Multi Robot Cooperative Control Based on Edge Computing and Multi Agent Reinforcement Learning with Cooperative Planning

In this era, multi-robot systems play vital role in the sector of environmental monitoring, disaster response, and smart manufacturing. Previous researchers have suggested various traditional methods but still there are challenges such as high latency, bandwidth limitations, and single points of failure, making them unsuitable for time-sensitive and large-scale operations. This article is proposed to develop a system for Multi robot Cooperative Control based on edge computing by employing robust methods such as; Hierarchical Inference, dynamic workload distribution, real-time collaboration on multi-agent. Further, the hierarchical Inference is classified into three layers; cloud layer which is designed using Open Artificial Intelligence (AI) DALL-E, fog layer employed for data preprocessing using various techniques such as median filtering for noise filtering and Linear Discriminant Analysis (LDA) for dimensionality reduction and edge layer is designed by incorporating MobileViT technology. Then, the extracted features are allocated with specific task using federated learning technique namely federated averaged (FedAvg). Finally the Multi Agent Reinforcement Learning (MARL) with Cooperative Planning (CP) is imposed for real-time collaboration in multi agent and has resulted with 0.99 Accumulative Coverage Score (ACS) when compared to the existing Large Language Model (LLM).