Persistent Memory in Multi-Agent Systems: the OMAS Approach
Résumé
In long living multi-agent applications the problem of memory becomes important. Agents need to remember what they have done, maintain data persistently despite crashes, power failures, or turning the machines on and off. Other important problems concern the maintenance of agents and the evolution of the applications. In this paper we only consider systems of cognitive hybrid agents capable of multi-processing, and examine the various problems concerning the persistency of information during the agent lifecycle, ranging from keeping a piece of information while executing a specific task to organizing a long term memory for applications that need to run for many months. As an example of such system we take the OMAS platform developed at UTC.