Multi-Agent Communication Protocol in Collaborative Problem Solving : A Design Science Approach

Contemporary multi-agent systems (MAS) face significant communication challenges, particularly in dynamic environments characterized by heterogeneity, uncertainty, and scalability demands. Existing protocols, such as FIPA-ACL and KQML, are inherently static, insufficiently adaptive, and unable to efficiently integrate heterogeneous agents. Addressing this gap, the objective of this study is to design, implement, and evaluate a dynamic, context-aware, and scalable communication protocol that facilitates effective collaboration among heterogeneous agents (ML-based and rule-based) in complex problem-solving scenarios. Employing a Design Science Research methodology, the study developed a hybrid protocol integrating vertical (hierarchical) and horizontal (peer-to-peer) coordination, supported by middleware-driven dynamic role negotiation, conflict detection, and adaptive resolution strategies. Experimental results indicate notable improvements in throughput, context-awareness, and conflict resolution efficacy compared to traditional static protocols. However, mediator agents exhibited performance bottlenecks under high communication loads, and scalability constraints emerged under extensive scale-up scenarios. Despite simulation-based methodological limitations, the developed protocol demonstrates significant potential to enhance collaborative decision-making within MAS. This research contributes a robust, adaptable, and practically validated communication framework, offering pathways for future enhancements in mediator optimization, real-world validation, and advanced dynamic learning mechanisms.