Objective Existing unmanned surface vehicle (USV) path planning algorithms rely heavily on rule-based libraries and high-precision sensor inputs, making them unsuitable for “information-scarce” environments characterized by cognitive blind spots, perception loss, and semantic ambiguity. Such conditions hinder their ability to respond to sudden, unknown threats. This study proposes an intelligent agent framework leveraging large language models (LLMs) to explore its feasibility and advantages for USV path planning under such information-scarce conditions.

Method The proposed intelligent agent framework leverages a pre-trained LLM as its cognitive core, and is designed for high modularity and extensibility. The framework integrates a diverse library of tool functions (including sensor APIs for real-world data acquisition) and a human-computer interaction interface for natural language engagement, significantly enhancing the LLM's perceptual and operational capabilities. To facilitate continuous learning and experience accumulation, the framework incorporates long-term and short-term memory mechanisms alongside distinct online and offline evolutionary strategies, enabling the agent to autonomously self-optimize through persistent interactions and accrued experience. At the core of the framework is a highly structured prompting template, which systematically organizes critical contextual information such as the agent's persona, environmental context, current states, sensor observations with confidence scores, task objectives, chain-of-thought (CoT) and ReAct prompts, prior knowledge, decision-making priorities, and conditional logic rules. This structured prompting effectively guides and constrains the LLM's cognitive processes, enabling it to perform complex tasks including environmental perception, dynamic reasoning, interactive communication, and goal-directed path planning.

Results Experimental results indicate that, in environments with static obstacles, the proposed intelligent agent outperforms or performs at least comparably to representative traditional algorithms such as A*, RRT, PSO, and Hybrid A* across key metrics including path length, number of turns, and average turning angle. The agent achieved a composite performance score of 1.00, significantly higher than those of the traditional algorithms (0.47, 0.47, 0.77, and 0.76, respectively). In dynamic environments, the LLM effectively invokes tool functions, initiates human-in-the-loop interactions when appropriate, and performs real-time path planning based on sensor inputs. Under challenging conditions such as cognitive blind spots, perceptual gaps, and semantic ambiguity, the agent adapts its planning strategy by leveraging confidence assessments, historical memory, and commonsense knowledge, demonstrating robust reasoning, generalization capabilities, and human-like intelligence.

Conclusion These findings indicate that the LLM-based path planning agent exhibits significant advantages in both simple obstacle avoidance and complex, information-scarce scenarios. This approach holds strong potential for advancing USVs toward higher levels of autonomy.