Multi-agent systems operating in complex environments, with diverse task requirements and environmental constraints, often lead to a significant increase in computational complexity. To address this challenge, this study proposes a hierarchical planning-based multi-agent path optimization strategy. First, the K-means clustering algorithm is employed to partition task points into multiple sub-regions, thereby reducing the complexity of task allocation. Subsequently, the Particle Swarm Optimization (PSO) algorithm and the constrained Grey Wolf Optimizer (PSO-CGWO) are employed for global path planning and local planning, respectively. By adopting a hierarchical approach to task allocation and path planning, this method significantly reduces computational complexity. This strategy provides an effective solution for implementing multi-agent systems in complex mission environments. Experimental results demonstrate that PSO-CGWO significantly outperforms PSO-GWO in computational efficiency, exhibiting shorter average execution times and smaller standard deviations across different task scenarios. Notably, PSO-CGWO maintains higher stability with an increasing number of tasks. These findings validate the advantages of PSO-CGWO in path planning within complex task environments, enhancing task execution efficiency and reliability.