Intelligent multiagent coordination based on reinforcement hierarchical neuro-fuzzy models.

Area of Science:

• Artificial Intelligence
• Multi-Agent Systems
• Computational Intelligence

Background:

• Intelligent multi-agent systems require sophisticated coordination mechanisms for effective operation in complex environments.
• Existing coordination strategies often struggle with scalability and adaptability in dynamic settings.

Purpose of the Study:

• To develop and evaluate two novel hybrid neuro-fuzzy models for hierarchical coordination of multiple intelligent agents.
• To introduce and assess two new coordination mechanisms: market-driven (MA-RL-HNFP-MD) and graph-based (MA-RL-HNFP-CG).

Main Methods:

• Integration of Reinforcement Learning Hierarchical Neuro-Fuzzy models with novel coordination mechanisms.
• Development of two specific models: MA-RL-HNFP-MD and MA-RL-HNFP-CG.
• Evaluation using benchmark applications: pursuit game and robot soccer simulation.

Main Results:

• Both proposed coordination mechanisms significantly improved multi-agent system performance.
• The MA-RL-HNFP-MD and MA-RL-HNFP-CG models demonstrated superior coordination capabilities.
• Performance gains were validated through rigorous testing on benchmark applications.

Conclusions:

• The developed hybrid neuro-fuzzy models offer effective solutions for hierarchical coordination in multi-agent systems.
• The novel market-driven and graph-based coordination mechanisms provide substantial performance enhancements.
• This research contributes to advancing intelligent agent interaction in complex computational systems.