Emergent Language in Multi-Agent Systems: A Multi-Task Learning Approach

Multi-agent systems, which consist of multiple intelligences capable of interacting with the environment through communication and interaction, have been widely used in many domains, such as intelligent robotics, transportation control, and distributed decision making. With the development of deep learning, there is an increasing focus on allowing intelligent bodies to interact through natural language. However, while traditional natural language processing methods rely heavily on labeled datasets, emergent languages can be formed and evolved spontaneously through interactions between intelligences without the need for pre-defined grammar rules. The main advantage of emergent languages is that they do not rely on manually labeled data, and thus can be applied more flexibly to a wide range of scenarios. This study focuses on emergent languages in multitasking environments, addressing the problem that emergent languages in existing studies are limited to single-tasking scenarios. By designing a multitasking intelligent body interaction simulation scenario combined with a reinforcement learning framework, this paper proposes a novel multi-intelligent body interaction model and algorithm based on emergent language. Experimental results show the effectiveness and adaptability of our proposed method in a multi-task environment.