TITLE:
Research Report on Multi-Agent Collaboration in Shipping: Expert Agent Construction and Application
AUTHORS:
Chong Pan
KEYWORDS:
Shipping Industry, Large Language Model, Multi-Agent System (MAS), Shipping Professional Knowledge Base, Online Knowledge Graph, AI Agent
JOURNAL NAME:
Journal of Software Engineering and Applications,
Vol.18 No.9,
September
24,
2025
ABSTRACT: As a vital pillar of global trade, the shipping industry possesses vast and complex expertise, data, and scenarios. Compiling shipping research reports demands high professional skills from data scientists and industry analysts, proving both time-consuming and labor-intensive. With the rapid advancement of artificial intelligence technology, the integration of large language models, RAG systems, and Multi-Agent systems with specialized shipping scenarios has emerged as a key direction for assisting shipping research and enhancing data analysis efficiency. Traditional RAG single-agent systems integrated with large models typically follow a linear “retrieval-generation” workflow when generating research reports. This approach suffers from disconnects between retrieval and generation, neglects conceptual connections, amplifies content hallucinations, and lacks the ability to generate comprehensive narratives and insights. This paper addresses the limitations of RAG-based single-agent systems in handling complex tasks requiring deep reasoning, multi-step planning, and tool collaboration. Drawing on multi-agent system principles, it designs a hierarchical collaborative expert agent architecture comprising management and execution layers. Detailed functional specifications for each module are provided. It proposes a four-stage framework for generating shipping report content and conducts a case study analyzing the impact of the Red Sea crisis on the global shipping landscape. Results demonstrate that this expert agent achieves significant improvements in completeness, accuracy, and research depth compared to single-agent RAG systems when generating reports. The findings demonstrate that the constructed shipping report expert agent substantially enhances the completeness of shipping content retrieval and the quality of report generation, elevating the intelligence level of shipping information analysis. This holds significant practical implications for advancing the intelligent transformation of the shipping industry.