Multi-agent Framework Based on Coordinated Sensing and Control with Online Evolutive Learning

The integration of Internet of Things (IoT) and Artificial Intelligence (AI) technologies has emerged as a pivotal research focus, fostering the development of diverse intelligent systems. These systems usually require autonomous decision-making and real-time feedback, and the multi-agent system (MAS) is one of the main paradigms for realizing such systems. Within MAS, collaborative perception and control are key components. Although existing research has made progress in these two aspects, most of them study the multi-agent perception and control algorithms separately in MAS, which may be insufficient when facing the non-stationary environment. In this paper, we propose a multi-agent coordinated sensing and control framework based on online evolutive learning. Specifically, we present a framework consisting of multiple sensors and controllers, where the sensors are based on target detection models, and the controllers are based on multi-agent reinforcement learning algorithms. Coordinated optimization from the controllers to the sensors, along with a generative model providing prior knowledge, enables agents to achieve online evolutive learning. To validate the effectiveness of the proposed framework, we construct a multi-agent pathfinding simulation environment, and the experiments demonstrate both the superiority and the necessity of our approach.