Functional Integrity of Some Class of Multi-Agent Systems

Krzysztof Cetnarowicz

Abstract



The interest in decentralized systems that arose in the eighties led to the development of multi-agent systems based on the concept of an autonomous agent. Multi-agent systems have a number of advantages and open up many new possibilities in creating IT systems, but many problems related to the operation of these systems remain to be solved. Such problems include the functional integrity of multi-agent systems. The functional integrity of a multiagent system can be defined, generally speaking, as the preservation of the basic functions of the system during its functioning. It can be analyzed from the point of view of various system functions (which should remain preserved) as well as from the point of view of various factors that may affect the loss or preservation of functional integrity by the system. The study examines the functional integrity of a multi-agent system depending on the number of agents (global and individual types). Agents during the system work generate agents of the same or different type, depending on their capabilities and needs of the system. The agent, having the opportunity to perceive the environment in his environment, does not have direct access to certain global information in the system. As a result, in many cases a given agent, when generating a new agent, cannot take into account all the important factor systems from the point of view of functional integrity. 
As a consequence, this leads to an excessive number of agents (system blockage), or an insufficient number, and to a complete lack of agents of a given type (loss of system function implemented by agents of a given type). The authors analyzing the behavior of multi-agent systems [8] point out that in addition to the numerous advantages that decentralized systems may have, they may also have a significant disadvantage. It consists in the fact that system elements, acting in a decentralized way, too often will take the initiative to communicate with other system elements. This will cause a large number of transmissions that are often unnecessary, and as a consequence, overloading the system with excessive transmissions and, as a result, a decrease in system performance, or even blocking it. The considerations were based on a simulation study of a certain class of multi-agent systems.


Full Text:

PDF

References


Cetnarowicz K.: M-agent architecture based method o f development o f multiagent systems. In Proc. o f the 8th Joint EPS—APS International Conference on Physics Computing, ACC Cyfronet Kraków Poland, 1996

Cetnarowicz K.: Technology o f decentralized midti-agent system creation based on the M-agent architecture. In Proc. MIFSYD’96, II-AGH, Leibniz/IMAG, Technical Rep. No 11.1/96, Institute of Computer Science AGH Kraków, Poland, 1996

Cetnarowicz E., Nawarecki E., Cetnarowicz K.: Agent oriented technology o f decen tralized system based on the M-agent architecture. In Proc. of the MCPL'97, IFAC/IFIP conference, CTI - Technological Center for Informatics Foudation, Campinas - SP, Brazil, LAG - Grenoble, France, BIBA, Bremen, Germany, 1997

Cetnarowicz K., Nawarecki E., abińska M.: M-agent architecture and its applica- tion to the agent oriented technology. In Proc. o f the DAJMAS’97. International work- shop: Disributed Artificial Intelligence and Muld-Agent Systems, St. Petersburg, Russia, 1997

Demazeau Y., Muller J.P., editors: Decentralized A.I. North-Holland ISBN 0^144—88705-9,1990

Demazeau Y., Muller J.P., Perram J., editors: Modelling Autonomous Agent in a Multi- Agent World. Odense University, 1994

Wallach Y.: Alternating Seąuential/Parallel Processing. Berlin, Springer-V erlag 1982

Wooldridge M., Jennings N.: Formalizing the cooperativeproblem solvingproces. In Demazeau et al. [7], 15-26




DOI: https://doi.org/10.7494/csci.1999.1.1.3571

Refbacks

  • There are currently no refbacks.