Citation

Balducelli C, D'Esposito C. Safety Sci. 2000; 35(1-3): 59-73.

Copyright

(Copyright © 2000, Elsevier Publishing)

DOI

unavailable

PMID

unavailable

Abstract

Emerging efficient and intelligent behaviors from co-operative activities of many autonomous agents comprise unexpected events that often take place inside animal social organizations. This type of evidence induced artificial intelligence researchers to re-design the artificial computing system architectures, from monolithic and hierarchical structures toward networked and component distributed environments. In this work the co-operative capacity of three different software agents is experimented to front the management problem, emerging during fires emergencies, relative to fire-proof resources optimization and dangerous products evacuation inside large oil storage and production plants. The three software components include, respectively, a repository or memory of past solutions (managed by case-based reasoning methods), the capacity to discover new solutions (using evolutionary algorithms), and the capacity to verify solutions (using numerical simulation models). The main result of the work was that, from the co-operative activities of these three software components, together with the human agent, the capacity to learn and adapt solutions for the current problem arises as a proper and additional feature of such hybrid system. The general problem of anti-fire resources optimization and evacuation of risk products, during fire emergencies, inside a petrol-chemical plant, is firstly described. Then, the models and the software algorithms, implemented in the three mentioned components, are illustrated in the central part of the work. Finally, a set of test cases are reported, for different scenarios in the physical domain, experimented and analyzed.