Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

VTS-LLM: Domain-Adaptive LLM Agent for Enhancing Awareness in Vessel Traffic Services through Natural Language

Sijin Sun, Liangbin Zhao, Ming Deng, Xiuju Fu

TL;DR

This work tackles the challenge of enabling intuitive, domain-aware natural language interaction in Vessel Traffic Services (VTS) by introducing VTS-LLM, a domain-adaptive LLM agent for interactive decision support. It reframes risk-prone vessel identification as a knowledge-augmented Text-to-SQL task and builds a domain-specific multimodal dataset to support robust NL-to-SQL reasoning in maritime contexts. The methodology integrates NER-based relational reasoning, an agent-driven domain knowledge injection framework, a Semantic Algebra Intermediate Representation (SAIR), and a Query Rethink mechanism, achieving state-of-the-art performance (e.g., 77.80% overall accuracy) and revealing systematic effects of linguistic style on Text-to-SQL. The work demonstrates the feasibility and value of natural language interfaces for proactive, LLM-driven real-time maritime traffic management and outlines future directions for broader applicability and robustness under diverse regulatory contexts.

Abstract

Vessel Traffic Services (VTS) are essential for maritime safety and regulatory compliance through real-time traffic management. However, with increasing traffic complexity and the prevalence of heterogeneous, multimodal data, existing VTS systems face limitations in spatiotemporal reasoning and intuitive human interaction. In this work, we propose VTS-LLM Agent, the first domain-adaptive large LLM agent tailored for interactive decision support in VTS operations. We formalize risk-prone vessel identification as a knowledge-augmented Text-to-SQL task, combining structured vessel databases with external maritime knowledge. To support this, we construct a curated benchmark dataset consisting of a custom schema, domain-specific corpus, and a query-SQL test set in multiple linguistic styles. Our framework incorporates NER-based relational reasoning, agent-based domain knowledge injection, semantic algebra intermediate representation, and query rethink mechanisms to enhance domain grounding and context-aware understanding. Experimental results show that VTS-LLM outperforms both general-purpose and SQL-focused baselines under command-style, operational-style, and formal natural language queries, respectively. Moreover, our analysis provides the first empirical evidence that linguistic style variation introduces systematic performance challenges in Text-to-SQL modeling. This work lays the foundation for natural language interfaces in vessel traffic services and opens new opportunities for proactive, LLM-driven maritime real-time traffic management.

VTS-LLM: Domain-Adaptive LLM Agent for Enhancing Awareness in Vessel Traffic Services through Natural Language

TL;DR

This work tackles the challenge of enabling intuitive, domain-aware natural language interaction in Vessel Traffic Services (VTS) by introducing VTS-LLM, a domain-adaptive LLM agent for interactive decision support. It reframes risk-prone vessel identification as a knowledge-augmented Text-to-SQL task and builds a domain-specific multimodal dataset to support robust NL-to-SQL reasoning in maritime contexts. The methodology integrates NER-based relational reasoning, an agent-driven domain knowledge injection framework, a Semantic Algebra Intermediate Representation (SAIR), and a Query Rethink mechanism, achieving state-of-the-art performance (e.g., 77.80% overall accuracy) and revealing systematic effects of linguistic style on Text-to-SQL. The work demonstrates the feasibility and value of natural language interfaces for proactive, LLM-driven real-time maritime traffic management and outlines future directions for broader applicability and robustness under diverse regulatory contexts.

Abstract

Vessel Traffic Services (VTS) are essential for maritime safety and regulatory compliance through real-time traffic management. However, with increasing traffic complexity and the prevalence of heterogeneous, multimodal data, existing VTS systems face limitations in spatiotemporal reasoning and intuitive human interaction. In this work, we propose VTS-LLM Agent, the first domain-adaptive large LLM agent tailored for interactive decision support in VTS operations. We formalize risk-prone vessel identification as a knowledge-augmented Text-to-SQL task, combining structured vessel databases with external maritime knowledge. To support this, we construct a curated benchmark dataset consisting of a custom schema, domain-specific corpus, and a query-SQL test set in multiple linguistic styles. Our framework incorporates NER-based relational reasoning, agent-based domain knowledge injection, semantic algebra intermediate representation, and query rethink mechanisms to enhance domain grounding and context-aware understanding. Experimental results show that VTS-LLM outperforms both general-purpose and SQL-focused baselines under command-style, operational-style, and formal natural language queries, respectively. Moreover, our analysis provides the first empirical evidence that linguistic style variation introduces systematic performance challenges in Text-to-SQL modeling. This work lays the foundation for natural language interfaces in vessel traffic services and opens new opportunities for proactive, LLM-driven maritime real-time traffic management.
Paper Structure (24 sections, 3 equations, 5 figures, 7 tables)

This paper contains 24 sections, 3 equations, 5 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Vessel-of-Interest Identification in VTS via Text-to-SQL
  3. Identification of Vessels of Interest
  4. Knowledge-augmented Text-to-SQL with Operational Linguistic Queries
  5. Dataset
  6. Methodology
  7. NER-based Relational Reasoning
  8. Agent-based Domain Knowledge Injection
  9. Domain Knowledge-Aware Semantic Parsing
  10. Natural Language Processing and Interactive Query Engine
  11. Multimodal DBs and Functional Call
  12. Semantic Algebra Intermediate Representation
  13. Semantic-aware Entity Understanding and Domain Enhancement
  14. Structured Algebraic Transformation Process
  15. Fusion of spatial and functional instructions
  16. ...and 9 more sections

Figures (5)

  • Figure 1: Simplified overall structure of the VTS-LLM agent.
  • Figure 2: Hierarchical NER and Reasoning Framework for Maritime Query Processing
  • Figure 3: Agent-based Maritime Knowledge Injection Framework.
  • Figure 4: Schematic workflow of the SAIR module for VTS scenario.
  • Figure 5: Demos of our developed system with VTS-LLM agent in VTS scenario.