KDI Home    Contact Us   

Center for Information and Systems Engineering (CISE) Emerging Technologies Weekend Web site

New Research Findings Speed Up Manufacturing Supply Chains and Decongest Communication Networks

Innovative research in multiple scale based complex system decomposition and algorithm development is breaking ground across a variety of scientific fields. A multi-disciplinary team of scientists from among the nation's top universities and research organizations has decentralized decisions in complex systems that bog down manufacturing, tie up Internet traffic, and stifle growth and development in others areas of the global economy. Funded by the National Science Foundation's Knowledge and Distributed Intelligence (KDI) initiative, the research program has provided major industries with a scientifically valid set of methods for making better planning and operational decisions faster and more efficiently than previously possible. These methods have advanced the disciplines of operations research and systems science, decision making and control, and quality of service and coordination, resulting in positive implications for world markets.

Starting with the Manufacturing Enterprise

With the goal of bringing about "synergistic and decentralized decision making in complex systems," this pioneering work began five years ago in the area of manufacturing, accompanied by applications in communication networks and other scientific fields including computational physics and complex social and economic organizations. Manufacturing enterprises provided a good starting point for this KDI research for a couple of reasons. One, the economic and technological importance of manufacturing is compelling. Two, the scope of manufacturing is wide, thereby fostering inter-disciplinary research that is applicable to diverse areas. The ability to cross-fertilize research is an important objective of the KDI program.

What Makes Manufacturing Complex

In order to understand what makes manufacturing an appropriate paradigm for the research program, it helps to know why the enterprise is complex, large-scale, and stochastic. Advances in technology have changed the way manufacturers operate. Not since the 1800s, when machines powered by electricity and gas revolutionized industry worldwide, have technological advances had such a major impact on the production of goods. The proliferation of technological breakthroughs has given rise to elaborate systems that are vulnerable to inefficiencies and random events.

The random events are known as "stochastic." This esoteric term means something that is difficult to predict, such as the weather. An example of a stochastic event in manufacturing would be machine failure—although Murphy's Law guarantees that your printer will jam at the precise moment you are rushing to meet a deadline. Nonetheless, we really don't know when vital office equipment will cease to function. We simply know the outcome: frustration and disruption to workflow. Now, multiply those affects several times over and you get an idea of the enormous impact that stochastic phenomena can have in manufacturing. A random event in manufacturing can transcend every layer of a production system, affecting quality control, production time, and costs. If not quickly and appropriately managed, a stochastic event can cause system-wide failure and disruption. It boils down to how well and how timely decisions are made.

Decision Making in Manufacturing

As highly stratified systems, production chains are guided by a number of multi-faceted, interconnected decisions. The decisions range from frequent to infrequent, the outcomes of decisions from predictable to random, and the impact of decisions from immediate to long-term. For example, a decision to make capital investments in a plant is less frequent and more time consuming than the second-by-second decision making that goes into operating a metal-cutting tool. In addition, those separate decisions are made by different individuals or teams out of necessity since centralized decision making is impractical. However, manufacturing enterprises today engage in limited coordination and information exchange amongst decision makers. Certain decisions such as Materials Requirement Planning are more often than not made on the basis of inaccurate information about the capabilities and needs of manufacturing facilities and customers resulting in shortages or excessive inventories and obsolescence. Inefficient decision making processes are unfortunately relied upon today in the absence of tools that promote enterprise integration and coordinated decision making. The enterprise of the new millennium can benefit from decentralized, autonomous decisions of teams within the enterprise that act in concert to achieve a mutually beneficial dynamic equilibrium. The new research findings have the potential to achieve this goal likening enterprise decision making teams to the various organs in a healthy human body that act and react to external and internal stimuli to achieve homeostatic control. In this homeostatic control paradigm, decision making is autonomous yet coordinated, creating synergy between the multiple, intertwined layers defining large-scale complex systems.

The research team from Boston University, Massachusetts Institute of Technology, Tufts University, and Los Alamos National Laboratory has been investigating how to decentralize and create synergy in complex stochastic systems. The team members have been applying their expertise in control theory, systems engineering, computational physics, mathematics, economics, electrical engineering, and computer science to provide organizations operating in a competitive global economy with a novel set of tools and methodologies for achieving homeostasis.

Promising Outcomes

Designed for general use, the methodology is realizing promising outcomes in the separate fields of manufacturing and network communications. Supply chains involving production, transportation, storage, and distribution facilities that circle the globe deliver most consumer products today. Each facility in these long supply chains makes production decisions, planning, and executing to the best of its capabilities. However, coordination along the supply chain requires exchange of information that is sufficient to regulate individual facility actions that minimize excessive shortages or inventories across facilities. Today's limited information exchange approach to supply chain coordination assumes that products flow through the supply chain with a constant delays incurred at each facility. This simplification of the individual facility dynamics is convenient but results in less than optimal coordination.

The KDI-funded research has developed collaborative decision making tools that rely on the decentralized solution of multiple sub-problems—one at each facility—that compute and summarize the salient features of facility specific stochastic dynamics and convey them to a coordinating master problem. The coordinating master problem converts the actual delay characteristics of individual facilities to production targets that guide each facility to behave in a manner consistent with overall supply chain objectives. Pilot studies performed with real industrial size problems indicate that the inventory and shortages (or stock outs) are reduced to better than half the level encountered under today's state of the art industry practice. The ability to reduce inventories and speed up the supply chain is of enormous significance. High inventories and long delays are not only costly. They run counter to rapid technological progress and innovation that continuously improve product quality rendering inventoried merchandise obsolete and worthless.

Likewise, the Internet is a major enabler in today's global economy, providing a widely used medium for communication and commerce. As one of the most popular and increasingly accessible advances in technology, the Internet has transformed the way the world shares information and makes decisions. Internet usage—from e-mail, to online research, to e-commerce, to other applications—has accelerated the pace of society and the demand for faster, better services.

New technological advances and recent developments in research are leading the way towards an "enhanced" next-generation Internet. This new medium is expected to surpass the current "best effort" capability of the existing technology and evolve into a network that can provide multiple services at multiple quality-of-service grades to accommodate various consumer needs. The problem of congestion on the information superhighway is as nerve-racking as bumper-to-bumper traffic on city streets. Like overcrowded arterial roads and thoroughfares, Internet routers are often operating above capacity. These online traffic jams can compromise the quality of service provided on the Internet, leading to larger delays and frustration. Congestion will have an even greater impact on so-called real-time services including streamed video, digitized voice (Internet telephony), or access to some online application that is sensitive to delays (e.g., online trading, access to databases, net-meeting with exchange of multimedia content) Even rare congestion phenomena can lead to severe degradation of the quality of real-time services.

The research team has developed two main approaches for tackling the congestion problem and reliably support real-time services. The first approach involves the introduction of proper protocols to regulate traffic that enters the Internet and ensure compliance with appropriate quality-of-service practices. To that end, the researchers have relied on the mathematical theory of large deviations that enabled them to assess the likelihood of rare events. The second approach rests on the introduction of congestion-dependent pricing for Internet services, pretty much like congestion road pricing used in Singapore and recently introduced in London. The researchers have developed pricing mechanisms that can smooth spikes in demand and alleviate congestion by providing proper incentives to users.

Learn More about the Research Findings

The multiple scale decomposition and algorithmic development results and findings have been already adopted by the research community. Los Alamos National Laboratory scientists are using extensively and developing further the Computational Physics advances related to this research. Pilot studies have been completed on industrial scale data that demonstrated the significance of this production planning and supply chain quality of service work. Widespread industry adoption of breakthroughs of the type achieved in this KDI project usually lags by a decade. For example, the existing advanced planning and scheduling systems available now commercially were developed after 1995, whereas the research results were obtained in the 1980s. The interdisciplinary KDI team at Boston University, MIT, Tufts, and Los Alamos National Laboratory are working hard to disseminate results in many ways. They have relied on the well-tried avenues of publications, a Web site, conference presentations, and regular weekly seminars where industry is invited to participate along with graduate students and academics. The recently established Center for Information and Systems Engineering (CISE)—see www.bu.edu/systems—is promoting interdisciplinary research and industry interactions with the purpose of facilitating research transfer. An industrial advisory board to the department of Manufacturing Engineering at Boston University meets several times per year, attracting representatives from 15 plus industries with national and international presence. These industrial advisory board meetings, the industrial collaborations promoted by CISE and the placement of our Ph.D. students in the industry constitute our direct efforts towards research transfer. Finally, on November 2002, the Center for Information and Systems Engineering and the department of Manufacturing Engineering at Boston University organized an Emerging Technologies Weekend focusing on "Production and Supply Chain Logistics in the Global Communications Economy" that attracted more than a hundred industry representatives (see www.bu.edu/mfg/etseminar/mfg_prog_outr_emer_nov.html).

Back to Top of Page