Leveraging Large Language Models for Dynamic Multi-Objective Optimization in UAV Sensor-Target Assignment

The multi-sensor target assignment problem in Unmanned Aerial Vehicles (UAVs) plays a vital role in optimizing a range of tasks, including navigation, target tracking, and obstacle avoidance. It focuses on the efficient allocation of sensors to specific targets, improving task performance, accuracy, and operational efficiency. Traditional multi-objective optimization algorithms often rely on static operators, which limits their adaptability to dynamic environments. To address this challenge, we introduce LLM-MOEA, a novel Large Language Model (LLM)-driven multi-objective evolutionary algorithm. LLM-MOEA dynamically adjusts optimization operators, leveraging population-level insights through an innovative prompt design and integrates the optimization trajectory to enhance the algorithm's decision-making capabilities. Experimental results demonstrate that LLM-MOEA significantly outperforms traditional algorithms in terms of solution quality, convergence speed, and robustness. Ablation studies further confirm the importance of key components. This work highlights the potential of combining LLMs with evolutionary algorithms to solve complex multi-objective optimization problems and enhance the flexibility of optimization methods.