RemoteRF was developed in the Wireless Lab at UCLA to simplify experimental validation in wireless research. By separating compute from radio hardware, it turns shared SDRs and RF instruments into a multi-user, network-accessible testbed:
- Local compute, remote radios: Run signal processing, parameter sweeps, or AI/ML training on your workstation or GPU server while streaming IQ samples to and from remote SDRs in Python.
- Exclusive reservations: Book time slots on one or more devices so other users cannot overwrite carrier frequencies, gains, or buffers during an experiment.
- Switch setups in software: Move between cabled loopback, over-the-air, MIMO, multi-room, and multi-site deployments simply by changing reservation tokens.
Iterative Experimentation Workflow
Wireless research typically moves back and forth between offline algorithm design and over-the-air testing:
-
Stage 1
Offline Development & Simulation
Build and debug waveforms, signal processing blocks, or AI/ML pipelines locally in Python using simulated channels, saved IQ datasets, or virtual radios without using physical hardware.
-
Stage 2
Cabled & Loopback Calibration
Reserve a cabled loopback or cross-device cabled radio to verify your signal processing pipeline over a repeatable, interference-free link before transmitting over the air.
-
Stage 3
Over-the-Air Capture & Live Evaluation
Switch your script's device tokens to over-the-air, multi-SDR, or MIMO setups to stream live IQ samples across real channels. Run closed-loop experiments in real time or log captures to disk for offline analysis.
-
Stage 4
Offline Debugging, Retraining & Re-Evaluation
Analyze captured signals or model performance offline, refine your algorithms or retrain models on your computer or GPU server, and re-test on the same hardware setup.
Research Guides
Run cabled loopback, multi-SDR, and MIMO setups, control multiple radios from one script, and collect repeatable IQ datasets.
Stream IQ samples between remote SDRs and GPU servers for deep learning, site-specific wireless techniques, RF fingerprinting, and reinforcement learning.
Use AI coding agents and RemoteRF's MCP server to run closed-loop DSP debugging on live links, port experiments across SDR families, and automate parameter sweeps.
Control SDRs alongside custom RF instruments, build multi-room testbeds, and share hardware across your research group.
Using or Deploying RemoteRF
Reserve time on active RemoteRF nodes to run over-the-air experiments, capture IQ data, or test algorithms without setting up hardware.
Run remoterf-server on a Linux machine in your lab to give students and collaborators shared access to your group's SDRs and RF equipment.
Publications & Citing RemoteRF
If you use RemoteRF in your academic research, dataset collection, or experimental evaluation, please cite the following paper:
- Ethan Y. Ge and Ian P. Roberts, “RemoteRF: A Global, Open-Source Sandbox for Wireless Research and Education,” arXiv preprint arXiv:2606.22161. [arXiv:2606.22161 ↗]
@misc{remoterf_2026,
title={{RemoteRF}: A Global, Open-Source Sandbox for Wireless
Research and Education},
author={Ethan Y.~Ge and Ian P.~Roberts},
year={2026},
url={https://arxiv.org/abs/2606.22161},
}