From Passive Mirrors to Active Agents: Holonic Digital Twins for Physical Artificial Intelligence Over Networks

Artificial intelligence (AI) has revolutionized multiple sectors, ranging from health care to entertainment. Remarkably, despite this progress, today’s AI tools, including deep learning and generative AI (e.g., large language models), still fail when embedded into physical systems, such as robots or vehicles, that operate under the physical laws of the real world. This limitation stems from the inability of physical AI systems to maintain reliable world models for long-horizon planning under uncertainty and to generalize to unseen scenarios. In this context, wireless networks, through their pervasive sensing and communication infrastructure, can act as orchestrators of physical intelligence. However, current wireless architectures optimize for throughput, latency, and reliability and cannot support real-time physical AI coordination where agents must maintain a shared spatiotemporal context. To overcome such challenges, in this article, a network of holonic digital twins (HDT-Nets) is proposed to deliver the real-time physical AI inference requirements through holonic agents that actively reason about their environment, rather than passively mirror physical assets. Each HDT is realized as a hierarchical structure spanning the physical agent and the network edge, reasoning autonomously at the local level while cooperating with neighboring HDTs to form larger collectively intelligent units at the network level. In HDT-Nets, causal Markov blankets (MBs) spanning sensing, communication, and control domains determine which agents must coordinate and enable counterfactual reasoning over multidomain interventions. Active inference, operating within these boundaries, unifies perception, action, and learning by minimizing expected free energy, simultaneously deciding which beliefs to transmit based on their cognitive value to the receiver. Category theory guarantees that these transmitted beliefs preserve semantic structure when crossing between heterogeneous agents with incompatible representations. Finally, integrated information theory (IIT) provides a quantifiable metric to capture when collective intelligence exceeds independent operation and how network intelligence evolves through coordinated learning and information exchange.