Abstract
Advancements in Internet technology increase the need for concurrent, robust and distributed systems. Existing software engineering methodologies can prove inadequate to address the design and development of such increasingly complex systems. Multi-agent systems are a popular alternative and prove highly adequate to address complex distributed challenges in real-world domains. However, multi-agent systems are subject to limitations and constraints. Firstly, multi-agent systems are commonly inadequate when addressing the real-time properties of complex systems in critical real-time constrained environments. Secondly, existing multi-agent systems commonly require substantial re-engineering to reuse or adapt them to alternative domains. The thesis introduces the Holonic Superposition Collaborative Multi-Agent Systems Architecture (Splinter). Splinter introduces a design pattern and an architecture to design and implement runtime pluggable, intelligent and holonic multi-agent systems that are programming language agnostic. A holonic multi-agent system comprises holonic agents, where a holonic agent recursively embeds holonic agents until a certain level of granularity exists. Holonic multi-agents prove robust, dynamic and efficient in complex domains.
Ph.D. (Computer Sciences and Software Engineering)