Architecture of High-Performance Fuzzy Multi-Agent System
Ivan Parasiuk, Sergiy Yershov · 2013
Multi-agent simulation framework, which followed fuzzy type-2 fuzzy agent model and speeded-up the agent-execution and agent-messaging using computational power of GPUs in high-performance cluster environment, is developed. Architectural models of intelligent agents and multi-agent systems based on type-2 fuzzy logic, which enable to present more informative the degree of uncertainty of fuzzy rules in the specification of the behavior of these agents and systems. Approach for construction of fuzzy multi-agent systems in high-performance environment based on the specification of fuzzy graph transformation and adaptation of fuzzy rules by coordinator are explored. To reduce the cost of communication between agents we propose a method that is based on the generation of virtual blocks of agents on memory hierarchy of cluster system.