Advancements in artificial intelligence (AI) and improved performance of embedded systems have facilitated the widespread adoption of deep learning network model predictions in both cloud services and edge computing. These distributed intelligent nodes can be amalgamated into a cohesive service. As an application of Distributed Artificial Intelligence (DAI) [1], the Multi-Agent System (MAS) extends agent technology by employing a network of loosely coupled autonomous agents that interact within an environment to pursue a shared objective [2]. This study leverages the conceptual framework of MASs, drawing inspiration from physiology, to propose an integration framework for complex AI systems. A navigation system for the visually impaired serves as the practical application of this framework.
MAS technology, recognized for its parallel computation capabilities, robustness, scalability, affordability, and reusability, is particularly well-suited for large-scale systems [3]. Within the realm of agent organization architecture, a holonic agent structure with a recursive design is notably effective in modeling complex systems [4, 5]. Similarly, systematic design approaches begin with system specifications, followed by a hierarchical and modular construction of the system's functional architecture using the IDEF function modeling method, thereby facilitating the efficient design of intricate systems [6].
Physiology serves as a potent inspiration for designing complex systems [7]. Hierarchical organizations with recursive submodules are a common structural paradigm in biological networks, including neural, metabolic, ecological, and genetic regulatory networks [8]. These structures range from static configurations, where cells aggregate into tissues, tissues into organs, and cells contain smaller organelles, to dynamic systems like the circulatory and nervous systems, which respectively manage material delivery and information transmission throughout the body [9]. In this study, a framework was developed to integrate complex intelligent agents by emulating the organizational and coordination mechanisms of the human body.
Multi-Agent Systems are predominantly utilized for node coordination across various sectors [10, 11], including traffic control [12, 13], smart electric grids [14, 15], and team coordination [16–18]. While the application of MAS techniques to centralized architectures is less common in fields such as robotics or self-driving cars, this paper presents a case study where a centralized MAS architecture is applied to a navigation system. This exploration highlights the potential for broader applicability of MAS methodologies in centralized settings.
In addition to structural considerations, inter-organizational communication is a pivotal aspect of system design. Protocols such as the Message Queuing Telemetry Transport (MQTT) [19] and the Object Management Group’s (OMG) Data Distribution Service (DDS) [20] are widely used for publishing and subscribing to sensor or event data. DDS, in particular, offers numerous configurable policies that enhance the end-to-end Quality of Service (QoS) for applications. In this study, these communication features were leveraged to orchestrate the operations of complex intelligent systems.
This paper is organized into five sections. Section 1 introduces the context and relevance of the study. Section 2 presents a design that draws bionic inspiration from both physiology and MAS theory. Section 3 details the experimental validation of the proposed strategy using a navigation system for the visually impaired. Section 4 discusses the findings of these experiments. Section 5 concludes the paper with a summary of the results and an outline of future research directions.