Search

In this paper we consider the optimization of integral fixed-time (traffic light) schedules for isolated intersections. Such integral schedule has an integral period duration and, moreover, for this schedule all switches to green, yellow and red are scheduled at an integral second. These schedules are desired in practice because they are clear, presentable and easy to work with. Much research has been devoted to the optimization of fixed-time (traffic light) schedules for isolated intersection. However, these methods optimize over all fixed-time schedules and do not necessarily return an integral one. One can simply obtain an integral schedule from any fixed-time schedule by rounding the period duration and the switching times to the nearest integer. However, such an approach might result in an integral schedule that is infeasible. In this paper we give a more rigourous approach to find the optimal integral fixed-time schedule that minimizes the average delay that road users experience at the intersection.

We generalize the standard multi-class queueing network model by allowing both standard queues and infinite virtual queues which have an infinite supply of work. We pose the general problem of finding policies which allow some of the nodes of the network to work with full utilization, and yet keep all the standard queues in the system stable. Toward this end we show that re-entrant lines, systems of two re-entrant lines through two service stations, and rings of service stations can be stabilized with priority policies under certain parameter restrictions. The analysis throughout the paper depends on model and policy and illustrates the difficulty in solving the general problem.

In this paper we present an aggregate simulation model for a real-life emergency department, which is based on the concept of effective process times and which uses a token system to model patients claim- ing multiple resources simultaneously. Although it has been developed for a specific hospital, the model is exible, and capable to describe different set- Tings. The modeling steps, model specification and model validation are explained in detail. By using a process-based simulation language, the resulting model is transparent, intuitive and easy to use in quantitatively evaluating proposed changes in the operational processes of the emergency department.

only.

In this chapter we will present three approaches to model manufacturing systems in an aggregate way leading to fast and effective (i.e., scalable) simulations that allow the development of simulation tools for rapid exploration of different production scenarios in a factory as well as in a whole supply chain. We will present the main ideas and show some validation studies. Fundamental references are given for more detailed studies.

The paper considers standard fluid models of multi-product multiple-server production systems where setup times are incurred whenever a server changes product. Concerned is the general approach to the problem of optimizing the long-run average cost per unit time that offers to first determine an optimal steady state (periodic) behavior and then to design a feedback scheduling protocol ensuring convergence to this behavior as time progresses. The second part of this program is treated and a systematic presentation of a novel approach to it is offered. This approach gives rise to protocols that are cyclic and distributed: the servers do not need information about the entire system state. Each of them proceeds basically from the local data concerning only the currently served queue, although a fixed finite number of one-bit notification signals should be exchanged between the servers during every cycle. The approach is illustrated by simple albeit instructive examples concerning polling systems, single server systems with processor sharing scheme, and the re-entrant two-server manufacturing network with non-negligible setup times introduced by Kumar and Seidman. For the last network considered in the analytical form, the optimal steady-state (periodic) behavior is first determined. Based on the desired steady state behavior and the presented theory, we designed simple distributed feedback switching control laws for all examples. These laws not only obtain the required behaviors but also make them globally attractive, irrespective of the system parameters and initial state.

We generalize the standard multi-class queueing network model by allowing both standard queues and in??nite virtual queues which have infinite supply of work. We pose the general problem of finding policies which allow some of the nodes of the network to work with full utilization, and yet keep all the standard queues in the system stable. Towards this end we show that re-entrant lines, systems of two re-entrant lines through two service stations, and rings of service stations can be stabilized with priority policies under certain parameter restrictions. We further establish simple diffusion limits for the departure and work allocation processes. The analysis throughout the paper depends on model and policy and illustrates the difficulty in solving the general problem.

A third party logistic provider operates a distribution network with minimal control over supply and demand. The operation is characterized by three levels: a strategic level that includes the location and sizing of warehouses, a tactical level that determines the links between customers, warehouses and producers and an operational level that determines the size of the flows through the links at any given time. An algorithm to optimize the operational level is determined for a given network structure. Starting with a fully connected network a reduced network is found by deleting the least used links until the operational costs increase dramatically. The topological structures of the reduced network as a function of backlog and transportation costs are determined.

No abstract.

Performance evaluation using queuing models is common practice in semiconductor manufacturing. Analytical closed-form expressions and simulations models are popular in capacity planning and the analysis of equipment configurations. However, the complexity of semiconductor processes complicates the modeling of the equipment. Analytical models lack the required accuracy, whereas simulation models require too many details, making them impractical. Aggregation is a way to overcome this difficulty. The various details are not modeled in detail, but their contribution is lumped in the aggregate model, which makes the model more appropriate for both analysis and simulation. This paper gives an overview of our efforts to develop a top-down aggregate modeling approach for semiconductor equipment, starting from the effective process time concept inspired by the Factory Physics book of Hopp and Spearman. In our modeling approach the aggregate model parameters can be estimated directly from industrial data, without the need to quantify the various details.

This paper considers adaptability for robust periodic capacity allocation problems, usingthe example of crane capacity allocation in container terminals. A new model is presented that allowsto compute an adaptive plan via formulation of a linear program over a scenario graph. Computationalresults are given that quantify to the value of adaptability for reducing the price of robustness, i.e, thetotal capacity that is necessary to guarantee all service times for all considered realizations of the disturbances.

We consider a container port, which consists of multiple terminals all run by the same terminaloperator. The terminal operator provides a number of vessel lines with the logistics services of container unloading,loading, transporting and storing. He faces the combined decisions on strategic, tactical, and operationallevel to provide a certain service level while running its operations with the smallest possible amount of resources.Since these combined decisions are often too complex to be solved at once, the problem is often dividedinto a number of subproblems, which are then solved subsequently, or alternatingly where possible. Existingstudies subdivide the combined problems in container operations according to the common categorization andsubsequently solve strategic, tactical and operational problems. In this research however, we propose a typicallydifferent subdivision, such that each subproblem i) can be solved within the time allowed, ii) has not yet beenaddressed in literature, an iii) is practically interesting in its own right. In this paper, we address this uniquesubdivision in general and one of the resulting subproblems in particular: minimizing the fuel usage of thestraddle carriers by intelligent berth and terminal planning. This research has financially been supported bythe terminal operator PSA HNN in Antwerp, Belgium. Results of a case study suggest significant reductions inthe fuel usage of the straddle carries. Currently, results of this research have actually been implemented in oneof the terminals in Antwerp.

In this paper we consider output feedback tracking of ships with position and orientation measurements only. Ship dynamics are highly nonlinear, and for tracking control, as opposed to dynamic positioning, these nonlinearities have to be taken into account in the control design. We propose an observer-controller scheme which takes into account the complete ship dynamics, including Coriolis and centripetal forces and nonlinear damping, and prove that the closed-loop system is semi-globally uniformly stable. Furthermore, a gain tuning procedure for the observer controller scheme is developed. Experimental results are presented where the observer-controller scheme is implemented onboard a Froude scaled 1:70 model supply ship. The experimentally obtained results are compared with simulation results under ideal conditions and both support the theoretical results on semi-global exponential stability of the closed-loop system.

Predicting the cycle time distribution as a function of throughput is helpful in making a trade-off between workstation productivity and meeting due dates. To predict cycle time distributions, detailed models are almost exclusively used, which require considerable development and maintenance effort. Instead, we propose a so-called aggregate model to predict cycle time distributions, which is a lumped-parameter representation of the queueing system. The lumped parameters of the model are determined directly from arrival and departure events measured at the workstation. The paper demonstrates that the aggregate model can accurately predict the cycle time distribution of workstations in semiconductor manufacturing, in particular the tail of the distribution

For manufacturing workstations, there is a trade-o ff between throughput and meeting product due dates.Models for predicting flow time distributions are helpful in this respect. A model of a workstation typically has to include the process time distributions of the machines in the workstation, the dispatching rule, and other disturbances that aff ect the flow time performance. A detailed simulation model may be used to predict the flow time distributions of a workstation. However, such a model typically involves considerable development time, and obtaining all model parameters may be difficult. In this paper we propose a single-server aggregate model with a workload-dependent aggregate process time distribution, and a workload-dependent overtakingdistribution. The model involves little development time. The process time and overtaking ditribution in the aggregate model are determined from lot arrival and departure events measured at the considered workstation. Two simulation test cases are included to demonstrate the proposed method.

This research is focused on using a highly detailed simulation model to create a physical security system to prevent intrusions in a building. Security consists of guards and security cameras. The problem is represented as a binary optimization problem. A new heuristic is proposed to do the security configuration optimization, the Hybrid Genetic Ant Optimization (HGAO). This heuristic combines a genetic algorithm (GA) and an ant colony optimization (ACO). HGAO is compared with GA for two security optimization scenarios.

In this report we consider the optimal control problem of emptying a deterministic single server multiclass queuing system without arrivals. We assume that the server is able to serve several queues simultaneously, each at its own rate, independent of the number of queues being served. We showed that the optimal sequence of modes is ordered by rate of cost decrease. However, queues are not necessarily emptied. We propose a dynamic programming approach for solving the problem, which reduces the multiparametric QP (mpQP) to a series of problems that can be solved readily.

We consider a container operator, who serves a number of shipping lines by discharging and loading their periodically arriving container vessels. Disruptions on vessels’ travel times lead to stochastic arrivals in the port. To cope with these disturbances, the operator and each vessel line agree on two types of arrivals: arrivals i) within, and ii) out of a so-called arrival window. If a vessel arrives within its window, the operator guarantees a maximal process time. If not, the operator is not bound to any guaranteed process time. The problem is to construct a periodic window-based i) arrival, ii) departure and iii) time-variant crane capacity plan to minimize the maximal crane capacity reservation. In this paper, we propose a mixed integer linear program (MILP) that minimizes the maximal crane capacity reservation while window agreements are satisfied for all scenarios in which vessels arrive within their windows. Results of a case study suggest that with slight modifications to an existing plan, significant reductions in the maximal crane capacity reservation can be achieved. As a particular case, the MILP determines the conventional optimal window-ignoring plan. Results suggest that although the windowignoring plan on itself requires less crane capacity than the window-based plan, it is much more sensitive to the arrival window agreements.

We present an MILP formulation to construct a strategic, cyclic terminal allocation and time scheduling for a set of container vessels, which are processed by a multi-terminal port operator. Our case study is about balancing the workload over time, minimizing inter-terminal transport and minimizing deviations from the existing time schedule.

The dynamics of manufacturing systems has been a subject of study for several decades. Over the last years, manufacturing systems have become more and more complex. A good understanding of the dynamics of manufacturing systems has therefore become even more important. The goal of this report is to introduce a large variety of models that have been used for modeling manufacturing systems. By means of examples it is illustrated how certain modeling techniques can be used to derive models that can be used for analysis and/or control. Also, references are provided that can be used as a starting point for getting more details. Furthermore, recent developments in the modeling, analysis and control of manufacturing systems are presented.

We present a case study based on a multi-terminal container operation in Antwerp, Belgium, where a set of cyclically calling container vessels is processed. The operator faces the problem of strategically allocating a terminal, a berthing interval, and a variable number of quay cranes to the vessels in the set. Restricting properties are terminal quay lengths, number of quay cranes and storage capacities. Currently, the operator's objective is to satisfy the preferences of the vessel lines, with respect to a terminal and berthing time, as much as possible. We are interested in the benefit of modifying a given allocation, i.e. the potential crane and inter-terminal costs savings if specific changes to a given allocation are allowed. An MILP is implemented in a two-step optimization, which enables us to efficiently investigate the benefit of modification. Experimental results suggest that small changes in a given allocation may lead to significant cost savings.

The objective of this report is to develop a method to control a manufacturing system using feedback control. The manufacturing system of interest is a three machine multi-product flow line with stochastic process times, producing eight different product types in large quantities. The performance of the feedback controlled flow line is compared to the performance of the flow line controlled by several common control methods: push, pull, conwip, and POLCA control. The main performance criteria for the flow line are assumed to be the throughput of the system and the inventory level.

In this report we will present three approaches to model manufacturing systems in an aggregate way leading to fast and effective (i.e., scalable) simulations that allow the development of simulation tools for rapid exploration of different production scenarios in a factory as well as in a whole supply chain. We will present the main ideas and show some validation studies. Fundamental references are given for more detailed studies.

In this paper we consider a port consisting of a cluster of inter-related terminals, where container vessels arrive cyclically. The problem is to strategically assign a terminal and a time interval of berthing to each of the vessels in the cycle. Restricting properties are terminal quay lengths and quay crane capacity. Conflicting objectives are i) minimizing the number of required quay cranes, ii) minimizing the amount of inter-terminal traffic and iii) minimizing the total weighted deviation from desired berthing intervals. We formulate both a straightforward and an alternative mixed integer linear program to model this system. Results show that the alternative model is much faster solvable and enables to optimize real-life problems within a couple of hours.

The dynamics of manufacturing systems has been a subject of study for several decades. Over the last years, manufacturing systems have become more and more complex. A good understanding of the dynamics of manufacturing systems has therefore become even more important. The goal of this report is to introduce a large variety of models that have been used for modeling manufacturing systems. By means of examples it is illustrated how certain modeling techniques can be used to derive models that can be used for analysis and/or control. Also, references are provided that can be used as a starting point for getting more details. Furthermore, recent developments in the modeling, analysis and control of manufacturing systems are presented.

In this paper we study the control of switching servers, which can for example be found in manufacturing industry. In general, these systems are discrete event systems. A server processes multiple job types. Switching between the job types takes time and during that time, no jobs can be processed, so capacity is lost. How should a server switch between the job types in an efficient way? In this paper we derive the optimal process cycle with respect to work in process levels for a server with two job types and finite buffer capacities. The analysis is performed using a hybrid fluid model approximation. After the optimal process cycle has been defined, a state feedback controller is proposed that steers the trajectory of the system to this optimal cycle. Workstations are often placed in series to form a flowline of servers. Our goal is to control flowlines of switching servers in a way that the work in process level is minimized. In a flowline, only the most downstream workstation influences the work in process level of the system, since upstream workstations simply move jobs from one server to the other. If it is possible to have the most downstream workstation process in its optimal cycle and the other workstations can make this happen, then optimal work in process levels are achieved. This paper investigates under which conditions the upstream workstations can make the most downstream workstation work optimally. Conditions on the upstream workstations are derived and the class of flowlines is characterized for which the optimal process cycle of an isolated downstream workstation can become the optimal process cycle for the flowline. For a flowline consisting of two workstations, a state feedback controller is proposed and convergence to the optimal process cycle is proved mathematically. An extensive case study demonstrates how the controller performs, for both the hybrid fluid model and in a discrete event implementation with stochastic inter-arrival and process times.

only.

This contribution deals with the modelling of manufacturing systems forcontrol. First the concept of effective process times is introduced as ameans to arrive at relatively simple discrete event models ofmanufacturing systems based on measured data. Secondly, a controlframework is presented. Thirdly, a validation study is presented whichshows that the currently available PDE-models for describing manufacturingsystems need further improvement. Finally some criteria are specified which a PDE-model should at least meet in order to be considered valid.

Current literature on modeling and control of manufacturing systems can roughly be divided into three groups: flow/fluid models, queueing theory, and discrete event models. Most fluid models describe linear time-invariant controllable systems without any dynamics. These models mainly focus on throughput and are not concerned with cycle time. Queuing theory deals with relationships between throughput and cycle time, but is mainly concerned with steady-state analysis. In addition, queueing models are not suitable for control theory. Discrete event models suffer from ``state-explosion''. Simple models of manufacturing systems can be studied and analyzed, but for larger problems the dimension of the state grows exponentially. In addition, most control problems studied are supervisory control problems: the avoidance of undesired states. An important class of interesting manufacturing control problems asks for proper balancing of both throughput and cycle time for a large nonlinear dynamical system that never is in steady state. None of the mentioned models is able to deal with these kind of control problems. In this paper, models are presented with are suitable for addressing this important class of interesting manufacturing control problems.

In this paper a trajectory tracking problem for the three-dimensional kinematic model of the unicycle-type mobile robot is considered. It is assumed that only two of the tracking error coordinates are measurable. By means of a cascaded systems theory we develop observers for each of the error coordinates and show K-exponential convergence of the tracking error in the combined closed-loop observer-controller systems. The results are illustrated with computer simulations.

The subject of this thesis is the design of tracking controllers for certain classes of mechanical systems. The thesis consists of two parts. In the first part an accurate mathematical model of the mechanical system under consideration is assumed to be given. The goal is to follow a certain specified trajectory. Therefore, a feasible reference trajectory is assumed to be given i.e., a trajectory that can be realized for the system under consideration. The tracking error at each time is defined as the difference between where the system is and where it should be. The problem now is to design a controller for the system which is such that the tracking error converges to zero, no matter where the system is initialized nor at which time-instant. A new design methodology is presented, based on the theory of cascaded systems, i.e., systems that can be seen as a special interconnection of two stable subsystems. This new approach is applied to three different models.

In this note we address the problem of set-point control of robot manipulators with uncertain gravity knowledge by combining several previous contributions to PID control. The main contribution is a linear PID controller which ensures global asymptotic stability of the closed loop. The key feature of the controller is that the integration is started after a short transient. In the case of unmeasurable velocities, a similar ``delayed'' PI2D controller is shown to globally asymptotically stabilize the manipulator.