Seminar by Prof. Arvind Bansal
Applying Bioinformatics Model for Fault Tolerant Multi -Agents
Prof. Arvind Bansal
Kent State University
Ohio, USA
Date: Mon, Nov 08, 2004
Time: 5:00 PM
Venue: CS-101
Abstract
Current day networked intelligent agent based systems can be classified as genetic algorithm based evolutionary systems, collective intelligence based team based cooperative systems, and game theoretic profit maximization based systems. These teams of agents are being used over the Internet. As the popularity and the size of Internet grow, the complexity of multi-agents systems would also grow. In past, evolutionary algorithms and heuristic optimization based techniques (such as ANT colony systems) have experimented with random changes (movement) and heuristic reward based evolution for self-adaptation. However, current multi-agent systems have limited capability of fault-tolerance and adaptability to handle extreme stress conditions. Since 1999 the bioinformatics study has provided great insight in the working of single cell based bacterial organisms and the stress pathways. We believe that bio-mimicking the natural hierarchy of modeling will provide the capability of natural fault tolerance and systemic level adaptation to handle sudden extreme stresses.
This paper introduces a new message based biological model of intelligent multi-agent based systems that represents agents as self-correcting dynamically modifiable genes ^ a reconfigurable set of dynamically regulated built-in functions, and system of agents as dynamically adaptable event-trigger controlled interacting pathways that can be altered and reconfigured in response to external stress and events. The model supports the integration of message, code, trigger, and belief states, and supports interchangeability of message, code, and trigger to provide dynamic adaptive control. The model and its implementation have been described.