FlexRDZ: Autonomous Mobility Management for Radio Dynamic Zones
Aashish Gottipati, Jacobus Van der Merwe
TL;DR
FlexRDZ introduces a closed-loop autonomous RDZ manager that combines Hierarchical Task Networks with a digital twin to maintain RDZ health under mobile transmissions. The system uses a TIREM-based propagation model, accelerated by a neural RF-map approximator, and a GTPyhop-based HTN planner to generate interpretable action sequences in near real-time. In simulation on the Salt Lake City POWDER urban RDZ, FlexRDZ reduces mobile interference by up to $20$ dBm and lowers total leaked signal power while preserving test-transmitter uptime, outperforming PPO and naive baselines in leakage and interference metrics. The work demonstrates the potential of combining AI planning, digital twins, and physics-based modeling for autonomous, transparent management of dynamic spectrum environments, with future work focusing on generalization and real-world benchmarking.
Abstract
FlexRDZ is an online, autonomous manager for radio dynamic zones (RDZ) that seeks to enable the safe operation of RDZs through real-time control of deployed test transmitters. FlexRDZ leverages Hierarchical Task Networks and digital twin modeling to plan and resolve RDZ violations in near real-time. We prototype FlexRDZ with GTPyhop and the Terrain Integrated Rough Earth Model (TIREM). We deploy and evaluate FlexRDZ within a simulated version of the Salt Lake City POWDER testbed, a potential urban RDZ environment. Our simulations show that FlexRDZ enables up to a 20 dBm reduction in mobile interference and a significant reduction in the total power of leaked transmissions while preserving the overall communication capabilities and uptime of test transmitters. To our knowledge, FlexRDZ is the first autonomous system for RDZ management.
