An Expert Data Generation Method for Multi-Agent Cooperative Planning Method

In environments with limited information, intelligent agents have restricted access to their current location, target location, and local environment details. To enhance their perception capabilities, this study introduces a fusion model called GraphSAGE-DH-Bug, which combines GraphSAGE and DH Bug algorithms. More importantly, the article presents the process of generating expert training databases and proposes three criteria for evaluating their effectiveness. These contributions establish the groundwork for applying the GraphSAGE-DH-Bug fusion model to multi-agent exploration in unknown environments. The experimental results validate the effectiveness of the generated dataset, which accurately represents the extended perception range of agents through the inclusion of neighboring agents.

I. Introduction

Multi-agent cooperative exploration is a rapidly growing field with applications in various domains, such as robotics, autonomous vehicles, environmental monitoring, and disaster response. In these scenarios, a team of intelligent agents must work together to efficiently explore and gather information in unknown and dynamic environments. Unlike traditional single-agent systems, multi-agent systems benefit from the ability to divide tasks, share information, and collaboratively plan their actions, leading to improved efficiency and adaptability.