This paper presents the IoT Smart Hub, a local, modular infrastructure designed to enhance adaptive interoperability, semantic context perception, and action validation in Internet of Robotic Things (IoRT) ecosystems. A new M2M2PA (Machine-to Machine to-Perception-and- Actuation) paradigm is also proposed, integrating semantic reasoning and real-time feedback into the communication cycle. Preliminary results demonstrate its capacity to structure contextual data and support cognitive agents, such as Large Language Models (LLMs), for natural language interaction and future reinforcement learning integration.