A Scalable Agent Architecture

Research in the field of Multi Agent Systems (MASs) focuses on the design of intelligent interactions between agents. Consequently, agents often require multiple layers of function-ality, subcomponents, and communication channels leading to more complex agents. The development process of such agents is also intricate and often results in unclear implementations, software errors, and security vulnerabilities. To mitigate this complexity and streamline the development process, the paper proposes a scalable and modular agent architecture that functions as a system in a system. The key concept here is to model an agent as a MAS of its own. Agents are divided into components such as resources, sensors, and conversations, which communicate and cooperate with each other like agents in a MAS. Applying this concept enhances the manageability of agent functionalities by adhering the principle of divide and conquer while remaining within the concept of MASs. The proposed approach enables the implementation of each agent as a system of sub-agents, making them deployable in a cloud based environment and leading to highly scalable systems.