Agent-Based Distributed Manufacturing Process Planning and Scheduling: A State-of-the-Art Survey

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 36, NO.
4, JULY 2006 563
Agent-Based Distributed Manufacturing Process

Planning and Scheduling: A State-of-the-Art Survey
Weiming Shen, Senior Member, IEEE, Lihui Wang, and Qi Hao
Abstract—Manufacturing process planning is the process of se- ingly important for manufacturing enterprises to increase their
lecting and sequencing manufacturing processes such that they productivity and profitability through greater shop floor agility
achieve one or more goals and satisfy a set of domain constraints. to survive in a globally competitive market [98].
Manufacturing scheduling is the process of selecting a process plan
and assigning manufacturing resources for specific time periods to This paper describes the complexity of manufacturing
the set of manufacturing processes in the plan. It is, in fact, an op- process-planning and scheduling problems (Section II), and re-
timization process by which limited manufacturing resources are views the research literature in manufacturing process planning
allocated over time among parallel and sequential activities. Manu- (Section III), manufacturing scheduling (Section IV), and the
facturing process planning and scheduling are usually considered integration of process planning and scheduling (Section V),
to be two separate and distinct phases. Traditional optimization
approaches to these problems do not consider the constraints of particularly focusing on agent-based approaches in these areas.
both domains simultaneously and result in suboptimal solutions. Major issues in these research areas are discussed (Section VI),
Without considering real-time machine workloads and shop floor research opportunities and challenges addressed (Section VII),
dynamics, process plans may become suboptimal or even invalid and a brief conclusion stated (Section VIII).
at the time of execution. Therefore, there is a need for the integra- The objective of this paper is not to provide an extensive sur-
tion of manufacturing process-planning and scheduling systems for
generating more realistic and effective plans. After describing the vey of general manufacturing process-planning and scheduling
complexity of the manufacturing process-planning and scheduling systems, but to focus on the agent-based approaches and their
problems, this paper reviews the research literature on manufac- applications in manufacturing process planning and scheduling.
turing process planning, scheduling as well as their integration, An earlier survey of multiagent systems for intelligent manu-
particularly on agent-based approaches to these difficult problems. facturing systems, including agent-based manufacturing process
Major issues in these research areas are discussed, and research
opportunities and challenges are identified. planning, scheduling, and control, can be found in [92]. More
discussions on the applications of agent technology to collabo-
Index Terms—Agents, distributed manufacturing systems, man- rative design and manufacturing can be found in [94].
ufacturing scheduling, multiagent systems, process planning.
I. INTRODUCTION II. PROBLEM COMPLEXITY

ANUFACTURING process planning and scheduling are
M usually considered to be two separate activities in man-
ufacturing. Manufacturing process planning determines how a
The problem of manufacturing process planning and schedul-
ing has been introduced in Section I. This section discusses the
complexity of the problem and the difficulty in solving it.
product will be manufactured. It is the process of selecting and The scheduling problem exists not only in manufacturing
sequencing manufacturing processes and parameters so that they enterprises, but also in organizations like publishing houses,
achieve one or more goals (e.g., lower cost, shorter processing universities, hospitals, airports, and transportation companies.
time, etc.) and satisfy a set of domain constraints. Manufactur- It is typically NP-hard, i.e., it is impossible to find an optimal
ing scheduling, on the other hand, is the process of assigning solution without the use of an essentially enumerative algorithm,
manufacturing resources over time to the set of manufacturing with computation time increasing exponentially with problem
processes in the process plan. It determines the most appropriate size. However, the manufacturing scheduling problem is one
time to execute each operation, taking into account the temporal of the most difficult of all scheduling problems. More detailed
relationships between manufacturing processes and the capac- discussions and analyses of scheduling problems can be found
ity limitations of the shared manufacturing resources. The as- in [5], [29].
signments also affect the optimality of a schedule with respect A well-known manufacturing scheduling problem is the clas-
to criteria such as cost, tardiness, or throughput. In summary, sical job shop scheduling where a set of jobs and a set of ma-
scheduling is an optimization process where limited resources chines are given. Each machine can handle at most one job at a
are allocated over time among both parallel and sequential activ- time. Each job consists of a chain of operations, each of which
ities [136]. Such an optimization process is becoming increas- needs to be processed during an uninterrupted time period of
given length on a given machine. The purpose is to find the best
Manuscript received December 5, 2003; revised February 3, 2005 and June schedule, i.e., an allocation of the operations to time intervals
29, 2005. This paper was recommended by Associate Editor V. Marik. on the machines, that has the minimum total duration required
The authors are with Integrated Manufacturing Technologies Institute, Na- to complete all jobs. The total number of possible solutions
tional Research Council Canada, London, ON, Canada (e-mail: weiming.shen@
nrc.gc.ca; lihui.wang@nrc.gc.ca; qi.hao@nrc.gc.ca). for a classical job shop scheduling problem with n jobs and m
Digital Object Identifier 10.1109/TSMCC.2006.874022 machines is (n!)m [5].
1094-6977/$20.00 © 2006 Canadian Crown Copyright
564 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 36, NO. 4, JULY 2006
The problem becomes even more complex in the following artificial intelligence techniques. They are oriented toward the
situations. needs of large companies, especially those producing products
1) When other manufacturing resources, such as operators with large variety and small batch sizes. However, a truly gen-
and tools, are also considered during the scheduling pro- erative process-planning system that can meet industrial needs
cess. For a classical job shop scheduling problem with n and provide an appropriate generic framework, knowledge rep-
jobs, m machines, and k operators, the total number of resentation methods, and inference mechanisms has not been
possible solutions could be ((n!)m )k . developed so far [134].
2) When both process planning and manufacturing schedul- Various approaches to CAPP have been proposed in the
ing are to be done at the same time. Traditional approaches literature [2], [25]. Research studies on process planning in-
that treat process planning and manufacturing scheduling clude object-oriented approaches [105], [132], GA-based ap-
separately can result in suboptimal solutions for the two proaches [70], [131], neural-network-based approaches [21],
phases. Integrating the two phases into one optimization [69], Petri net-based approaches [53], feature recognition or
problem, by considering the constraints of both domains feature-driven approaches [114], [119], and knowledge-based
simultaneously, can theoretically result in a global optimal approaches [108], [118]. These approaches and their combina-
solution, but it increases the solution space significantly. tions have been applied to some specific problem domains, such
3) When unforeseen dynamic situations are considered. In a as tool selection [24], [56], tool path planning [7], [45], machin-
job shop manufacturing environment, rarely do things go ing parameters selection [3], [37], process sequencing [129],
as expected. The system may be asked to include addi- and setup planning [75], [125].
tional tasks that are not anticipated, or to adapt to changes Recently, the research focus on process planning has shifted
to several tasks, or to neglect certain tasks. The resources toward solving problems in distributed manufacturing environ-
available for performing tasks are subject to changes. Cer- ments. Tu et al. [115] introduced a method called incremental
tain resources can become unavailable, and additional re- process planning (IPP) for one-of-a-kind production (OKP) in
sources can be introduced. The beginning time and the such environments. The IPP is used to extend or modify a prim-
processing time of a task are also subject to variations. A itive plan (a skeletal process plan) incrementally according to
task can take more or less time than anticipated, and tasks new features that are identified from a product design until no
can arrive early or late. Other uncertainties include power more new features can be found. A complete process plan gen-
system failures, machine failures, operator absence, and erated by the IPP may include alternative processes.
unavailability of tools and materials. An optimal schedule,
generated after considerable effort, may rapidly become
unacceptable because of unforeseen dynamic situations B. Agent-Based Approaches
on the shop floor and a new schedule may have to be gen- Apart from centralized AI approaches [e.g., genetic algo-
erated. This kind of rescheduling problem is also called rithms (GAs), neural networks, fuzzy logic, and expert systems],
dynamic scheduling or real-time scheduling. agent technology is emerging as a solution for distributed AI that
has attracted a wide attention. Instead of being one large expert
III. APPROACHES TO MANUFACTURING PROCESS PLANNING system, cooperative intelligent agents are being used in devel-
oping distributed CAPP systems. The agent-based approach is
A. Traditional Approaches
also being recognized as an effective way to realize adaptive-
Traditionally, manufacturing process planning is a task that ness and dynamism of process planning. The following are some
transforms design information into manufacturing processes and examples of agent-based process-planning systems.
determines the sequence of operations [15]. Maintaining the 1) Shih and Srihari [99] proposed a distributed AI-based
consistency of process plans and keeps them optimized is a framework for process planning. Their approach decom-
difficult task. Since 1965, when Nieble [74] reported the first poses the entire production control task into several sub-
computer-aided process planning (CAPP) system, numerous re- tasks, each of which is implemented by an intelligent
search efforts have been reported in this area. agent. By working collaboratively, the agents can reach
Generally, CAPP approaches can be classified into two cat- a solution for the problem.
egories: variant and generative. The success of the variant ap- 2) CoCAPP [133], [134] was proposed to distribute com-
proach depends on group technology and computerized database plex process-planning activities to multiple specialized
retrieval. When a new part enters a factory, a previous similar problem solvers and to coordinate them to solve com-
process plan is retrieved from the database and modified to suit plex problems. The CoCAPP attempts to satisfy five
the new part. This method is especially suitable for compa- major requirements: autonomy, flexibility, interoperabil-
nies with few, and relatively fixed, product families and a large ity, modularity, and scalability. It builds cooperation and
number of parts per family. Most of the earlier CAPP systems coordination mechanisms into distributed agents using
can be categorized under the variant approach [2]. The genera- knowledge-based techniques. Each agent in the system
tive approach, on the other hand, can be used to automatically deals with a relatively independent functional domain in
generate an optimal process plan according to the part’s fea- process planning.
tures and manufacturing requirements. Most of the generative 3) Zhang et al. [132] proposed an agent-based adap-
systems in the literature are knowledge-based systems utilizing tive process-planning (AAPP) system on top of an
SHEN et al.: AGENT-BASED DISTRIBUTED MANUFACTURING PROCESS PLANNING AND SCHEDULING 565
object-oriented manufacturing resources modeling However, the whole framework seems to have been de-
(OOMRM) framework. The OOMRM describes man- signed at a high level. No practical systems were reported.
ufacturing resources’ capability and capacity in an 7) Similarly, Lim and Zhang [55] introduced an APPSS sys-
object-oriented manner, while the AAPP is implemented tem, which is made up of a number of agents and functional
as an integrated process-planning platform. Instead of modules. This system is mainly used for the dynamic re-
automating process-planning tasks completely, the AAPP configuration and optimization of resource utilization in
system provides an interactive mode to enable experienced manufacturing shop floors by considering the real-time
manufacturing engineers to make decisions at crucial process-planning and scheduling issues.
points. Five agents are used in the AAPP to carry out 8) Kornienko et al. [50] considered process planning as a
part information classification, manufacturing resources typical constraint satisfaction problem to generate an op-
mapping, process planning, human planning, and ma- timized plan in a distributed way satisfying all restrictions
chining parameter retrieval. A contract net-based scheme in the presence of different disturbances. An agent plays
is utilized as the coordination protocol between agents. different “roles” and has a primary algorithm (determined
4) Sluga et al. [102] introduced a virtual work system (VWS) by interactive pattern) and a set of emergency states to
as the essential building block for in a distributed man- handle local emergencies or global emergencies. In case
ufacturing environment. The VWS represents a manu- an agent is in emergency state recognized by the activity
facturing work system in the information space, and is guard agent, it could either resolve the emergency by itself
structured as an autonomous agent. It is a constituent or request a rescue agent to handle it.
entity of an agent network in which dynamic clusters In addition to the above systems, there are also other simi-
of cooperating agents are solving manufacturing tasks. lar research efforts toward agent-based process planning [78],
The decision-making in process planning is based on [110]. All these systems tend to solve the process-planning prob-
a market mechanism consisting of bidding–negotiation– lem by cooperation and negotiation among intelligent agents.
contracting phases. The VWS approach aims at enabling The agents making up the systems usually use the function de-
dynamic decision-making based on the actual state of the composition approach as described in Section VI.
manufacturing environment.
5) CyberCut [103] is a research project that aims at devel-
oping a networked manufacturing service for rapid part IV. APPROACHES TO MANUFACTURING SCHEDULING
design and fabrication on the Internet. A critical part
of this service is an automated process-planning mod- A. Traditional Approaches
ule that is capable of generating process plans to sat- Because of its highly combinatorial aspects (NP-complete),
isfy the desired geometries and specified requirements. dynamic nature, and practical usefulness for industrial applica-
Three types of agents are designed to facilitate CyberCut: tions, the scheduling problem has been widely studied in the
primary process-planning agent, environmental planning literature by various methods: heuristics, constraint propagation
agent, and burr minimization tool path planning agent [22]. techniques, constraint satisfaction problem formalisms, Tabu
The multiagent planning module incorporates conven- search, simulated annealing, GAs, neural networks, fuzzy logic,
tional and specialized planning agents for environmental etc. [136].
consideration and burr minimization. However, the inter- As direct methods are not available for complex scheduling
actions between agents are based on human decisions. problems, search methods are usually adopted to solve these
6) IDCPPS [14] was reported to be an integrated, distributed, problems. However, the simplest generate-and-test search strat-
and cooperative process-planning system. The process- egy is not a reasonable approach for large complex problems.
planning tasks are broken into three levels, namely, initial Many local search algorithms are more appropriate. These al-
planning, decision-making, and detail planning. The initial gorithms require a cost function, a neighborhood function, and
planning deals with the manufacturability evaluation of a an efficient method for exploring the neighborhood.
design and the generation of alternative processing routes A variety of neighborhood search methods have been cre-
based on feature reasoning. The decision-making level ated including climbing, simulated annealing, etc. These meth-
takes place when the orders have been released for produc- ods offer heuristic refinements to the generate-and-test. Heuris-
tion on the shop floor. The result of this step is a ranked list tic approaches try to replace the exhaustive search strategies
of near-optimal alternative plans that considers the avail- with some sophisticated experience. With the aid of heuristics
ability of shop floor resources. The detail planning is exe- in searching strategies, good solutions (though possibly non-
cuted just before manufacturing begins. This step finishes optimal) to hard problems can be found within greatly reduced
the final selection of machines, tools, cutting parameters, computation time.
and the calculation of machining cost and time. Different The Petri Net approach and its variants, due to its graphical
functional modules are grouped into different agents, in- representation and mathematical analysis of the control logic
cluding the three process-planning agents dealing with the of a manufacturing system, provide a powerful approach to
above three-level planning, plus the task agents, resource model, control, and schedule an automated system, in both
agents, and coordination agents (CAD/Process coordina- its information flows and its material flows. Colored timed
tion agent and Process/Production coordination agent). object-oriented Petri Nets (CTOPN) [123] further incorporates
structured, reusable, and easily maintainable control/decision extensive bibliography on multiagent scheduling in manufac-
knowledge that can be used in scheduling/dispatching. turing systems is compiled by Schiegg [88].
Constraint satisfaction is another search procedure that oper- Agent-based approaches have several potential advantages
ates in the space of constraint sets rather than in the solution set for distributed manufacturing scheduling [95].
space [59], [60], [68]. a) These approaches use parallel computation through a large
The objective of multisite scheduling [86] is to support the number of processors, which may provide scheduling sys-
scheduling activities of a global scheduler or schedulers in dis- tems with high efficiency and robustness.
tributed production plants in a cooperative way. A schedule b) They can facilitate the integration of manufacturing pro-
generated on a global level must be translated into detailed cess planning and scheduling.
schedules as part of the local scheduling process. In the case of c) They make it possible for individual resources to trade off
a disturbance, feedback between the local and global levels is local performance to improve global performance, leading
essential. Global-level data are derived from aggregated local to cooperative scheduling.
data, and are normally imprecise or estimated. d) Resource agents may be connected directly to physical
Several approaches take advantages of search strategies in devices they represented for so as to realize real-time dy-
which even cost-deteriorating neighbors are accepted. Simu- namic rescheduling (of course, not immediate reschedul-
lated annealing uses an analogy with the physical process of ing after any change in the working environment for the
annealing, in which a pure lattice structure of a solid is made sake of system stability). It may therefore provide the man-
by heating up the solid in a heat bath until it melts, then cool- ufacturing system with higher reliability and device fault
ing it down slowly until it solidifies into a low-energy state. tolerance.
As designed, simulated annealing is a randomized neighbor- e) Schedules are achieved by using mechanisms similar to
hood search algorithm and it has been successfully applied to those being used in manufacturing supply chains (i.e.,
solve many single-objective scheduling problems. Tabu search negotiation rather than search). In this way, the manufac-
combines deterministic iterative improvements with the pos- turing capabilities of manufacturers can be directly con-
sibility of accepting cost-increasing solutions occasionally— nected to each other and optimization is possible at the
to direct the search away from local minimum [32]. In GAs, supply chain level, in addition to the shop floor level and
learning occurs through a solution selection process. GAs dis- the enterprise level.
cover superior solutions to global optimization problems adap- f) Other techniques may be adopted at certain levels for
tively (akin to the evolution of organisms in the natural world) decision-making, e.g., simulated annealing [48] and GAs
by searching for small, local improvements rather than big [33], [96].
jumps in a solution space. Fuzzy logic-based scheduling is used
to support the scheduling activities in a multisite scheduling
scenario [86]. In this system, a global scheduler or sched- C. Research Literature on Agent-Based
ulers in distributed production plants work in a cooperative Manufacturing Scheduling
way, based on adequate modeling and processing of imprecise Research in agent-based manufacturing scheduling has been
data. A robust prescription is created for the local scheduling more active and has a richer literature base than that in agent-
systems. based manufacturing process planning. This section provides a
All the traditional scheduling methods, whether analytical, detailed review in a structured way.
heuristic, or metaheuristic (including GAs, Tabu search, sim- 1) Earlier Attempts: Shaw may be the first person who pro-
ulated annealing, artificial neural networks, fuzzy logics), en- posed using agents in manufacturing scheduling and factory
counter great difficulties when they are applied to real-world control. He suggested that a manufacturing cell could subcon-
situations. This is because they use simplified theoretical mod- tract work to other cells through a bidding mechanism [89], [90].
els and are essentially centralized in the sense that all computa- Yet Another Manufacturing System (YAMS) [80] is another ex-
tions are carried out in a central computing unit. The intelligent ample of an early agent-based manufacturing system, wherein
agent technologies, on the other hand, suggest an innovative, each factory and factory component is represented as an agent.
lightweight approach to scheduling problems. This essentially Each individual agent has a collection of plans as well as knowl-
distributed approach is more flexible, efficient, and adaptable to edge about its own capabilities. The Contact Net protocol [104]
real-world dynamic manufacturing environments. is used for interagent negotiation.
2) Methodologies and Techniques: Different methodologies
and techniques have been proposed, developed, and used in the
B. Agent-Based Approaches
literature for agent-based manufacturing scheduling.
Within the past decade, a number of researchers have applied a) CORTES [84], [111] uses micro-opportunistic techniques
agent technology in attempts to resolve scheduling problems. for solving the scheduling problem through a two-agent
Applications include manufacturing flow shop scheduling [18], system, where each agent is responsible for scheduling a
[113] and job shop scheduling [49], [59], [60], transportation set of jobs and for monitoring a set of resources.
scheduling [27], power distribution scheduling [44], computing b) Baker [6] proposed a market-driven contract net for heter-
resource scheduling [31], meeting scheduling [100], medical archical agent-based scheduling. This agent architecture
test scheduling [38], and project management [54], [127]. An performs a type of forward/backward scheduling.
c) Logistics Management System (LMS) [28] applies inte- 3) Approaches and Architectures: To satisfy the require-
gration decision technologies to dispatch-scheduling in ments for next-generation manufacturing systems, researchers
semiconductor manufacturing. It uses functional agents, have proposed and developed a number of approaches and archi-
one for each production constraint, and a judge agent to tectures for agent-based manufacturing scheduling and control.
combine the votes of all the perspectives. Each agent par- a) Burke and Prosser [10] described a distributed asyn-
tially models those aspects of the environment that are chronous scheduling (DAS) system. The DAS architec-
needed to satisfy its objective. Its uniqueness is a voting ture consists of three types of entities: knowledge re-
protocol for communication among agents. sources, agents, and a constraint maintenance system. The
d) Liu and Sycara [59] proposed a coordination mecha- agents were originally developed as a multiagent heterar-
nism called Constraint Partition and Coordinated Reac- chy to represent only resources (O-agents). The final de-
tion (CP&CR) for job shop constraint satisfaction. This velopment includes agents for aggregations of resources
system assigns each resource to a resource agent respon- (T-agents) and an agent for overseeing the whole schedul-
sible for enforcing capacity constraints on the resource, ing process (S-agent). This final scheduling system orga-
and each job to a job agent responsible for enforcing nizes agents into a hierarchical architecture, in which the
temporal precedence and release-date constraints within S-agent assigns operations to the T-agents and the T-agents
each job. Moreover, a coordination mechanism called An- assign these operations further to O-agents, respectively.
chor&Ascend is proposed for distributed constraint op- While DAS is able to make a correct schedule, however,
timization. Anchor&Ascend employs an anchor agent to it has no method for optimizing that schedule.
conduct local optimization of its subsolution and inter- b) Scheduling in architecture for distributed dynamic manu-
acts with other agents that perform constraint satisfaction facturing scheduling (ADDYMS) is decomposed into two
through CP&CR to achieve global optimization [60]. levels [12]: the first level involves the assignment of a
e) In AARIA [79], the manufacturing capabilities (e.g., peo- manufacturing work cell to a task, and the second consists
ple, machines, and parts) are encapsulated as autonomous of the determination of a local resource as well as other
agents. Each agent seamlessly interoperates with other aspects, such as workers and tools, which may possibly be
agents in and outside the factory boundary. AARIA shared among a number of work cells. Corresponding to
used a mixture of heuristic scheduling techniques: for- these two levels, there are two kinds of agents: site agents
ward/backward scheduling, simulation scheduling, and and resource agents. The system is composed of several
intelligent scheduling. Scheduling is performed by job, connected site agents, each of which is in turn connected
resource, and operation. with its subsite agents and some local resource agents.
f) Miyashita [68] proposed an integrated architecture for dis- c) Lin and Solberg [58] showed how a market-like control
tributed planning and scheduling using the repair-based model could be used for adaptive resource allocation and
methodology together with the constraint-based mecha- distributed scheduling. They modeled the manufacturing
nism of dynamic coalition formation among agents. A shop floor exactly like a market place, where each task
prototype system called CAMPS is implemented, in which agent enters the market carrying certain “currency” and
a set of intelligent agents try to coordinate their actions for bargains with each resource agent on which it can be
satisfying planning/scheduling results by handling several proposed. At the same time, each resource agent com-
intra- and interagent constraints. petes with other agents to get a more “valuable” job. The
g) Usher [116] presented an experimental approach for per- market mechanism, using multiple-way and multiple-step
formance analysis of a multiagent system for job routing negotiation, is incorporated to coordinate different agents,
in job-shop settings: i) under various information levels for including part agents, resource agents, database agents,
constructing and evaluating bids, and ii) under actual real- and communication agents.
time process data for the negotiation process. Some simple d) Interrante and Rochowiak [43] proposed using active
but practical mechanisms are proposed and implemented. scheduling in the development of a multiagent architecture
h) Lu and Yih [61] proposed a framework that utilizes au- for dynamic manufacturing scheduling.
tonomous agents and weighted functions for distributed e) Murthy et al. [72] described an agent-based scheduling
decision-making in elevator manufacturing and assem- system based on the A-team architecture, in which func-
bly. This system dynamically adjusts the priorities of sub- tional agents generate, evaluate, improve, and prune a pool
assemblies in the queue buffer of a cell by considering the of candidate solutions. This system can be considered to
real-time status of all subassemblies in the same order. be a blackboard system.
i) In [4], an agent-based scheduling system, incorporating f) Kouiss et al. [49] proposed a multiagent architecture for
game theoretic based agent cooperation, is presented to dynamic job shop scheduling. Each agent represents a
solve the n-job three-stage flexible flow shop scheduling work center and performs a local dynamic scheduling by
problem. With scheduling task represented by a series of applying an adaptive dispatching rule. Depending on local
digraphs, MIP (mixed integer programming, minimizing and global considerations, a new selection of dispatching
makespan) is used by individual agents to schedule their rule is carried out when a predefined event occurs. The
jobs, and the final solution is reached by agent cooperation selection method is improved through the optimization of
using game theory. the thresholds used to detect symptoms (events). Agents
can also coordinate their actions to perform a global dy- was specially designed as an evaluation tool (testbed) for agent-
namic scheduling. However, a global agent is needed to based manufacturing systems (ABMS) based on its embedded
detect the symptom of the shop floor. ABMS model. It is a parallel simulator developed on Paragon su-
g) Sousa and Ramos [106] proposed a dynamic scheduling percomputer and the analytical results could clearly show the ef-
system architecture composed of the holons representing fects (marked by viability) of nonautonomous and autonomous
tasks together with the holons representing manufacturing factors on individual agents as well as the whole agent system.
resources. The Contract Net protocol is adapted to handle The shop floor agents’ (SFA) project at NCMS [73] was focused
temporal constraints and deal with conflicts. Sousa et al. on application of agent-based systems for shop floor schedul-
[107] further proposed an extended Contract Net Protocol ing and machine control. The project members include three
with constraints propagation for explicit representation of manufacturers (AMP, GM, and Rockwell Automation/Allen-
the precedence relationships between the operations of a Bradley) and a software company (Gensym). The objective of
task (with a cooperation phase between service providers). the project is the design and development of agent-based sys-
It shows some novelty compared with other variants of the tems to support three industrial scenarios, one sponsored by
Contract Net Protocol. each of the manufacturing members of the project [82]. Cowl-
h) van Brussel et al. [117] proposed the PROSA reference ing et al. [17] reported an application of intelligent agents to the
architecture for holonic manufacturing systems (HMS). It dynamic scheduling in steel production.
uses a hierarchical system architecture to organize holons Note that multiagent manufacturing systems are being consid-
(similar to agents and implemented as agents in most HMS ered to have substantial overlap with HMS [63]. The distributed
demonstrations as discussed below) and has been used as information-processing mechanisms in individual holons and
the basic architecture for implementing holonic manufac- holonic systems are generally resulted in the implementation of
turing scheduling systems [19], [67], [63]. The architec- multiagent systems [19] and the results of agent research com-
tural components include order holons, product holons, munity (including the standardization efforts like FIPA [26])
and resource holons. have been used to implement the concepts of HMS. This leads to
i) Wang et al. [124] proposed a heterachical multiagent sys- the trend of merging the two academic research communities—
tem and distributed ruler-based scheduling mechanism. MAS and HMS [63]. More discussions on this topic can be found
The scheduling system is decomposed into three subsys- in [9] and [11]. Within the MAS/HMS community, a number
tems, i.e., management subsystem, resource subsystem, of applications have been reported, including applications on
and part subsystem, and they are further decomposed into agent-based manufacturing process planning and scheduling.
several units, which are assigned to several agents, i.e.,
management agents (MA), resource agents (RA), and part
agents (PA). Rulers are encapsulated in agents and sched- V. APPROACHES TO PROCESS PLANNING AND
ules are able to be established and executed in a parallel SCHEDULING INTEGRATION
and distributed fashion. Traditionally, manufacturing process planning determines
j) An agent-based market-driven scheduling architecture how a product should be manufactured by focusing on the pure
was proposed by Yen and Wu [128] to integrate all kinds geometric and technological requirements of tasks. It assigns
of existing manufacturing scheduling systems over the machines, cutting tools, and cutting parameters to each process
Internet. Each standalone scheduling system is endowed based on somewhat unrealistic assumptions, such as unlimited
with agent feature by migrating the legacy system into resources and an idle shop floor. Without the consideration of
an Internet scheduling agent. A network of heterogeneous real-time machine workload and shop floor dynamics, the pro-
scheduling agents collaborate or compete with each other cess plans made offline during the planning stage are often out
for scheduling tasks using a market-driven protocol— of touch with shop floor operations at the time of task execution.
Vickrey Auction. Therefore, there is a need for the integration of manufacturing
k) Similarly, Goldberg et al. [34] uses a market-driven mech- process-planning and scheduling systems for generating more
anism for task and resource allocation in a larger dis- realistic and effective plans to be used in the shop floor.
tributed, multirobot control and coordination architecture. The topic of manufacturing process planning and scheduling
Individual robot interacts with others to advance its self- integration has been the research focus of many researchers for
interest; however, the aggregate effect of a robot commu- more than ten years. The importance of the nonlinear process
nity can manifest mutual beneficiary effects. plan concept (NLPP, also called flexible process plan, alternative
4) Testbeds and Industrial Applications: Only a few testbeds process plan, or multiple process plans) has been identified as
and real industrial applications have been developed and re- a milestone for process planning and scheduling integration by
ported in the literature. LMS [28] has been used in commercial almost all relevant publications.
production at IBM. The Manufacturing Scheduling and Control Since no recent comprehensive survey is available in the area
Testbed (MASCOT) [81] was a simulated testbed for manu- of manufacturing process planning and scheduling integration,
facturing scheduling and control. It provides a communication this section has been expanded to include all the approaches
infrastructure, a shared ontology, a shared interface based on that in this area. It provides a comprehensive state-of-the-art review
ontology, and a base set of realistic modules for the design, inte- for both agent-based approaches as well as other traditional
gration, and operation of agile enterprises. The TIE/Agent [40] approaches.
Note that some of the research projects or systems cited in this eral SA. The improved performance makes SA a candidate
section may have been mentioned in Sections II–IV. However, for solving the complex optimization of process planning
they are classified into different categories and reviewed from and scheduling.
different perspectives. g) Tan [112] provided a literature review and many cri-
tiques toward the existing integration methods. He stud-
ied and extensively experimented various mathematical
A. Traditional Approaches approaches to solve the problem, including linear mixed
1) Centralized Optimization Algorithms: Many researchers integer programming (LMIP), polynomial LMIP, Tabu
tried to combine process planning and scheduling as a single search, and branch-and-bound procedure. Tan introduced
scheduling optimization problem. The complexity created by a concept of cooperative solver that combines the strength
introducing alternative process plans and alternative resources of different algorithms to provide overall superior results
to the already NP-hard scheduling problem has been analyzed with the capability of considering multiple optimization
in Section II. With the much more complex situations in real- objectives.
world manufacturing process planning and dynamic shop floors, Aiming at reducing the computation complexity of a big cen-
the combined solution space becomes so large that an effective tral optimization algorithm, some researchers tried to split the
optimization algorithm becomes an absolute necessity. As a optimization problem into several steps, each with a distinguish-
result, a variety of optimization approaches have been studied ing objective.
and reported as follows. a) Brandimarte and Calderini [8] presented a hierarchical
a) The concurrent assignment algorithm proposed by Chen Tabu-search method on top of LMIP to deal with process
and Khoshnevis [16] uses a time-window scheme, an ef- selection and job shop scheduling. The problem is mod-
fective matrix, an operation matrix, and an optimal assign- eled with two objective functions, considering both the
ment algorithm to optimize the combination of multiple operation cost and the makespan.
parts with multiple features and multiple machines. b) Saygin and Kilic [87] proposed hierarchical integer lin-
b) Zijm [135] addressed the integration problem by searching ear programs with different objective functions to se-
alternatives for operations on the critical path based on lect alternative machines (shortest path) and alternative
graph theory, where the NLPPs are represented by graph process plans (dissimilarity maximization), and to finish
theoretical expressions. time scheduling of the obtained linear process plan. The
c) A branch-and-bound approach to optimize priority- three-step mathematical procedures make up an offline
weighted earliness of jobs scheduled in a mould manu- scheduling algorithm with good load balancing over po-
facturing shop floor was proposed by Gan and Lee [30]. tential resources. However, the rescheduling issue is only
Some simple heuristics are used to speed up the algorithm. based on alternative machines. Some simple heuristics are
This algorithm together with the rescheduling mechanism used but cannot solve all the situations in dynamic shop
was stated to be effective in terms of objective function floors.
and other performance measures. 2) Close Loop Optimization: Rather than listing alternative
d) Two algorithms were proposed and compared in [47]. The plans, some researchers argued that NLPP-oriented approaches
preprocessing algorithm combines the features of branch for integrating process planning and scheduling are still in
and bound and mixed integer programming techniques. an offline mode. All the acceptable schedules made at the
Another algorithm is an iterative one based on a heuristic predictive scheduling stage are almost immediately subject to
improvement method. The authors concluded that with the changes on the shop floor owing to the rescheduling factors,
expansion of problem size, the preprocessing algorithm such as machine breakdowns, materials shortage, order cancel-
displays good quality results and consumes moderate com- lation, due date changes, and so on [87]. Although many of the
putation time while the heuristic one is getting worse. systems adopting the above solutions were said to be responsive
e) Lee and Kim [52] introduced a simulation-based GA to shop floor disturbances, they cannot be called dynamic as
approach, where a simulation model is used to compute process planning is performed before scheduling, with the same
the performance measures (predict scheduling) and a optimization algorithm invoked to do the initial scheduling
GA is used to evaluate and select the best process plan and rescheduling. In the close loop process-planning (CLPP)
combination. The linear process plan obtained is then sent approaches, process plans are created based on the dynamic
to a scheduling module to do the allocation task. In the feedback from the production scheduling or production control
scheduling approach proposed by Sugimura et al. [109], modules.
the sequence generation of machining features is realized a) Mamalis et al. [62] used least cost and depth first search
by a GA and the selection of machining equipment is technique of the decision tree along with discrete event
completed by dynamic programming for each machining simulation to evaluate the suggested NLPPs. offline pro-
sequence. cess plans and schedules are developed in the first phase of
f) Palmer [77] applied simulated annealing (SA) to the inte- process planning and scheduling integration. In the second
gration problem and justified its optimization results. The phase, the system works out eventual changes of the ini-
cybernetic optimization by simulated annealing (COSA) tially defined production conditions (mainly the utilization
[39] attempts to mitigate the slow convergence rate of gen- of factory resources), and regenerates alternative process
plans. The authors called this phase an online integration B. Agent-Based Approaches
of process planning and scheduling.
For complex optimization tasks, agent approaches provide a
b) Within the ESPRIT project COMPLAIN, Kempenaers
distributed intelligent solution by multiagent negotiation, coor-
et al. [46] presented an architecture that integrates
dination, and cooperation.
the ideas of manual/automated process planning and
a) Gu et al. [35] proposed a multiagent system where the
scheduling, together with the idea of feedback loops. The
process routes and schedules of a part are accomplished
full closed loop solution includes production constraints
through the contract net bids. Process-planning tasks, in-
as the feedback from the scheduling module to the
cluding STEP file parsing and interpretation, tolerance
process-planning module to facilitate the generation or
analysis, operation planning, setup planning, and machine,
regeneration of NLPPs. The COMPLAN project is the
tool, and fixture selection, are all distributed to individual
continuation of FLEXPLAN [51], which uses a Petri-net
machine agents. There is no centralized process planner
to model and analyze the flexibility provided by NLPPs
in this system although same planning techniques are ap-
(represented by AND/OR graphs).
plied to every machine. The task allocation and process
3) Distributed Process-Planning (DPP) Approaches: The
alternative selection are achieved through the hierarchi-
basic idea of DPP is to look upon the integration problem as
cal bidding processes between machine agents and shop
a process-planning task. To integrate with scheduling, DPP ad-
floor manager, between upper level machine agents and
vocates a more flexible plan scenario by separating the whole
lower level machine agents, and between machine agents
process to several steps.
and tool agents, etc. A cost model is built to facilitate
a) Huang et al. [41] tried to separate the integration problem
decision-making in the process of negotiation.
of process planning and scheduling into three phases:
b) Denkena et al. [20] also presented a multiagent architec-
preplanning, pairing planning, and final planning. The
ture to determine operation routes and schedules. Their
preplanning is a technical manufacturing analysis of
approach separates the rough process-planning task as a
the product to identify machining requirements and
centralized shop floor planner from the detailed process
constraints. Pairing planning matches the required job
planning conducted through agent negotiations. The pro-
operations with the operation capabilities of the available
posed multiagent planning architecture (centralized rough
production resources. The final planning prepares de-
level process planning and decentralized planning and
tailed manufacturing plans of the product for the selected
scheduling) builds a flexible, reliable, and fault-tolerant
equipment. The interaction of process planning and
information logistics to enable supply chains, temporal
scheduling takes place in all three phases, starting from
logistic networks, or virtual enterprises.
offline machine potential information support to machine
c) IDCPPS [14] also uses agent-based techniques to imple-
group allocation and finally to specific machine selection
ment its process planning and scheduling integration.
within the machine group upon plan execution. From the
d) A cascading auction protocol [66] was proposed as a
viewpoint of process planning, the whole process plan
framework for integrating process planning and heter-
is not deducted at once but progressively. This work is
archical shop floor control. The integration of the real-
based on the results of the IPPM [130].
time online process planning (alleviating the selection of
b) IDCPPS [14] is an integrated, distributed, and cooperative
routing alternatives, resource alternatives, detail process
process-planning system. Similarly, the process-planning
planning, etc.), and shop floor control (resource alloca-
tasks are separated into three levels, namely, initial plan-
tion, scheduling, transportation, tooling, and fixturing) is
ning, decision-making, and detail planning. The responsi-
accomplished progressively through a recursive auction
bilities of each level are very similar to what has been pre-
process carried out in parallel among part management
sented by Huang et al. [41]. The results of these three steps
agent and multiple resource management agents. In this
are general process plans, a ranked list of near-optimal al-
approach, the complete part production process plan could
ternative plans and the final detailed linear process plans,
be constructed, including tooling, fixturing, and trans-
respectively. The integration with scheduling is considered
portation plans.
at each stage with process planning.
Because of the autonomous and cooperative nature of
c) Wang and Shen [120] presented a new DPP methodology
agents, scheduling and process-planning functions can be
by integrating machining feature-based planning, function
integrated from either a higher level of separate functional
block-based control, and agent-based distributed decision-
decomposed agents [14] or a lower level of resource agents
making. It proposes to use two-level decision-making—
plus a coordinator [35].
supervisory planning and operation planning. The super-
visory planning focuses on product data analysis, machine
selection, and machining sequence planning, and the op- VI. MAJOR DESIGN AND IMPLEMENTATION ISSUES
eration planning considers the detailed working steps of
the machining operations inside each process plan and is A. Agent Encapsulation
accomplished by intelligent NC controllers. The function Among different approaches used for agent encapsulation in
block-based integration architecture of DPP with manu- agent-based manufacturing systems, two approaches are dis-
facturing scheduling, execution, and control is discussed tinct: the functional decomposition approach and the physical
in more detail in [121]. decomposition approach.
In the functional decomposition approach, software agents are composition approach, rather than the functional decomposition
used to encapsulate modules detailed to functions, such as order approach.
acquisition, process planning, scheduling, material handling, This section does not discuss in detail the different kinds of
transportation management, and product distribution. There are agent architectures. A detailed discussion on desired charac-
no explicit relationships between software agents and physi- teristics and commonly used modules of agents in agent-based
cal entities. In the physical decomposition approach, software collaborative design and manufacturing systems can be found
agents are used to represent entities in the physical world, such in [94].
as workers, machines, tools, fixtures, products, parts, features, In agent-based manufacturing scheduling systems, the ap-
operations, etc. There exist explicit relationships between soft- plied agent negotiation protocols require individual agents to
ware agents and physical entities. reply to the incoming offers, to compete, and to negotiate or
The functional decomposition approach tends to share many to bargain with other agents. As a result, rich knowledge bases
state variables across different functions. Separate agents must and powerful learning and reasoning mechanisms are very im-
share many state variables, thus leading to problems of consis- portant. Each agent should have at least knowledge about the
tency and unintended interactions. The physical decomposition capability, availability, and cost of the physical resource (e.g., a
approach naturally defines distinct sets of state variables that can machine) represented by itself. Some sophisticated agents need
be managed efficiently by individual agents with limited inter- to have knowledge about other agents in the system, the prod-
actions, but it needs a large number of resource-related agents, ucts to be manufactured, and know-how (historical experience,
thus leading to other problems, such as communication over- successful cases), etc.
head and complex agent management. However, the functional The decision scheme of an individual agent depends primar-
decomposition approach is very useful to integrate existing sys- ily on two aspects: coordination or negotiation mechanisms
tems (e.g., CAD tools, MRP systems, databases, etc.) so as to used by the multiagent system and its local decision-making
resolve legacy problems. mechanisms based on knowledge. For example, a Contract Net
Corresponding to the two distinct approaches for agent en- protocol needs each individual agent to reply to an offer with
capsulation, two types of agent-based distributed manufacturing requested information such as cost, start time, processing time,
scheduling systems can be distinguished according to the fol- etc. [80]. A game-theory-based multiagent system needs agents
lowing characteristics. to follow game rules [4], [36]. While a multiagent system imple-
a) Scheduling is an incremental search process that can in- mented with a conversation scheme will need each agent to fol-
volve backtracking. Agents, responsible for scheduling or- low the conversation policies [57]. Local decision-making may
ders, perform local incremental searches for their orders use rule-based and case-based mechanisms reasoning on top of
and may consider multiple resources. The global schedule the knowledge the agent possesses. To update an agent’s knowl-
is obtained through the merging of local schedules. This edge, learning mechanisms are needed. Such learning mecha-
is very similar to centralized scheduling. nisms may range from case-based reasoning to neural network
b) An agent represents a single resource (e.g., a work cell, a and fuzzy logic-based reasoning.
machine, a tool, a fixture, a worker, etc.) and is responsible
for scheduling this resource. This agent may negotiate with
other agents to carry out the overall scheduling. C. System Structures
Examples of the second type of scheduling systems can be Agent system architectures provide the organizing frame-
found in [10], [28], [66], [72], [84], [101], [106], and [124]. works within which agents are designed and constructed. The
Most agent-based manufacturing scheduling systems proposed various system architectures proposed in the literature for
and developed in literature use the second approach. agent-based manufacturing process planning and scheduling
Things become complex when considering process-planning can be classified into three categories: hierarchical, federate,
tasks or their integrated systems. Most agent-based process- and autonomous.
planning systems (including scheduling integrated systems) fol- A typical modern manufacturing enterprise consists of a
low the functional decomposition of agents [22], [55], [99], number of physically distributed, semiautonomous units, each
[132], [134]. How to build multiagent systems to solve complex with a degree of control over local resources and with differ-
applications using the physical agent decomposition approach, ent information requirements. In this situation, many practical
or a combination of both approaches, is a possible future re- agent-based industrial applications still use the hierarchical ar-
search direction. chitecture, even though this has the disadvantages introduced
by centralized systems. In fact, agent-based distributed manu-
facturing process-planning and scheduling systems using func-
B. Agent Modeling
tional decomposition (each agent represents a function or a
In agent-based manufacturing process-planning and schedul- department in a traditional system) usually have this kind of
ing systems, bidding-based negotiations or market-like ap- hierarchical structure. Examples can be found in HMS [106],
proaches are commonly used. This section will address the de- ADDYMS [12], DAS [10], and LMS [28].
cision schemes that an individual agent should have to realize Because of the serious problems caused by the central
effective agent-based scheduling. Discussions in the following characteristics associated with hierarchical structure, the
paragraphs are also related to the systems using the physical de- federated architectures are increasingly being considered as a
compromise solution for industrial agent-based applications, planning and scheduling systems, which are complex, dynamic,
especially for large-scale engineering applications. Federated and composed of a large number of resource agents. A com-
multiagent architectures are able to coordinate multiagent bination of above-mentioned three approaches called “hybrid
activities via facilitation or mediation as a means of reducing approach” was proposed in MetaMorph II [91] for developing
communication overhead and ensuring system stability and more flexible, modular, scalable, and dynamic manufacturing
scalability. A fully federated agent-based system has no explicit systems.
shared facility for storing active data; rather, the system stores
all data in local databases and handles updates and changes
through message passing. The federated approach promises to D. Coordination and Negotiation
be a good foundation upon which to develop open, scalable When a system uses the functional decomposition approach, it
multiagent system architectures. is similar to traditional integrated systems. Such a system usually
Different federated architectures have been proposed in the needs a predefined coordination mechanism. Discussions in the
literature for agent-based manufacturing systems. Four ap- following paragraphs are therefore more related to the systems
proaches are dominant: facilitators, brokers, matchmakers, and using the physical decomposition approach.
mediators. Negotiation protocols are used in most agent-based manufac-
In the facilitator approach, several related agents are com- turing scheduling systems for resource allocation. The Contract
bined into a group. Communication between agents takes place Net Protocol or its modified versions are the most commonly
always through an interface called facilitator. Each facilitator is utilized, although some other protocols such as the voting mech-
responsible for providing a gateway between a local collection anism by Fordyce and Sullivan [28] have also been considered.
of agents and remote agents, usually by providing two main The Contract Net Protocol was first proposed by Smith [104]
services: 1) routing outgoing messages to the appropriate desti- and demonstrated on a distributed sensing system. To summa-
nations; 2) translating incoming messages for consumption by rize briefly, an agent (manager) having some work to subcontract
its local agents. This approach has been widely used in agent- broadcasts its offer and waits for other agents (contractors) to
based collaborative design systems [94]. send back their bids. After some delay, the best offers are re-
Brokers (also called broker agents) are similar to facilitators tained and contracts are allocated to one or more contractors
in some aspects, such as monitoring and notification. The dif- that process the subtask. The Contract Net Protocol is a coordi-
ference between a facilitator and a broker is that a facilitator nation method for task allocation, providing dynamic allocation
is responsible only for a designated group of agents, whereas a and natural load balancing. However, as the number of nodes
broker can be contacted by any agent dynamically in case they increases, the communication workload on the network soars,
need services to finish their tasks. The matchmaking mecha- which may lead to a situation where agents spend more time
nism is a superset of the brokering mechanism, since it uses in processing messages than doing actual work. Thus, various
the brokering mechanism to match agents. Yellow page agents improvements to the basic contract net approach have been pro-
and Directory Facilitators proposed by Foundation of Intelligent posed, such as the following:
Physical Agents (FIPA) [26] are similar to matchmakers. Direc- 1) sending offers to a limited number of nodes, instead of
tory Facilitator has been used in the iShopFloor project [93] for broadcasting them;
providing yellow page-like services to all other agents. 2) anticipating offers, i.e., contractors send bids in advance;
Mediator approach is another type of federated architecture. 3) varying the time when commitment is decided;
In addition to the functions of a facilitator and a broker, a me- 4) allowing de-commitment (breaking commitments);
diator assumes the role of system coordinator by promoting co- 5) allowing several agents to answer as a group (coalition
operation and learning among intelligent agents. Applications formation);
using mediators in intelligent manufacturing systems can be 6) introducing priorities for solving tasks.
found in [64], [65], [76], [91], and [93]. The CNP-based bidding mechanism used in manufacturing
In the third structure in the category (autonomous multiagent scheduling can be part-oriented [23], [58], resource-oriented
approach), each autonomous agent usually has the following [6], [12], or bidirectional [83].
characteristics: 1) it is not controlled or managed by any In basic CNP approaches, the choice of a contractor is done by
other software agents or human beings; 2) it can communi- comparing bids corresponding to a particular offer using what-
cate/interact directly with any other agents in the system and ever mechanisms that are relevant to the problem. In some more
also with other external systems; 3) it has knowledge about complex cases, e.g., involving de-commitment [85], penalties
other agents and its environment; and 4) it has its own goals are introduced thus bringing the approach closer to a market-like
and an associated set of motivations. Examples can be found approach [71]. Shen et al. [91] presented a promising approach
in [23], [58], [79], [83], and [128]. However, the autonomous by combining Contract-Net-based negotiation and mediator-
approach is only well suited for developing distributed systems based coordination for agent-based manufacturing scheduling
consisting of a small number of agents. and rescheduling.
In the mediator architecture, a static or dynamic hierarchy is Although Contact Net and its variants are usually used as
imposed for every specific task, which can provide computa- negotiation protocols in most agent-based scheduling systems,
tional simplicity and manageability. This type of architecture is market-based approaches are becoming more and more popu-
quite suitable for developing distributed manufacturing process- lar. Market-based or like protocols use the so-called bargaining
process or auction process, which is also simple and easy to use. tiation frameworks together with combined negotiation
Market-based or like approaches have recently been used in a techniques [97].
number of agent-based scheduling systems [6], [34], [54], [58], 3) Integration of process planning, manufacturing schedul-
[66]. ing, and control. Agent-based approaches provide a nat-
Some recent research work is also showing some advan- ural way to integrate manufacturing process planning,
tages by combining agent-based approaches with other tech- scheduling, and execution control. It also provides the
niques such as GAs, neural networks, fuzzy logic [86], and possibility of simultaneous optimization of process plan-
some mathematical modeling methods. Market-like or bidding- ning and manufacturing scheduling. However, it increases
based methods emphasize system flexibility and responsiveness significantly the complexity of the problem. Much more
over optimality of solutions, and are, therefore, more suitable effort needs to be devoted to this research topic, includ-
for dynamic rescheduling. Search methods, like GAs and sim- ing complexity analysis and formal modeling of such
ulated annealing, focus more on the optimality of solutions and integration.
are, therefore, more suitable for advance scheduling. A combi- 4) Integration with real-time information. Without consid-
nation of these methods can be used in intelligent shop floors ering real-time information including changing customer
requiring both advance scheduling and dynamic scheduling. For requirements and shop floor dynamics, the process plans
example, Daouas et al. [18] proposed combining agent-based and manufacturing schedules generated offline may be-
negotiation with simulated annealing search, while Shen [93] come invalid at the time of task execution. Real-time dy-
proposes to combine agent-based negotiation with GA-based namic scheduling has been an important research topic
search for scheduling optimization. [49], [106], but significant efforts are needed before solu-
Sycara et al. [106] proposed a different approach using “tex- tions can be widely used in industry.
ture measures,” where all agents share a common information 5) Integration of agent-based approaches with other ap-
base, called coordination agent. This approach requires each proaches. As mentioned early in the paper, market-like
agent to compute its own texture measure and submit it to the or bidding-based methods emphasize system flexibility
coordination agent, and then read the integrated texture mea- and responsiveness over optimality of solutions. They are
sure to make its own decision. After individual agents make more suitable for dynamic rescheduling. While search
their decisions, they submit their solutions to the coordination methods like GAs and simulated annealing focus more
agent that in turn regulates the possible conflicts. This approach on the optimality of solutions and are, therefore, more
could be used for agents to predict possible conflicts, but not to suitable for advance scheduling. It is interesting to inte-
eliminate conflicts. grate these two kinds of methods into powerful manu-
Some researchers have also realized the game-like nature of facturing process-planning and scheduling systems, e.g.,
independent scheduling decisions, and try to use game theory integration of agent-based negotiation with GA-based
to make their agents smarter [36]. search [96], with artificial neural-network-based learning
Very recently, a new adaptive negotiation approach has been or fuzzy logic-based learning, with simulated annealing-
proposed by our group to address complex negotiation situations based optimization [18], [79], with Petri-net/Color Petri-
[122]. net-based coordination [57], [123], with heuristic search
methods [79], and with matrix-based methods [1].
6) Combination of individual problem-solving and coordi-
VII. RESEARCH OPPORTUNITIES AND CHALLENGES nation/negotiation schemes. Obviously, there is a tradeoff
This paper provides a state-of-the-art survey of research on between individual problem solving at the individual agent
agent-based manufacturing process planning and scheduling. level and the coordination/negotiation scheme at the sys-
Based on this survey, as well as on our first-hand research and tem level. A big challenge is how to combine them using
development experience in this area, future research opportuni- the integration approaches mentioned above.
ties, as well as challenges, are identified as follows. 7) Integration with function blocks approach. The function
1) Theoretical investigation of methodologies needed for blocks specification [42], a powerful modeling approach
agent-based distributed manufacturing process planning in the industrial process control field, has been considered
and scheduling. Theoretical investigation of methodolo- as a critical media to integrate data, events, and decision-
gies, including implementation methodology, is needed to making processes of various manufacturing activities. In-
consolidate the current research results in this area and fa- tegration of agent-based approaches with function blocks
cilitate the implementation of real industrial applications. seems to be a promising solution for the integration of
2) Negotiation mechanisms, protocols, frameworks. As most manufacturing process planning, scheduling, and execu-
agent-based manufacturing process planning and schedul- tion control [121].
ing use bidding-based or market-like negotiation mecha- 8) Benchmark for agent-based distributed manufacturing
nisms/protocols, research and development of more pow- process-planning and scheduling problems. Benchmarks
erful negotiation mechanisms and protocols are needed. are needed to evaluate various systems proposed and de-
Combinatorial market-based negotiation protocols are of veloped using different approaches [13], as well as to
much interest in the near future. Advanced negotiation compare with systems implemented using traditional tech-
frameworks are to be investigated, e.g., adaptive nego- niques/approaches.
9) Assurance of system security. System security is not the in Manufacturing, A. Famili, D. S. Nau, and S. H. Kim, Eds. Menlo Park,
focus of this paper, but a major concern of implement- CA: AAAI Press, 1992, pp. 199–214.
[13] S. Cavalieri, L. Bongaerts, M. Macchi, M. Taisch, and J. Wyns, “A bench-
ing Internet-enabled manufacturing systems, which is the mark framework for manufacturing control,” in Proc. 2nd Int. Workshop
assurance that proprietary information owned by the or- Intell. Manuf. Syst., Leuven, Belgium, 1999, pp. 225–236.
ganization or information about company operations is [14] F. T. S. Chan, J. Zhang, and P. Li, “Modelling of integrated, distributed and
cooperative process planning system using an agent-based approach,” in
available only to authorized individuals or organizations. Proc. Inst. Mech. Eng., Part B—J. Eng. Manuf., vol. 215, no. B10, 2001,
This problem must be solved before agent-based manu- pp. 1437–1451.
facturing process-planning and scheduling systems can be [15] P. T. Chang and C. H. Chang, “An integrated artificial intelligent computer-
aided process planning system,” Int. J. Comput. Integr. Manuf., vol. 13,
implemented in industrial settings [126]. no. 6, pp. 483–497, 2000.
[16] Q. Chen and B. Khoshnevis, “Scheduling with flexible process plans,”
Prod. Planning Control, vol. 4, no. 4, pp. 333–343, 1993.
[17] P. I. Cowling, D. Ouelhadj, and S. Petrovic, “A multi-agent architecture
VIII. CONCLUDING REMARKS for dynamic scheduling of steel hot rolling,” J. Intell. Manuf., vol. 14,
It is becoming clear that agent-based approaches offer many no. 5, pp. 457–470, 2003.
[18] T. Daouas, K. Ghedira, and J. P. Muller, “How to schedule a flow shop
advantages for distributed manufacturing process-planning and plant by agents,” in Applications of Artificial Intelligence in Engineering.
scheduling systems: modularity, reconfigurability, scalability, Billerica, MA: Computational Mechanics Inc., 1995, pp. 73–80.
upgradeability, and robustness (including fault recovery). The [19] S. M. Deen, Agent-Based Manufacturing—Advances in the Holonic Ap-
proach. Berlin, Germany: Springer-Verlag, 2003.
results achieved so far in the agent research community provide [20] B. Denkena, H. K. Tonshoff, M. Zwick, and P. O. Woelk, “Process planning
excellent motivation for further development of solutions in this and scheduling with multiagent systems,” in Knowledge and Technology
area. Moreover, at present, there are no other ways to solve these Integration in Production and Services—Balancing Knowledge and Tech-
nology in Product and Service Life Cycle. Norwell, MA: Kluwer, 2002,
complex problems. However, whether the potential advantages pp. 339–348.
of agent-based approaches can actually be realized in industrial [21] C. R. Devireddy and K. Ghosh, “Feature-based modeling and neural
systems will depend on the selection of a suitable system ar- network-based CAPP for integrated manufacturing,” Int. J. Comput. In-
tegr. Manuf., vol. 12, no. 1, pp. 61–74, 1999.
chitecture for agent organization and an appropriate approach [22] D. Dornfeld, P. K. Wright, F. C. Wang, P. Sheng, J. Stori, V. Sundararajan,
for agent encapsulation; on the design and implementation of N. Krishnan, and C. H. Chu, “Multi-agent process planning for a net-
effective mechanisms and protocols for communication, coop- worked machining service,” in Tech. Paper NAMRC XXVII, vol. MS99–
175, 1999, pp. 1–6.
eration, coordination, and negotiation; and on the design and [23] N. A. Duffie and R. S. Piper, “Non-hierarchical control of manufacturing
implementation of advanced internal architectures and efficient systems,” J. Manuf. Syst., vol. 5, no. 2, pp. 137–139, 1986.
decision schemes of individual agents. [24] K. O. Edalew, H. S. Abdalla, and R. J. Nash, “A computer-based intelligent
system for automatic tool selection,” Mater. Design, vol. 22, pp. 337–351,
2001.
[25] H. A. Elmaraghy, “Evolution and future perspectives on CAPP,” CIRP
REFERENCES Ann. (Keynote), vol. 42, no. 2, pp. 739–751, 1993.
[26] FIPA (Foundation of Intelligent Physical Agents). [Online]. Available:
[1] M. Akbulut and S. V. Kamarthi, “Agent-based scheduling system to http://www.fipa.org/
achieve agility,” presented at the SPIE Int. Conf. Network Intelligence: [27] K. Fischer, J. P. Muller, M. Pischel, and D. Schier, “A model for cooper-
Internet Based Manufacturing, Boston, MA, 2000. ative transportation scheduling,” in Proc. ICMAS’95, San Francisco, CA,
[2] L. Alting and H. C. Zhang, “Computer aided process planning: The 1995, pp. 109–116.
state-of-the-art survey,” Int. J. Prod. Res., vol. 27, no. 4, pp. 553–605, [28] K. Fordyce and G. G. Sullivan, “Logistics management system (LMS):
1989. Integrating decision technologies for dispatch scheduling in semiconduc-
[3] B. Arezoo, K. Ridgway, and A. M. A. Al-Ahmari, “Selection of cutting tor manufacturing,” in Intelligent Scheduling, M. Zweben and M. S. Fox,
tools and conditions of machining operations using an expert system,” Eds. San Francisco, CA: Morgan Kaufmann, 1994, pp. 473–516.
Comput. Ind., vol. 42, pp. 43–60, 2000. [29] S. French, Sequencing and Scheduling: An Introduction to the Mathemat-
[4] A. Babayan and D. He, “A distributed scheduling methodology for ics of the Job-Shop. New York: Wiley, 1982.
a two-machine flowshop using cooperative interaction via multiple [30] P. Y. Gan and K. S. Lee, “Scheduling of flexible-sequenced process plans
coupling agents,” Int. J. Prod. Res., vol. 42, no. 4, pp. 777–7104, in a mould manufacturing shop,” Int. J. Adv. Manuf. Technol., vol. 20,
2004. pp. 214–222, 2002.
[5] T. P. Bagchi, Multiobjective Scheduling by Genetic Algorithms. Norwell, [31] A. Glockner and J. Pasquale, “Coadaptive behavior in a simple distributed
MA: Kluwer, 1999. job scheduling,” IEEE Trans. Syst., Man, Cybern., vol. 23, no. 3, pp. 902–
[6] A. D. Baker “Manufacturing control with a market-driven contract net,” 907, May–Jun. 1993.
Ph.D. thesis, Rensselaer Polytechnic Inst., Troy, NY, 1991. [32] F. W. Glover and M. Laguna, Tabu Search. Norwell, MA: Kluwer, 1997.
[7] R. M. Boogert, H. J. Kals, and F. J. Van Houten, “Tool paths and cut- [33] D. Goldberg, Genetic Algorithms in Search, Optimization and Machine
ting technology in computer-aided process planning,” Int. J. Adv. Manuf. Learning. Reading, MA: Addison-Wesley, 1989.
Technol., vol. 11, pp. 186–197, 1996. [34] D. Goldberg, V. Cicirello, and M. B. Dias, “Task allocation using a
[8] P. Brandimarte and M. Calderini, “A hierarchical bicriterion approach to market-based planning mechanism,” in Proc. Int. Conf. AAMAS’03, vol. 2,
integrated process plan selection and job shop scheduling,” Int. J. Prod. Melbourne, Australia, 2003, pp. 996–997.
Res., vol. 33, no. 1, pp. 161–181, 1995. [35] P. Gu, S. Balasubramanian, and D. H. Norrie, “Bidding-based process
[9] R. W. Brennan and D. H. Norrie, “Agents, holons and function blocks: planning and scheduling in a multi-agent system,” Comput. Ind. Eng.,
Distributed intelligent control in manufacturing,” J. Appl. Syst. Stud., vol. 32, no. 2, pp. 477–496, 1997.
vol. 2, no. 1, pp. 1–19, 2001. [36] Z. Guan, M. Lei, B. Wu, Y. Wu, and S. Yang, “Application of decentralized
[10] P. Burke and P. Prosser, “The distributed asynchronous scheduler,” in cooperative problem solving in dynamic flexible scheduling,” in Proc.
Intelligent Scheduling, M. Zweben and M. S. Fox, Eds. San Francisco, SPIE, Bellingham, WA, 1995, pp. 179–183.
CA: Morgan Kaufmann, 1994, pp. 309–339. [37] K. Hashmi, M. A. El Baradie, and M. Ryan, “Fuzzy logic based intelligent
[11] S. Bussmann, “An agent-oriented architecture for holonic manufactur- selection of machining parameters,” Comput. Ind. Eng., vol. 35, no. 3–4,
ing control,” in Proc. 1st Int. Workshop Intell. Manuf. Syst., Lausanne, pp. 571–574, 1998.
Switzerland, 1998, pp. 1–12. [38] L. T. Herren, B. L. Robey, and P. K. Fink, “Multi-agent knowledge-based
[12] J. Butler and H. Ohtsubo, “ADDYMS: Architecture for distributed dy- system for planning and scheduling medical tests,” in Proc. IEEE Symp.
namic manufacturing scheduling,” in Artificial Intelligence Applications Comput-Based Med. Syst., Jun. 13–16, 1993, pp. 159–164.
[39] J. W. Herrmann, M. Eleischer, E. Lin, V. Mathur, and J. Glasser, [64] F. Maturana, W. Shen, and D. H. Norrie, “MetaMorph: An adaptive agent-
“Affordable space systems manufacturing: Intelligent synthesis tech- based architecture for intelligent manufacturing,” Int. J. Prod. Res., vol. 37,
nology, process planning, and production scheduling,” presented at no. 10, pp. 2159–2174, 1999.
the ASME DETC/CIE 2001, Pittsburgh, US, 2001, DETC2001/ [65] F. Maturana, S. Balasubramanian, and D. H. Norrie, “A multi-agent ap-
CIE-21281. proach to integrated planning and scheduling for concurrent engineering,”
[40] C. Y. Huang and S. Y. Nof, “Evaluation of agent-based manufacturing sys- in Proc. Int. Conf. Concurrent Engineering: Research and Applications,
tems based on a parallel simulator,” Comput. Ind. Eng., vol. 43, pp. 529– Toronto, ON, Canada, 1996, pp. 272–279.
552, 2002. [66] P. McDonnell, G. Smith, S. Joshi, and S. R. T. Kumara, “A cascading
[41] S. H. Huang, H. C. Zhang, and M. L. Smith, “A progressive approach for auction protocol as a framework for integrating process planning and
the integration of process planning and scheduling,” IEE Trans., vol. 27, heterarchical shop floor control,” Int. J. Flexible Manuf. Syst., vol. 11,
no. 4, pp. 456–464, 1995. no. 1, pp. 37–62, 1999.
[42] International Electrotechnical Commission (IEC), IEC 61499: function [67] D. C. McFarlane and S. Bussmann, “Developments in holonic production
blocks for industrial-process measurement and control systems (part-1: planning and control,” Prod. Planning Control, vol. 11, no. 6, pp. 522–536,
architecture), IEC TC65/WG6 Draft, 2000. 2000.
[43] L. D. Interrante and D. M. Rochowiak, “Active rescheduling and collabo- [68] K. Miyashita, “CAMPS: A constraint-based architecture for multi-agent
ration in dynamic manufacturing systems,” Concurrent Eng.: Res. Appl., planning and scheduling,” J. Intell. Manuf., vol. 9, no. 2, pp. 147–154,
vol. 2, no. 2, pp. 97–105, 1994. 1998.
[44] N. R. Jennings, J. M. Corera, and I. Laresgoiti, “Developing industrial [69] L. Monostori, Z. J. Viharos, and S. Markos, “Satisfying various require-
multi-agent systems,” in Proc. ICMAS’95, 1995, pp. 423–430. ments in different levels and stages of machining using one general ANN-
[45] J. Y. Jung and R. S. Ahluwalia, “Prismatic part feature extraction and based process model,” J. Mater. Process. Technol., vol. 107, pp. 228–235,
feature-based tool path selection,” J. Des. Manuf., vol. 3, no. 1, pp. 1–19, 2000.
1993. [70] N. Morad and A. Zalzala, “Genetic algorithms in integrated pro-
[46] J. Kempenaers, J. Pinte, J. Detand, and J. P. Kruth, “A collaborative process cess planning and scheduling,” J. Intell. Manuf., vol. 6, pp. 169–179,
planning and scheduling system,” Adv. Eng. Softw., vol. 25, no. 1, pp. 3–8, 1999.
1996. [71] T. Mullen and M. P. Wellman, “Some issues in the design of market-
[47] K. H. Kim and P. J. Egbelu, “Scheduling in a production environment oriented agents,” in Intelligent Agents II: Agent Theories, Architectures,
with multiple process plans per job,” Int. J. Prod. Res., vol. 37, no. 12, and Languages, M. Wooldridge, J. P. Müller, and M. Tambe, Eds. Lecture
pp. 2725–2753, 1999. Notes in Artificial Intelligence, vol. 1037, New York: Springer-Verlag,
[48] S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by simulated 1996, pp. 283–298.
annealing,” Science, vol. 220, pp. 671–680, 1983. [72] S. Murthy, R. Akkiraju, J. Rachlin, and F. Wu, “Agent-based coopera-
[49] K. Kouiss, H. Pierreval, and N. Mebarki, “Using multi-agent architecture tive scheduling,” in Proc. AAAI Workshop Constrains and Agents, 1997,
in FMS for dynamic scheduling,” J. Intell. Manuf., vol. 8, no. 1, pp. 41–47, pp. 112–117.
1997. [73] NCMS, “Shop floor agents,” Nat. Center for Manuf. Sci. Ann Arbor, MI,
[50] S. Kornienko, O. Kornienko, and P. Levi, “Flexible manufacturing process Tech. Rep., 1998.
planning based on the multi-agent technology,” in Proc. 21st IASTED Int. [74] B. W. Nieble, “Mechanized process selection for planning new designs,”
Conf. Applied Informatics, Innsbruck, Austria, Feb. 10–13, 2003, pp. 156– presented at the ASTME, 1965, Paper 737.
161. [75] S. K. Ong and A. Y. C. Nee, “Fuzzy-set-based approach for concurrent
[51] J. P. Kruth and J. Detand, “A CAPP system for nonlinear process plans,” constraint setup planning,” J. Intell. Manuf., vol. 7, no. 2, pp. 107–120,
Ann. CIRP, vol. 41, no. 1, pp. 489–492, 1992. 1996.
[52] H. Lee and S. S. Kim, “Integration of process planning and scheduling [76] D. Ouelhadj, C. Hanachi, and B. Bouzouia, “Multi-agent system for dy-
using simulation based genetic algorithms,” Int. J. Adv. Manuf. Technol., namic scheduling and control in manufacturing cells,” in Working Notes
vol. 18, no. 8, pp. 606–590, 2001. of the Agent-Based Manufacturing Workshop, Minneapolis, MN, 1998,
[53] K. Y. Lee and M. Y. Jung, “Flexible process sequencing using Petri net pp. 96–105.
theory,” Comput. Ind. Eng., vol. 28, no. 2, pp. 279–290, 1995. [77] G. J. Palmer, “A simulated annealing approach to integrated pro-
[54] Y. Lee, S. R. Kumara, and K. Chatterjee, “Multiagent based dynamic re- duction scheduling,” J. Intell. Manuf., vol. 7, no. 3, pp. 163–176,
source scheduling for distributed multiple projects using a market mech- 1996.
anism,” J. Intell. Manuf., vol. 14, no. 5, pp. 471–484, 2003. [78] H. G. Park and J. M. Baik, “Enhancing manufacturing product develop-
[55] M. Lim and Z. Zhang, “APPSS—An agent-based dynamic process plan- ment through learning agent system over internet,” Comput. Ind. Eng.,
ning and scheduling system,” in Proc. Volume from the IFAC Workshop, vol. 37, pp. 117–120, 1999.
2000, pp. 51–56. [79] V. D. Parunak, A. D. Baker, and S. J. Clark, “The AARIA agent archi-
[56] T. Lim, J. Corney, J. M. Ritchie, and D. E. R. Clark, “Optimizing tool tecture: An example of requirements-driven agent-based system design,”
selection,” Int. J. Prod. Res., vol. 39, no. 6, pp. 1239–1256, 2001. presented at the 1st Int. Conf. Autonomous Agents, Marina del Rey, CA,
[57] F. Lin, D. H. Norrie, W. Shen, and R. Kremer, “Conversation 1997.
specification—A schema-based approach to specifying conversation poli- [80] V. D. Parunak, “Manufacturing experience with the contract net,” in Dis-
cies,” in Issues in Agent Communication, F. Dignum and M. Greaves, tributed Artificial Intelligence, M. N. Huhns, Ed. New York: Pitman, 1987,
Eds. Lecture Notes in Computer Science, vol. 1916, New York: Springer- pp. 285–310.
Verlag, 2000, pp. 193–204. [81] ——, “MASCOT: A virtual factory for research and development in manu-
[58] G. Y. J. Lin and J. J. Solberg, “Integrated shop floor control using au- facturing scheduling and control,” Ind. Technol. Inst., Tech. Memo. 93-02,
tonomous agents,” IIE Trans.: Des. Manuf., vol. 24, no. 3, pp. 57–71, 1993.
1992. [82] ——, “Workshop report: Implementing manufacturing agents,” Nat. Cen-
[59] J. Liu and K. P. Sycara, “Distributed problem solving through coordination ter for Manuf. Sci., Ann Arbor, MI, Tech. Rep., 1996.
in a society of agents,” presented at the 13th Int. Workshop on DAI, 1994. [83] A. Saad, G. Biswas, K. Kawamura, M. E. Johnson, and A. Salama, “Eval-
[60] , “Exploiting problem structure for distributed constraint uation of contract net-based heterarchical scheduling for flexible man-
optimization,” in Proc. 1st Int. Conf. Multi-Agent Systems, ufacturing systems,” in Proc. 1995 IJCAI’95, Workshop Intell. Manuf.,
1995, pp. 246–253. Montreal, QC, Canada, 1995, pp. 310–321.
[61] T. P. Lu and Y. Yih, “An agent-based production control framework for [84] N. Sadeh and M. S. Fox, “CORTES: An exploration into
multiple-line collaborative manufacturing,” Int. J. Prod. Res., vol. 39, micro-opportunistic job-shop scheduling,” presented at the Work-
no. 10, pp. 2155–2176, 2001. shop on Manufacturing Production Scheduling (IJCAI-89), Detroit,
[62] A. G. Mamalis, I. Malagardis, and K. Kanbouris, “online integration of 1989.
process planning module with production scheduling,” Int. J. Adv. Manuf. [85] T. Sandholm and V. Lesser, “Issues in automated negotiation and electronic
Technol., vol. 12, no. 5, pp. 330–338, 1996. commerce: Extending the contract net framework,” in Proc. 1st Int. Conf.
[63] V. Mařik, D. McFarlane, and P. Valckenaers, Holonic and Multi-Agent Multi-Agent Systems, 1995, pp. 328–335.
Systems for Manufacturing (1st Int. Conf. on Industrial Applications of [86] J. Sauser, G. Suelmann, and H. J. Appelrath, “Multi-site scheduling
Holonic and Multi-Agent Systems). Berlin, Germany: Springer-Verlag, with fuzzy concepts,” Int. J. Approx. Reason., vol. 19, pp. 145–160,
2003. 1998.
[87] C. Saygin and S. E. Kilic, “Integrating flexible process plans with schedul- [111] K. P. Sycara, S. F. Oth, N. Sadeh, and M. S. Fox, “Resource allocation in
ing in flexible manufacturing systems,” Int. J. Adv. Manuf. Technol., distributed factory scheduling,” in Intelligent Scheduling, M. Zweben and
vol. 15, no. 4, pp. 268–280, 1999. M. S. Fox, Eds. San Francisco, CA: Morgan Kaufmann, 1991, pp. 29–40.
[88] P. Schiegg, Bibliography on Multi-Agent Scheduling in Manufacturing [112] W. Tan, “Integration of process planning and scheduling—A review,” J.
Systems. [Online]. Available: http://farm.ecs.umass.edu/˜pschiegg/bib/ Intell. Manuf., vol. 11, pp. 51–63, 2000.
lit.html [113] D. Trentesaux, C. Tahon, and P. Ladet, “Hybrid production control ap-
[89] M. J. Shaw, “Dynamic scheduling in cellular manufacturing systems: A proach for JIT scheduling,” Artif. Intell. Eng., vol. 12, no. 1, pp. 49–67,
framework for networked decision making,” J. Manuf. Syst., vol. 7, no. 2, 1998.
pp. 83–94, 1988. [114] Y. J. Tseng and S. B. Joshi, “Recognizing multiple interpretations of inter-
[90] M. J. Shaw and A. B. Whinston, “Distributed planning in cellular flexi- acting machining features,” Comput.-Aided Des., vol. 26, no. 9, pp. 667–
ble manufacturing systems,” Manage. Inform. Res. Center, Purdue Univ., 688, 1994.
West Lafayette, IN, Tech. Rep., 1983. [115] Y. Tu, X. Chu, and W. Yang, “Computer-aided process planning in virtual
[91] W. Shen, D. Xue, and D. H. Norrie, “An agent-based manufacturing en- one-of-a-kind production,” Comput. Ind., vol. 41, pp. 99–110, 2000.
terprise infrastructure for distributed integrated intelligent manufacturing [116] J. M. Usher, “Negotiation-based routing in job shops via collaborative
systems,” in Proc. 3rd Int. Conf. Practical Application of Intelligent Agents agents,” J. Intell. Manuf., vol. 14, no. 5, pp. 485–499, 2003.
and Multi-Agents (PAAM98), London, U.K., 1998, pp. 533–548. [117] H. V. Brussel, J. Wyns, P. Valckenaers, L. Bongaerts, and P. Peeters,
[92] W. Shen and D. H. Norrie, “Agent-based systems for intelligent man- “Reference architecture for holonic manufacturing systems: PROSA,”
ufacturing: A state-of-the-art survey,” Knowl. Inf. Syst., vol. 1, no. 2, Comput. Ind., vol. 37, no. 3, pp. 255–274, 1998.
pp. 129–156, 1999. [118] G. C. Vosniakos and B. J. Davies, “Knowledge-based selection and
[93] W. Shen, S. Lang, L. Korba, L. Wang, and B. Wong, “Reference archi- sequencing of hole-making operations for prismatic parts,” Int. J. Adv.
tecture for internet-based intelligent shop floors,” in Proc. SPIE Int. Conf. Manuf. Technol., vol. 8, pp. 9–16, 1993.
Network Intelligence: Internet Based Manufacturing, vol. 4208, Boston, [119] L. Wang and D. H. Norrie, “Process planning and control in a holonic
MA, 2000, pp. 63–71. manufacturing environment,” J. Appl. Syst. Stud., vol. 2, no. 1, pp. 106–
[94] W. Shen, D. H. Norrie, and J. P. Barthès, Multi-Agent Systems for Con- 126, 2001.
current Intelligent Design and Manufacturing. London, U.K.: Taylor & [120] L. Wang and W. Shen, “DPP: An agent-based approach for distributed
Francis, 2001. process planning,” J. Intell. Manuf., vol. 14, no. 5, pp. 429–440, 2003.
[95] W. Shen, “Distributed manufacturing scheduling using intelligent agents,” [121] L. Wang, Q. Hao, and W. Shen, “Function block based integration
IEEE Expert/Intell. Syst., vol. 17, no. 1, pp. 88–94, Jan./Feb. 2002. of process planning, scheduling and execution for RMS,” presented at
[96] , “Genetic algorithms in agent-based manufacturing scheduling sys- the CIRP 2nd Int. Conf. Reconfigurable Manufacturing, Ann Arbor, MI,
tems,” Integr. Comput.-Aided Eng., vol. 9, no. 3, pp. 207–218, 2002. 2003.
[97] W. Shen, Y. Li, H. Ghenniwa, and C. Wang, “Adaptive negotiation for [122] C. Wang, W. Shen, and H. Ghenniwa, “An adaptive negotiation frame-
agent-based grid computing,” in Proc. AAMAS 2002 Workshop on Agent- work for agent based dynamic manufacturing scheduling,” in Proc. IEEE
cities: Challenges in Open Agent Environments, Bologna, Italy, 2002, SMC, vol. 2, Washington, DC, Oct. 5–8, 2003, pp. 1211–1216.
pp. 32–36. [123] L. C. Wang and S. Y. Wu, “Modelling with colored timed object-oriented
[98] W. Shen and H. Ghenniwa, “Editorial of the special issue on agent-based petri nets for automated manufacturing systems,” Comput. Ind. Eng.,
manufacturing process planning and scheduling,” J. Intell. Manuf., vol. 14, vol. 34, no. 2, pp. 463–480, 1998.
no. 5, pp. 427–428, 2003. [124] Y. H. Wang, C. W. Yin, and Y. Zhang, “A multi-agent and distributed
[99] W. Shih and K. Srihari, “Distributed artificial intelligence in manufactur- ruler based approach to production scheduling of agile manufacturing
ing systems control,” Comput. Ind. Eng., vol. 29, no. 1–4, pp. 199–203, systems,” Int. J. Comput. Integ. Manuf., vol. 16, no. 2, pp. 81–92, 2003.
1995. [125] R. R. Wu and H. M. Zhang, “Object-oriented and fuzzy-set-based ap-
[100] T. Shintani and T. Ito, “An architecture for multi-agent negotiation using proach for setup planning,” Int. J. Adv. Manuf. Technol., vol. 14, pp. 406–
private preferences in a meeting scheduler,” in Proc. 5th PRICAI’98, 1998, 411, 1998.
pp. 47–60. [126] Y. Xu, L. Korba, L. Wang, Q. Hao, W. Shen, and S. Lang, “A security
[101] R. Sikora and M. J. Shaw, “Coordination mechanisms for multi- framework for collaborative distributed industrial control,” presented at the
agent manufacturing systems: Applications to integrated manufacturing IEEE Int. Conf. Industrial Informatics (CD-ROM), Banff, AB, Canada,
scheduling,” IEEE Trans. Eng. Manage., vol. 44, no. 2, pp. 175–187, May 2003.
1997. [127] Y. Yan, T. Kuphal, and J. Bode, “Application of multi-agent systems in
[102] A. Sluga, P. Butala, and G. Bervar, “A multi-agent approach to process project management,” in Working Notes of the Agent-Based Manufactur-
planning and fabrication in distributed manufacturing,” Comput. Ind. Eng., ing Workshop, Minneapolis, MN, 1998, pp. 160–170.
vol. 35, no. 3–4, pp. 455–460, 1998. [128] B. P.-C. Yen and O. Q. Wu, “Internet scheduling environment with market
[103] C. S. Smith and P. K. Right, “CyberCut: A world wide web based driven agents,” IEEE Trans. Syst., Man, Cybern. A, Syst., Humans, vol. 34,
design-to-fabrication tool,” J. Manuf. Syst., vol. 15, no. 6, pp. 432–442, no. 2, pp. 281–289, Mar. 2004.
1996. [129] S. H. Yeo, B. K. A. Ngoi, and H. Chen, “Process sequence optimization
[104] R. G. Smith, “The contract net protocol: High-level communication and based on a new cost-tolerance model,” J. Intell. Manuf., vol. 9, pp. 29–37,
control in a distributed problem solver,” IEEE Trans. Comput., vol. C-29, 1998.
no. 12, pp. 1104–1113, 1980. [130] H. C. Zhang and E. Merchant, “IPPM—A prototype to integrate process
[105] D. N. Sormaz and B. Khoshnevis, “Process planning knowledge rep- planning and job shop scheduling functions,” CIRP Ann., vol. 42, no. 1,
resentation using an object-oriented data model,” Int. J. Comput. Integr. pp. 513–518, 1993.
Manuf., vol. 10, no. 1–4, pp. 92–104, 1997. [131] F. Zhang, Y. F. Zhang, and A. Y. C. Nee, “Using genetic algorithms in
[106] P. Sousa and C. Ramos, “A dynamic scheduling holon for manufacturing process planning for job shop machining,” IEEE Trans. Evol. Comput.,
orders,” J. Intell. Manuf., vol. 9, no. 2, pp. 107–112, 1997. vol. 1, no. 4, pp. 278–289, Nov. 1997.
[107] P. Sousa, C. Ramos, and J. Neves, “The Fabricare scheduling prototype [132] Y. Zhang, S. C. Feng, X. Wang, W. Tian, and R. Wu, “Object oriented
suite: Agent interaction and knowledge base,” J. Intell. Manuf., vol. 14, manufacturing resource modeling for adaptive process planning,” Int. J.
no. 5, pp. 441–456, 2003. Prod. Res., vol. 37, no. 18, pp. 4179–4195, 1999.
[108] J. A. Stori and P. K. Wright, “A knowledge-based system for ma- [133] F. L. Zhao and P. S. Y. Wu, “A cooperative framework for process
chining operation planning in feature based, open-architecture manu- planning,” Int. J. Comput. Integr. Manuf., vol. 12, no. 2, pp. 168–178,
facturing,” presented at the ASME Design Technical Conf., Irvine, CA, 1999.
1996. [134] F. L. Zhao, S. K. Tso, and P. S. Y. Wu, “A cooperative agent modeling
[109] N. Sugimura, R. Hino, and T. Moriwaki, “Integrated process plan- approach for process planning,” Comput. Ind., vol. 41, no. 1, pp. 83–97,
ning and scheduling in holonic manufacturing systems,” in Proc. IEEE 2000.
ISATP2001—Assembly and Disassembly in the 21st Century, May 28–29, [135] W. H. M. Zijm, “The integration of process planning and shop floor
2001, pp. 250–255. scheduling in small batch part manufacturing,” Ann. CIRP, vol. 44, no. 1,
[110] J. Sun, Y. F. Zhang, and A. Y. C. Nee, “A distributed multi-agent en- pp. 429–432, 1995.
vironment for product design and manufacturing planning,” Int. J. Prod. [136] M. Zweben and M. S. Fox, Intelligent Scheduling. San Francisco, CA:
Res., vol. 39, no. 4, pp. 625–645, 2001. Morgan Kaufmann, 1994.
Weiming Shen (M’98–SM’02) received the B.Sc. Qi Hao received the B.Sc. degree in mechanical de-
and M.Sc. degrees in mechanical engineering from sign and manufacturing, the M.Sc. degree in elec-
the Northern Jiaotong University, Beijing, China, in tromechanical control and automation, and the Ph.D.
1983 and 1986, respectively, and the Ph.D. degree in degree in aeronautical and astronautical manufactur-
systems control from the University of Technologies ing engineering from the Northwestern Polytechnical
of Compiegne, Compiegne, France, in 1996. University, Shannxi, China, in 1992, 1994, and 1997,
From 1986 to 1992, he was a Lecturer in mechan- respectively.
ical engineering at the Northern Jiaotong University. From 1998 to 2001, she was an Associate Profes-
He has also held positions with MediReport, Paris, sor at the Northwestern Polytechnical University. She
France, and the Department of Mechanical and Man- has also held positions with Xi’an Butone, Inc. She is
ufacturing Engineering, University of Calgary. He is currently a Visiting Research Fellow at the Integrated
currently a Senior Research Scientist at the Integrated Manufacturing Technolo- Manufacturing Technologies Institute, National Research Council Canada, Lon-
gies Institute, National Research Council Canada, London, ON. He is an Adjunct don, ON. She has authored or coauthored more than 30 journal and confer-
Professor in systems design engineering at the University of Waterloo and an ence publications. Her research interests include manufacturing process plan-
Adjunct Professor in software engineering at the University of Western Ontario. ning and scheduling, ERP/MRP, virtual enterprises, supply chain management,
He has authored more than 180 papers in scientific journals and international and applications of intelligent agents and Internet/Web-based technologies to
conferences/workshops, and coedited 11 conference/workshop proceedings in these areas.
the related areas. He is also the author of Multi-Agent Systems for Concurrent
Intelligent Design and Manufacturing.
Dr. Shen is a Member of ASME, ACM, AAAI, and CSCSI. He is a Registered
Professional Engineer in Ontario.
Lihui Wang received the B.Sc. degree in machine de-

sign from the Beijing University of Printing, Beijing,
China, in 1982, and the M.Sc. degree in mechanical
engineering and the Ph.D. degree in intelligence sci-
ence from Kobe University, Kobe, Japan, in 1990 and
1993, respectively.
He has worked for two years at Kobe University
and for two years at the Toyohashi University of Tech-
nology, Japan, as an Assistant Professor. Currently,
he is a Senior Research Scientist at the Integrated
Manufacturing Technologies Institute, National Re-
search Council Canada, London, ON. He is also an Adjunct Professor at the
University of Western Ontario, London, ON. He has published more than 100
papers in refereed journals and conference proceedings, and has edited four
conference proceedings in manufacturing research.
Dr. Wang is a Senior Member of SME and a Member of the Japan Society
for Precision Engineering. He is a Registered Professional Engineer in Ontario.

Agent-Based Distributed Manufacturing Process Planning and Scheduling: A State-of-the-Art Survey

Uploaded by

Copyright:

Available Formats

Agent-Based Distributed Manufacturing Process Planning and Scheduling: A State-of-the-Art Survey

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Agent-Based Distributed Manufacturing Process Planning and Scheduling: A State-of-the-Art Survey

Uploaded by

Copyright:

Available Formats

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 36, NO.

4, JULY 2006 563

Agent-Based Distributed Manufacturing Process

I. INTRODUCTION II. PROBLEM COMPLEXITY

Lihui Wang received the B.Sc. degree in machine de-

You might also like