Transformer xApps for Semantic URLLC in Open RAN

This paper addresses reliability challenges in ultra-reliable low-latency communication (URLLC) by extending classical communication theory with semantic-aware approaches deployed as xApps in Open RANs Near-RT RIC (Near Real-Time RAN Intelligent Controller). We introduce two complementary frameworks. PULSE (Predictive Ultra-reliable Low-latency System Engine) leverages transformer-based semantic understanding to predict and reconstruct lost packets, reducing latency compared to retransmission-based recovery. DANTE (Dynamic Abstraction of Natural Text to Executable) ensures consistency and reliability at the command level by standardizing heterogeneous natural-language and vendor-specific instructions into consistent, executable formats. Together, PULSE and DANTE extend reliability from packet recovery to semantic command interpretation, centralizing intelligence at the RAN edge to reduce device overhead and enable consistent, low-latency communication. Experimental validation demonstrates that these frameworks outperform traditional approaches in robotic and drone scenarios, and the underlying methods generalize to other URLLC domains such as industrial automation, emergency response, and tactile internet applications.