Enhancing autonomous underwater vehicle reliability through functional decentralization and multi-agent interaction

The article presents a comparative analysis of existing control architectures – including centralized, two-tier, hybrid, and modular peer-to-peer – demonstrating their limitation in meeting modern requirements for scalability, energy efficiency, and survivability under conditions of environmental uncertainty and the ”information explosion”. To address these challenges, the authors propose an architecture based on the principles of a”non-strict hierarchy”, where each subsystem functions as a partially autonomous intelligent agent. The core of the system is a multi-agent intelligent model integrating communication and behavioral logic, domain ontologies, and advanced planning functions. The proposed structure is organized into functional clusters, including external communication, operational control, and executive devices, which facilitate the efficient distribution of computational loads across autonomous modules (agents). Furthermore, to resolve the critical issue of consensus in multi-agent systems, a method involving priority functions and tasks is introduced. The study concludes that this decentralized organizational approach minimizes the impact of hardware failures and communicational latencies, ensuring robust real-time decision-making in complex robotic environments