LEO satellite networking for direct-to-cell and satellite IoT: predictive handover, orbital scheduling, pass-aware delivery, jurisdictional policy, and satellite-edge learning. New: Predict, Don't React, ephemeris as lead time for the Direct-to-Cell control plane.
SkyHandover predicts direct to cell handovers from protocol invariants rather than signal strength alone. It uses cross layer features at the RAN edge to act before the satellite transition disrupts throughput.
SkyWindow holds IoT packets for better upcoming satellite windows instead of sending immediately into poor geometry. It uses orbital predictability to improve reliability without changing satellites or gateways.
Adaptive Orbital D2D chooses among LEO, GEO, and hybrid paths based on Doppler, load, and application needs. It makes orbital choice a runtime optimization rather than a static architecture decision.
D2D Decide determines when nearby phones should relay for a device with weak satellite signal. It avoids unnecessary discovery overhead while recovering links that would otherwise fail.
PathLink precomputes handover sequences for users moving through humanitarian corridors. It combines satellite trajectory prediction with mobility patterns to keep emergency connectivity stable.
PredictAct argues that direct to cell networks should use public orbital trajectories across radio, application, and policy planes. The same propagator can expose lead time at multiple layers.
Predictability Tax measures the energy, latency, and memory cost of handover predictors that report similar accuracy. It shows that predictor choice is under specified unless deployment cost is reported with F1.
SpaceRegulate pre stages legal policy before a satellite crosses jurisdiction boundaries. It turns predictable orbital movement into a policy handover primitive for space data planes.
FedSat LAM enables large AI models on resource constrained satellites using multihop offload, tree aggregation, and parameter efficient tuning. It keeps more learning near the data produced in orbit.
SafeZones protects mobility traces using sensitive location discovery, differential privacy, and synthetic trajectory generation. It balances trajectory privacy with service utility at the edge.