Synthetic data improves drone signal prediction with less real flight data
Measurement-Calibrated Synthetic Pretraining for Rapid Adaptation of UAV Channel Predictors
Machine Learning
Summary
Predicting wireless signals for drones is hard because the signals can change quickly and there isn’t much real data available. The paper shows that using synthetic data that mimics real drone signals can help train prediction models faster and with less real data. The authors also demonstrate a way to adjust these pretrained models quickly for new flights using only a small amount of data. This approach can make drone communication systems smarter and more adaptable in real-time.
What this means in practice
- •For drone communication engineers: Speed up training of wireless channel predictors for drones when little real flight data is available.
- •For wireless network developers: Reuse synthetic-data pretrained models to adapt channel predictors quickly for changing aerial network conditions.
Authors
Qiuheng Zhou, Wei Jiang, Donglin Wang, Hans D. Schotten
Abstract
Unmanned aerial vehicle (UAV) links can change faster than route-specific channel data can be collected, so a channel predictor must often be adapted from only a short target-flight calibration segment. This paper studies measurement-calibrated synthetic pretraining as an offline source prior for rapid adaptation of channel state information (CSI) predictors. The synthetic source is calibrated to flight-only power evolution, temporal shape variation, frequency selectivity, and phase disturbance, and is then combined with measured source-flight fine-tuning before target-flight adaptation. Under chronological held-out flight splits, synthetic-to-real (S2R) fine-tuning reduces the measured-data budget needed to reach useful prediction in the low-data regime. For 50 ms-ahead prediction, it crosses a latest-channel-estimate persistence baseline with 50 measured training windows, roughly 0.5 s of 10 ms-spaced supervised windows, whereas training from scratch requires 600 windows. A target-prefix adaptation study further shows that a lightweight gated recurrent unit predictor can reuse the pretrained source prior with one prediction-layer update epoch, positioning measurement-calibrated synthetic pretraining as a reusable prior for scarce target-flight measurements.