Researchers at the University of Tehran, a partner of TV BRICS, have introduced a new distributed artificial intelligence framework that enables multiple autonomous agents to learn and make decisions independently while coordinating under shared operational constraints, marking an important advance in the development of next-generation intelligent systems.
The newly developed approach addresses one of the major challenges in multi-agent reinforcement learning – allowing autonomous agents to cooperate efficiently without relying on a central controller. Instead, each agent uses only locally available information and communication with neighbouring agents, making the system more scalable, resilient and suitable for real-world deployment, as reported by
Mehr News Agency, a TV BRICS partner.
The method combines a distributed optimisation framework with an actor-critic reinforcement learning architecture, enabling intelligent agents to continuously adapt their behaviour while respecting common system constraints. Researchers also demonstrated that the algorithm converges to stable solutions, providing strong theoretical support for its practical application.
Experts believe the breakthrough could accelerate the development of autonomous technologies across a wide range of sectors, including robotics, intelligent energy networks, urban traffic management, self-driving systems, distributed computing and computational economics.
By eliminating the need for centralised coordination, the approach could also improve the efficiency, reliability and scalability of future AI-powered infrastructure.