PLANET Technology Information Day Prague, May 26, 2003
15:20 - 16:10

Michal Pěchouček (Czech Technical University, Czech Republic)
Agent-Based Production Planning

Abstract

The field of distributed artificial intelligence and multi-agent systems provides production planning and intelligent manufacturing systems in general with three distinctive pieces of technology: (i) architectures and design methodology for development of integrated enterprize resource planning systems (ii) technologies for distributed decision making algorithms and planning in particular and (iii) technologies for incorporeality, agentification, and legacy systems integration. There were several different approaches how to implement planning and balanced resource allocation in complex distributed systems (e.g. auctioning and advanced methods of negotiation, reactive agents and stigmergic approach, etc).

Within the presentation I intend to deliver a balanced overview between theoretical achievement and practical applications of the dsitributed planning technology in the Gerstner Laboratory, Czech Technical University. I will focus primarily to the concept of social knowledge and acquaintance models and show their potentials in planning application. The use of acquaintance-model-based interaction will be illustrated on the ExPlanTech multi-agent system. ExPlanTech has been implemented in order to support planning of patterns and forms manufacturing in an important Czech car industry supplier.

Agent based system is a collection of autonomous computational elements (independent programs) that perform collective behaviour in order to meet either their individual goals (self-interested agents) or an in-community shared goal (collaborative agents). An agent's acquaintance model is a computational model of agents awareness of the other existing agents, their properties, services, free resources, models of past and future behaviour, etc. In complex agent-based systems the collective decision making may involve excessive amount of agents' interaction. We claim that agents may make a reasonably good decision, that is based on imprecise information stored in their acquaintance models and they can adapt minor discrepancy and solve possible conflicts by classical negotiation techniques, rather than searching the entire space of possible collaboration patterns.

Biography

Dr. Michal Pěchouček works as assistant professor in Artificial Intelligence at the Department of Cybernetics, CTU FEE. He graduated in Technical Cybernetics from FEE-CTU in 1995, got his M.Sc. degree in IT: Knowledge Based Systems from University of Edinburgh and completed his Ph.D. in Artificial intelligence and biocybernetics at CTU in Prague in 1998. He is the Head of the Agent Technology Research Group at the Gerstner Laboratory for Intelligent Decision Making and Control. His research focuses on problems related to social knowledge, acquaintance models, forming coalitions, monitoring, meta-reasoning and reflection in multi-agent systems while at the same time he is interested in industrial applications of agent technologies. Dr. Pěchouček participated in and coordinated several international projects (EC funded) and acted as a principal investigator of direct research contracts (Rockwell Automation, US Air Force, Office of Naval Research, NASA). He is an author or co-author of about 50 publications in proceedings of international conferences and journal papers. In addition, he is the co-chair of the International Workshop on Industrial Applications of Holonic and Multi-Agent Systems (HOLOMAS 2000, 2001, 2002) and CEEMAS2003 - Central and Eastern European Conference on Multi-Agent Systems. Besides, he is a PC member of several workshops and conferences related to multi-agent system research. Michal Pěchouček consults to Rockwell Research Center in Prague and is a senior research consultant in CertiCon, corp.