Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 Systems-of-Systems, Distributed Simulations, and Enterprise Systems Three topics in modeling and simulation (M&S) receive special attention in this report: (1 ) the increasing complexity of systems-of- systems and the corresponding demands on modeling and simulation capacity; (2) the increasing desire for distributed simulations and their corresponding technical requirements; and (3) the long-term goal of having enterprise systems, or M&S systems that include all the aspects of a business enterprise from product development to manufacturing, to human resources, cost accounting, marketing, and sales. These three topics cut across several areas of M&S for defense acquisition and commercial manufacturing and present particular challenges for research and development. SIMULATING COMPLEX SYSTEMS-OF-SYSTEMS In both the commercial and defense worlds, the need to model complex systems-of-systems is increasing. Commercially, two examples are complex, multiunit manufacturing systems, and supply-chain systems in which interoperability between retailers and suppliers is demanded. One of the major challenges that DOD faces is the creation and sustainment of systems-of-systems to satisfy mission needs. It has been argued that future efforts to modernize DOD's weapons systems should put more emphasis on novel system concepts (Birkler et 63
64 MODELING AND SIMULA TION IN MANUFACTURING al., 2000). The present acquisition process, however, has difficulty in accommodating either rapid definition and development of such systems or implementation of new operational concepts. Accelerated development and demonstration of new concepts would be needed prior to the commitment of fu]] funding or fielding. In addition, the simulation-based acquisition (SBA) process envisioned for DOD requires an assessment of expected mission effectiveness early in the development of a new system. Mission effectiveness is a result of a system's ability to gather and share information and to survive and attack hostile targets. An assessment of such abilities is extremely difficult to make. Subtle design decisions can result in significant impacts on mission effectiveness. A modest alteration in the way a defense system is used or a minor modification to the scenario in which the system is immersed can also have a marked impact on effectiveness (Hall et al., 2000; Hall et al., 19991. It is becoming more accepted within the defense community that M&S technologies are an essential, and possibly the only, means of exploring, evaluating, and assessing the complexity of modern warfighting environments. The establishment of capabilities for representing systems- of-systems that work together to meet aggregated mission requirements, as well as integration and interoperability strategies, is therefore important. An M&S environment to support systems-of-systems evaluation would assist in evaluating a proposed defense system's mission effectiveness in the context of a specified set of possible design variations, operational use patterns, and engagement scenarios. Such an environment would need to contain a library of sensor, weapon, and command and control communication platform models that could be composed to model military systems-of-systems operating in physically realistic environments. It would need to support discrete event simulations involving large numbers of runs with different random number seeds and parameter settings for Monte Carlo sampling and optimization searches (see the subsection "Dealing with Complexity and Errors," in Chapter 5~. Since it might have to support a large number of mobile, communicating entities at a significant level of resolution, this M&S environment would need to be built on a middleware~ layer that efficiently manages interactions among such entities and their environments. In addition, such an environment would need to support ergonomic and informative human-computer interfaces, including visualization interfaces to display spatially referenced entities. It would also need to "'Middleware" is software that simplifies the use of network technology in applications by providing for sending message packets from one node to another. These services would otherwise have to be programmed from basics. Middleware enables large mainframe applications to migrate to distributed client/server applications and provides communication facilities across heterogeneous platforms.
M&S IN COMMERCIAL MANUFACTURING 65 support interactive scenario and experiment definition and analysis, including data modeling and statistical analysis of simulation output. Systems-of-systems are complex. One ofthe defining characteristics of complex systems is emergence, or emergent behavior. A complex system, such as a system-of-systems, can exhibit behaviors that are different from those of the separate agents or entities, such as the individual systems, that compose it (Jervis, 1997~. The aggregation of the agents' behaviors and the interactions between them can generate large- scale emergent behavior that is not part of the behavior repertoire of any of the agents and may be qualitatively different from them. Emergence manifests itself as the organization of consequential higher-order behavior from the separate agents. Even if the agents have no specific organizing behaviors, an overall organization can emerge as a consequence of the behavior of the individual agents and the interactions between them. Virtually all organizational behavior in such systems results from agents adapting to their environments and, in the process of so doing, affecting the environments ofthe other agents (Sage and Olson, 2001~. In a complex system, emergent behavior arises in a bottom-up fashion as the combination of the behavior of many individual agents. However, knowledge of the behavior of the agents does not allow prediction of the behavior of the entire system. Therefore, the effectiveness of the traditional reductionist approach to studying systems is significantly reduced when applied to complex systems: "It is not sufficient to think of the system in terms of parts or aspects identified in advance, then to analyze those parts or aspects separately, and finally to combine those analyses in an attempt to describe the entire system" (Gall-Mann, 1997~. What is the significance of emergence for simulating systems-of- systems? An M&S environment to support systems-of-systems evaluation must enable emergence to assess systems-of-systems performance realistically. The modeling methods used in the M&S environment must include those in which emergence is possible; generally, such methods explicitly model interagent interactions. Reductionist modeling methods, as noted, will often not be sufficient. DISTRIBUTED SIMULATIONS Advanced distributed simulations, which allow multiple participants connected by a network to interact simultaneously with each other, are becoming increasingly important in military and manufacturing applications. There is steady growth in the number of participants and amount of information being shared. Two examples of advanced
66 MODELING AND SIMULA TION IN MANUFACTURING distributed simulations are distributed mission training systems and distributed collaborative engineering environments. Distributed Mission Training Systems Distributed simulation training systems have successfully replicated battlefield conditions for soldiers and pilots. However, as the complexity of command and control systems continues to grow, an unmet need has arisen for joint coordinated mission training for participants at higher levels of military hierarchies. The payoff to the warfighter will be the ability to conduct various phases of mission execution within one training system, on demand, with minimal human-in-the-loop coordination. For example, a distributed simulation system for the U.S. Air Force Theater Battle Management Core System may include computer-generated forces, terrain, environment, and cognitive agents such as pilots to replace today's human- support role-players. The combat scenarios developed can provide training that differs from individual pilot training in several ways. Initially, skill or positional training requires small-scale forces and scenario elements. As the exercise level moves toward team or crew training, the simulation must support the ability to scale the conflict toward a major theater war by including many tactical missions, rules of engagement, special instructions, and pre-mission planning considerations to handle the enormous number of combat situations that could arise. The wide range of capabilities required by such training systems demands an architecture that is scalable. Distributed Collaborative Engineering Environments With the advent of high-speed networking technology and recent advances in modeling technology, a distributed and collaborative engineering environment is closer to reality. Several previous studies and ongoing efforts have used various terms to describe related or similar concepts, including simply "collaborative environment" (SBATF, 1998~; "advanced engineering environment" (NRC, 1 999a); "advanced acquisition environment" (Hollenbach, 2001~; "collaborative enterprise environment" (William, McQuay, U.S. Air Force Research Laboratroy, briefing to this committee); and "collaborative engineering environment" (Crisp, 2002~; among others. This committee has chosen the term "distributed and collaborative engineering environment" to capture this general concept, some aspects of which are described below. A distributed and collaborative engineering environment would include realistic, multiscale simulation models of all components of a system, including human beings and engineered systems such as weapons
M&S IN COMMERCIAL MANUFACTURING 67 systems. Such an environment would also include the ability to define a complete system by composing models of individual components, thereby simulating the complete system behavior and performance long before it is built. Simulation models could be distributed and developed by multiple organizations in widely scattered locations in such a manner that they would interoperate. Specific "application services" could be provided for use by simulation federations such as calculations of electromagnetic propagation through the atmosphere, as is being done by the Joint Virtual Battlespace (JVB) federation using simulation services provided by various U.S. Anmy Research, Development, and Engineering Centers (RDECs) (Richardson, 2002~. Such simulation models or application services could be sold or leased on a per-use basis. A distributed and collaborative engineering environment would be a completely digital, networked software system in which design and manufacturing engineers and organizations would collaborate to simulate and design complex new systems or to upgrade and maintain legacy systems. Such a design environment should have the ability to store comprehensive design information in virtual inventories so that one could quickly retrieve previous design solutions. The infonnation contained in these virtual inventories would encompass the complete life cycle of products, from initial design requirements to detailed functional decompositions, from computer-aided design (CAD) data to process plans. Moreover, this information would be stored not just as data but rather as semantically enhanced information, allowing for intelligent retrieval, fixture expansion, and sharing across interdisciplinary and organizational boundaries. Software tools would be seamlessly integrated using agent-based architectures.2 Such integration would occur at the syntactic level ("agentization" of software components) as well as at the semantic level (ontologies3 for interoperability), so that new tools could be dynamically incorporated into the system with little or no programming effort. Analysis of designs and replacement systems would occur completely digitally through virtual systems prototyping (Sinha et al., 20011. A composable simulation environment that integrates behavioral models with structural models would allow designers to quickly evaluate and compare a large number of design alternatives (Diaz-Calderon et al., 1998~. This design and simulation environment would accommodate both the fluidity and 2 An "agent-based architecture" is based on the notion of a software agent. A "software agent" can be thought of as a highly encapsulated piece of software that may be autonomous and also capable of negotiation. 3 In the context of knowledge sharing, the term "ontology" means a specification of a conceptualization. That is, an ontology is a description (like a formal specification of a program) of the concepts and relationships that can exist for an agent or a community of agents. Researchers have designed ontologies for the purpose of enabling knowledge sharing and reuse (Gruber 1993).
68 MODELING AND SIMULATION IN MANUFACTURING uncertainty inherent in earlier functional and conceptual design and the level of complexity involved in detailed design. Beyond the functional verification of designs, virtual prototypes would also be used for the analysis of the manufacturing process, maintenance, repair, logistics, training, and disposal of systems. Therefore, these virtual prototypes would provide a common infrastructure enabling specialists in different disciplines to evaluate the system from all life-cycle viewpoints. Support for the evolutionary aspect of both design information and the design process would exist. Instead of assuming that static considerations led to each design decision, design representations in the virtual inventories would capture the full design rationale not only the design options that were chosen, but also other possible options, and the information used to choose among them. In addition, evolution of the design process would be facilitated through dynamic, extensible ontologies. The existence of virtual inventories would significantly reduce the lead time and the cost of on-demand manufacturing and procurement of replacement parts and subsystems. This would in turn reduce both downtime of critical defense systems and the life-cycle cost of deploying these systems. Virtual inventories would exploit the knowledge residing in legacy systems that contain designs and process plans by providing ways to index data for search and retrieval and by adapting the data to fit new technological capabilities and new design requirements. The potential benefit is large, as it is widely recognized that practicing engineers spend large portions of their work time searching through legacy data, catalogs, and earlier engineering projects. It would permit consideration of a wide range of conceptual designs through use of tools that could provide fast, easy solutions with sufficient fidelity for preliminary consideration of designs. It would minimize errors, especially in long-term programs and system upgrades, by permitting the capture of design rationale for parts and systems and identifying ramifications of design changes during the product life-cycle, preventing many costly and potentially life-threatening errors. A distributed and collaborative engineering environment would give designers the ability to assess whether functional design requirements are met, with need for much less physical prototyping. Perhaps only in the final design stage might a physical prototype be necessary. Elimination of physical prototyping would result in considerable savings of both time and money. Designers would be able to evaluate more design alternatives and receive immediate feedback on design decisions, thereby developing better final designs. Virtual prototypes would also allow designers to consider life-cycle issues, such as evaluating manufacturing and assembly requirements throughout the design process, thereby avoiding costly and time-consuming engineering change orders. The prototypes would further
M&S IN COMMERCIAL MANUFACTURING 69 make it easier to generate maintenance and service instructions. The use of agents and ontologies would provide a way to coordinate the activities of multiple designers who might be separated in both geography and time. By combining an agent-based architecture with the use of ontological translation mechanisms, integration would be provided at both the syntactic and semantic levels. This would make it easy to integrate a wide range of heterogeneous tools, including a variety of legacy systems (e.g., CAD tools and databases) and tools yet to be developed. Overall, improved modeling and simulation tools would contribute to the ability to produce more robust designs and permit their integration into complex systems-of- systems, as well as improving the ability to support these design over extended product life-cycles. Limitations on Advanced Distributed Simulations The growth in the number of participants and the desire to share greater amounts of information are placing increasing demands on bandwidth and computational power. Attempts to overcome bandwidth limitations have tended to concentrate either on increasing the available bandwidth or on minimizing the demand for bandwidth made by applications. Methods of minimizing demand include data compression and multicast routing systems that incorporate software-based area-of- interest managers to direct packets across a network to particular groups of listeners. High level architecture (HLA) supports exploitation of multicasting hardware at the middleware layer by providing so-called data distribution management services. These services allow objects to specify attributes that they will publish and attributes to which they subscribe as well as associated regions in routing space. When publication and subscription regions overlap, attribute information flows from publisher to subscriber. The goal is to send data only where and when it is needed. To date there has been little interaction between these two approaches (NRC, 1997), although a solution to the bandwidth problem must come from a combination of these two to produce efficient use and allocation of bandwidth in accordance with application requirements. Any form of information technology combines information generation (computation and real-world inputs) and information transmission (communication). As is well known in computer performance modeling, for the best cost/performance ratio, generation and transmission must be matched to each other. If transmission capacity exceeds computational capacity, information overload occurs. Conversely, faster generation is pointless if transmission is the bottleneck. In today's technology, bandwidth demands exceed supply so that minimizing application bandwidth requirements and allocating bandwidth among applications are required to make simulation
70 MODELING AND SIMULATION IN MANUFACTURING possible. However, even with a significant increase in bandwidth supply, the latter approaches are still needed to match available computational capacity and maximize effective use of resources. ENTERPRISE SYSTEMS An "enterprise system" is a consistent suite of interoperable application programs that serves all major functions of a business enterprise, including product design, manufacturing, cost accounting, human resource management, sales and marketing, and purchasing. Enterprise systems had their origin in computer-integrated manufacturing (CIM) systems and evolved to materials requirements planning (MRP), enterprise-level systems, supply and value chains, and enterprise resource planning (ERP) systems, using commercial software tools such as Statistical Analysis System (SAS). The envisioned future defense acquisition process, SBA, will have to operate within the context of future commercial enterprise systems. SBA must therefore include consideration of the enterprise level of system acquisition that integrates concern for manufacturing with other major enterprise functions, such as cost accounting, human resource management, sales and marketing, and purchasing. Computer networks and the lnternet, in particular, have become a universal medium for enterprise-level software deployment. The network operating environment now greatly stretches the range of scalability, from a few users to millions of simultaneous users. This is true not only of consumer-oriented retail operations on the Internet, but also of business-to- business e-commerce and deployment of enterprisewide systems. The Internet is increasing in connectivity and node capability (Stiles, 2001), providing a highly interconnected and computationally powerful medium for companies to increase outsourcing arrangements and self-organize into v irtua] enterprises (B in stock, 2 0004 . However, many obstacles to achieving a true enterprise system remain. The increased connectivity and capability create new complexity and dynamics (the Internet as a holistic system), which have been neither fully understood nor addressed. In addition, designers of the architectures of such extended enterprise systems must consider many factors, including organizational issues governing the interactions of people and organizational structures and constraints governing these interactions; collaborative decision making, including the sharing of collaborative knowledge and the protection of proprietary knowledge; supply-chain structures for the integration, coordination, and management of activities in
M&S IN COMMERCIAL MANUFACTURING 71 networked environments; design approaches that facilitate outsourcing or strategic alliances; and decentralized mechanisms or policies that cause desirable emergent behaviors. Finally, the design of enterprise systems demands scalable system architectures. The committee used information on enterprise-level modeling and simulation functions developed by IMTI and extended it to form Table 4-1. The first two columns of the table summarize IMTI conclusions regarding the current state of practice and the ideal state of enterprise modeling and simulation functions. The committee developed the material in the remaining two columns regarding real-world limitations on each function and the requirements needed to achieve the ideal state. As indicated in Table 4-1, limitations on the development of enterprise modeling and simulation functions include the availability of models at the strategic, industry, and technology level; the availability of enterprise-level cost and resource models; the availability of supply-chain models; and limitations on communication and computation resources, such as bandwidth, computation speed, and memory. Requirements to reach the ideal state include scalable enterprise systems; integrated model frameworks with real-time data management; standard frameworks for model construction; and integration of scalable enterprise systems in the supply chain. Generalized Er'`erprise Reference Architecture and Methodology "Enterprise engineering" is a term used for the set of activities dealing with designing and redesigning business entities, either industrial systems, administrative systems, or service systems (Vernadat, 1996~. Enterprise engineering goes through the following stages: business entity identification; business entity conceptualization and definition; requirements definition; design specification; implementation description; building and testing; and finally, release of the system into operation. Once the business entity is in operation, such activities may continue with performance evaluation, change management, and continuous process improvement. A framework for enterprise engineering recently proposed by the enterprise integration community the Generalized Enterprise Reference Architecture and Methodology (GERAM:uses the computer- integrated manufacturing open system architecture (CIMOSA; see below). This architecture, shown in Figure 4-1, can be used to understand the interrelationships among enterprise processes, application domains, and industries, and it provides an overarching framework/architecture for integrating different simulation models in different domains.
72 MODELING AND SIMULA TION IN MANUFACTURING e E ~ ~ S ,~ ,cn ,a' a) ~ ~ a' ao a) a) -, ~ _ a, ~' ~ cn E =, =, ~ 3 a' ~ ~ ~ ° E a) 0 ~ ~ a) a) ~ O (D Q a) O) ( ~ ,o ~ ' o ~, ,oo ~ <t, E ~ ° ~ _ E ° ''' ° E ~ co ou ,~ E E ° C[) C~~ CD ° C~ o ._ o ._ Ct 3 ._ V: S~ Cd - o V) ._ ~L C) 1 LU m a) - co cn c~ a' c~ cn - _ . a) a) CO Q Q (n _C q 8 = e _ , b _ c L ~ r ~ y O r ~ ~ ~ _ c ~ E -,,, c,, ,,, o: ~ ~ , E (D ~ ,,, ~ C w 2 ° E o =: 2 a> ~n ~ ~ ~ 3 cn _ co ~ c u~ ~ c c~ o
M&S IN COMMERCIAL MANUFACTURING 73 arc E ' ~ E E E ; e 0 0 , ~ ~ -= e ,~c ~ | ~ ~ E o tt~ a=>' <0 C) ~ == e t~~~ I An An .~ O CD _ ci' . O A ~ _ o ._ Cal a: o ._ ._ Cal 5 ~: Cd ._ - o C~ ._ 5~ r~ m a' CO - CD - a) a, CTS - ~n - a C~ . _ CD Q Q CO o ._ . _ Q ~ a g =0 Cl: cn Cl) ~ 3 ~ ~ ~ ~ ~ ~ ~ 3 0) Q U) a~ E E ~ ~~, o~ C) o ~0 0) ~= E '~ ~0 E ~ ~ e ~ e °~ 0 0 =~ =0 0 0 ~ 0 E 0 0 ~ o5 ~ 3 ~ _ 3 8 o ~,8~ ,,,~ ~_E5~o _% o o o . . C) o C~
74 MODELING AND SIMULA TION IN MANUFACTURING (ienenc 1:~'~ws: -Lyams Em/ Cor'cepl: P`~w'~ment Des'g.n . , , ~., Im~ernant~ior~ Oper~mn D~mmi"~n P=t38t it., ~~ Gerard: ~~b . ~ -_ Be, at_ Reference _ If: ~ models / am P~irm~r mad ma'''''''''' - Or~n'~a~n ~ ~_. dot : 1~- l~;i~fl~1~i: , ~ Information _ ~ Hanson ,i~ Control Oman (operative Figure 4-1 GERAM architectural framework. Source: Reproduced as submitted by the IFAC-IFIP task force to ISO TC184/SC5/WG] for inclusion as an annex to ISO WD]5704, "Requirements for enterprise-reference architectures and methodologies." GERAM and CIMOSA are important in developing standards for SBA's dependence on enterprise integration and the role of M&S in supporting such integration (see Chapter 5~. CONCLUSIONS The next-generation M&S capabilities discussed in this chapter are essential to the achievement of the future defense acquisition vision. The ability to simulate complex systems-of-systems is essential for evaluating the mission effectiveness of future weapons systems within SBA. Advanced distributed simulations are essential for achieving combat-
M&S IN COMMERCIAL MANUFACTURING 75 training simulations that include models of weapons systems and for achieving the vision of distributed and collaborative engineering for commercial and defense manufacturing. Finally, enterprise systems are the ultimate goal for M&S in commercial manufacturing. SEA will have to function in the context of these systems and will benefit from the development of the capabilities needed. In order to achieve the capabilities for simulating complex systems- of-systems, additional R&D is needed in the following areas: library of sensor, weapon, and command and control communication platform models that could be composed to model military systems-of-systems operating in physically realistic environments; a simulation environment that supports Monte Carlo and/or optimization simulation involving large numbers of runs with different random number seeds and parameter settings; a simulation environment capable of efficiently managing the interactions among a large number of mobile, communicating entities at a significant level of resolution; a simulation environment that supports ergonomic and informative human/computer interfaces; a simulation environment that supports interactive scenario and experiment definition and analysis; improved organizational behavior models; and improved human behavior models. In order to achieve the necessary capabilities for advanced distributed simulations, additional R&D is needed in the following areas: scalable architectures; realistic, multiresolution models; 4 composability; virtual inventories to include life-cycle information on products and history and rationale for design; agent-based architectures; and management of bandwidth and computational power limitations. In order to achieve the necessary capabilities for enterprise systems, R&D is needed in the following areas: scalable enterprise systems, integrated model frameworks with real-time data management, standard frameworks for model construction, and integration of scalable enterprise systems in supply chain. 4 The need for research on multiresolution models is explained in Chapter 5.
