Automatic Generation of Theories of Coordination in Multi-Agent Systems

Automatic Generation of Theories of Coordination in Multi-Agent Systems

N. V. Findler (Arizona State University, USA)
Copyright: © 2007 |Pages: 6
DOI: 10.4018/978-1-59140-789-8.ch017
OnDemand PDF Download:


Coordination is defined as the process of managing dependencies between activities. Its fundamental components are the allocation of scarce resources and the communication of intermediate results. Coordination theory can be defined as the set of axioms, constructs and analytical techniques used to create a model of dependency management in multi-agent systems (MAS). Multi-agent systems (MAS) represent a significant interest in a variety of disciplines, such as artificial intelligence (and other domains of activity in computer science), political science, international relations, public health, public policy, social welfare, economics, demography, anthropology, communication studies, geography, history, sociology, urban planning, control theory, electrical engineering, military science, and so forth. The present author and his students have worked on several systems in different domains of MAS, such as: a. Manmade Technical Systems: The technical and economic aspects of distributed automated air traffic control; distributed automated control of urban street and highway ramp traffic signals; learning, planning and collaborating robots; distributed control of nationwide manufacturing operations; distributed decision support systems for optimum resource and task allocation over space and time. b. Natural Complex Systems: The behavior of natural organisms, causality and temporal relations, social structures and coordination. c. Human Behavior and its Simulation: Language development and studies of the properties of dictionaries; behavioral studies on reasoning and decision making; information, fact and knowledge retrieval; automatic teaching and evaluation of control operators; social networks, social and cultural anthropology aids; multi-agent systems simulating human societies. One important concern is how to verify the correctness of the computer representation of MAS and how to optimize their operations. Theories of Coordination (ToC) should satisfy these requirements. However, such are currently ad hoc and amorphous, in that there is no unified model of coordination, though there exist many constructs describing specific phenomena in MAS. With the current advent of large-scale agent-based societies, there is a need for theories that builders can use in designing MAS, instead of being forced to learn from trial and error every time such a society is built. In addition to such design tools, we generate a trouble-shooting tool to diagnose problems with existing deficient systems. We envisage steady feed-back from this component to the user when the MAS malfunctions. We are creating software tools and formal techniques to be used to analyze and design1 systems of interacting intelligent agents. The concept is based on our experience gained in many projects, primarily in developing tactical and strategic decision support aids for use in dynamic command and control environments with multiple resources, multiple tasks, multiple sources of information and multiple human and machine decision makers that have different roles and responsibilities and belong to a hierarchy with overlapping jurisdictions. (This was completed during our multi-year collaboration with the U.S. Coast Guard in computerizing their tactical and strategic planning processes.)

Complete Chapter List

Search this Book: