N. A. Wagner, C. Wietfeld, "O-RACES: Proactive AI-driven scheduling in Open RAN for 6G-networked humanoid robots," in IEEE INFOCOM Workshops, NetRobiCS Workshop on Networked Robotics and Communication Systems, Tokyo, Japan, May 2026.

Real-time communications is an essential feature of future 6G networks, enabling time-critical applications including cooperative robotic tasks in industry environments. Here, humanoid robots pose particular challenges due to their high complexity. When operating closed-loop via networked control, they exhibit stringent requirements on the wireless communications system. While single robots can be controlled locally, challenging collaborative tasks require high-fidelity networked control via a central entity. Reinforcement Learning (RL)-trained and edge-cloud-computed neural network policies enable agile networked robotics that do not rely on energy-intensive computing hardware on the robot. However, conventional reactive scheduling in 5G networks introduces significant uplink latency, limiting time-critical robotic control. To overcome this limitation, we propose the Open RAN Real-time AI-Coordinated Efficient Scheduling (O-RACES) proactive scheduling framework operating on open interfaces and designed to meet the stringent latency requirements of networked humanoid robots. We evaluate the scalability of O-RACES in a hybrid experimental environment using 16 virtual humanoid robots integrated within an end-to-end experimental Open RAN 6G research testbed. Results show that O-RACES reduces latency by up to 50% compared to reactive scheduling and resource usage by 40% versus static scheduling, paving the way for scalable collaborative 6G-networked robotics.