaluate and test logistic control for highly automated transport systems. Further tests for the OLS will focus on real operating conditions with disturbances, order interfaces to the real world, and a large number of real equipment should ultimately prove the feasibility and the advantages of the logistic control. The logistic control will be used for other research projects on complex transport systems as well. Examples are: underground city distribution systems, automated people movers, and automated container terminals. At this moment tests are being carried out to test how well the concepts work for other transport systems. The preliminary results seem very promising. Hopefully, the designed logistic control principles can be generalized for many types of automatic transport systems