Optimizing Multi-Agent Coverage Control with Deep Learning and Voronoi

This paper presents a hybrid approach combining multi-agent coverage control with deep learning to optimize agent coordination in dynamic environments. By partitioning the environment using Voronoi cells, agents iteratively adjust their positions toward centroids to maximize coverage while minimizing overlap. A neural network, specifically a multilayer perceptron, predicts these centroids online, eliminating the need for traditional, computationally expensive Voronoi partitioning. The model enhances efficiency and accuracy, making it well-suited for online applications such as autonomous systems. The results demonstrate significant computational benefits and improved agent performance in achieving optimal coverage.

کلیدواژه ها:

Multi-agent systems ، Coverage control ، Deep learning ، Voronoi partitioning ، Optimization