FelixOtt/sicdf_dataset

S-ICDF: Sionna-Simulated Interference Characterization and Direction Finding Dataset

Latest update: 2026-05-05

Project size: 1.3 TB

Introduction

We publish the S-ICDF dataset (Sionna-Simulated Dynamic Interference Characterization and Direction Finding) that was simulated in the ray-tracing environment Sionna. The environment is an 3D model of the industrial environment recorded at the Fraunhofer IIS L.I.N.K. test and application center. The dataset contains a variety of interference classes in GNSS signals.

The S-ICDF Dataset

S-ICDF is a large-scale indoor interference dataset for RF interference monitoring, generated with NVIDIA Sionna and its ray-tracing engine. The dataset is designed to support research on interference detection, classification, characterization, and direction finding (DF) under realistic indoor multipath conditions. It provides simulated raw complex baseband IQ data together with controlled variations in signal, channel, and antenna parameters, enabling reproducible benchmarking of both classical signal-processing methods and modern machine learning models.

Motivation

Jamming and spoofing threaten wireless and satellite-navigation systems by disrupting or manipulating radio-frequency (RF) signals, thereby degrading availability, integrity, and trust. Robust interference monitoring is therefore essential to identify anomalous signals and estimate their source direction. Although machine learning has shown strong potential for these tasks, progress is limited by the lack of large-scale datasets with realistic signal and channel variability.

Collecting such data in the real world is difficult. Intentional jamming is often legally restricted, and real measurements are typically affected by uncontrolled propagation, hardware impairments, and incomplete ground truth. S-ICDF addresses this gap by providing a fully controlled, simulation-based benchmark with realistic indoor multipath, configurable antenna arrays, and diverse interference waveforms.

Highlights

• Large-scale simulated interference dataset for indoor RF monitoring
• Generated with Sionna using physically grounded ray tracing
• Supports classification, characterization, and direction finding
• Includes moving antenna arrays and realistic time-varying multipath
• Covers 110 interference configurations
• Built from 102 real-world MXG-generated signal files across six interference classes
• Provides benchmark baselines with MUSIC, ESPRIT, SAGE, and ML models

Dataset Summary

Property	Value
Simulation engine	NVIDIA Sionna / Sionna RT
Environment	Indoor industrial hall-like scene
Center frequency	1.57542 GHz
Simulated bandwidth	100 MHz
IQ length per snapshot	1,024 complex samples
Number of simultaneous sources	1
Receiver motion	Yes
Array layouts	2×2 planar, 1×8 linear
Train/test split	80/20
Interference configurations	110
Signal library	102 signal files
Main interference classes	Chirp, FrequencyHopper, Noise, Multitone, Modulated, Pulsed

Simulation Environment

The dataset is generated in an industrial hall-like indoor environment designed to capture dense multipath propagation. The scene contains predominantly metallic walls, ceiling, and floor surfaces, as well as interior structures such as shelving units that act as reflectors and scatterers. Signal propagation is simulated with Sionna’s ray-tracing engine, which computes physically grounded paths between transmitter and receiver.

A single interference source is placed at a fixed position within the scene and transmits at the GPS L1 center frequency (1.57542 GHz). The receiver is modeled as a uniform antenna array that moves along a predefined trajectory through the hall while maintaining a fixed orientation relative to the source. This setup creates realistic time-varying observations that are particularly useful for direction finding under motion and synthetic-aperture effects.

Transmitter and Receiver Setup

• Source: one interference transmitter at a fixed location
• Center frequency: 1.57542 GHz
• Receiver arrays:
  • 2×2 planar array
  • 1×8 linear array
• Element spacing: configurable
• Antenna gain patterns: configurable
• Motion: receiver traverses the environment along a trajectory
• Signal representation: raw complex IQ samples
• Snapshot size: 1,024 samples per antenna element

Parameter Variation Strategy

S-ICDF is generated using a one-at-a-time (OAT) parameter variation strategy. In each experiment, exactly one parameter is varied over its predefined range, while all remaining parameters are fixed at default values. This design enables clear attribution of performance changes to the parameter under study and supports systematic sensitivity analyses.

For each parameter variation, a complete positioning dataset is generated with 216,089 samples, which is then divided into an 80/20 train/test split.

Variable Parameters

Simulation and Antenna Parameters

Parameter	Range	Default
Source bandwidth	1 MHz to 20 MHz	20 MHz
SNR	-20 dB to 20 dB	20 dB
Antenna distance	0.05 m, 0.09 m, 0.095 m, 0.2 m	0.09 m
Reflection depth	2, 5, 7	5
Refraction	True / False	True
Array layout	2×2, 1×8	2×2
Antenna gain pattern	dipole, isotropic, half-wave dipole, TR 38.901	dipole

At 1.57542 GHz, the antenna spacing range corresponds approximately to 0.26λ to 1.1λ.

Interference Waveforms

The dataset uses 102 real-world MXG-generated signal files spanning six interference classes:

• Chirp
• FrequencyHopper
• Noise
• Multitone
• Modulated
• Pulsed

These classes cover a broad range of spectral and temporal interference behaviors.

Waveform Details

Chirp
Two sweep profiles are included: linear and parabolic. Each is available with fast (0.025 ms), medium (0.25 ms), and slow (2.5 ms) sweep times. Bandwidth is varied from 2 MHz to 20 MHz.

FrequencyHopper
Carrier frequency hops across discrete channels. Main varying parameters are bandwidth and dwell time.

Noise
Spectrally flat emissions with varying occupied bandwidth. The dataset also includes BOC-modulated and AltBOC-like noise variants to introduce structured spectral shapes.

Multitone
Multiple equally spaced sinusoidal tones are transmitted simultaneously. The number of tones and the total occupied bandwidth are varied.

Modulated
Parameters such as modulation scheme, symbol rate, and occupied bandwidth are varied to generate structured temporal and spectral signal patterns.

Pulsed
Signals alternate between transmission and silence with a fixed 1:1 duty ratio. Pulse repetition interval and bandwidth are varied.

Intended Tasks

S-ICDF is intended for the evaluation of methods for:

• Interference classification
• Interference characterization
• Direction finding (DF) / direction of arrival (DoA) estimation
• Sensitivity analysis with respect to signal, channel, and antenna parameters
• Benchmarking classical and ML-based methods

Why S-ICDF?

Existing public GNSS/RF interference datasets are valuable benchmarks, but many lack one or more of the following properties:

• synchronized multi-antenna measurements,
• realistic indoor multipath,
• moving antenna trajectories,
• broad and controlled parameter variation,
• reliable ground truth for direction finding.

S-ICDF is designed to fill this gap by combining array-based IQ measurements, ray-traced multipath, receiver motion, and systematic parameter control in a single benchmark dataset.

Repository Contents

This repository provides the dataset release and supporting material for benchmarking. Depending on the release version, contents may include:

• raw IQ recordings,
• metadata and labels,
• train/test splits,
• example visualizations,
• baseline evaluation code,
• scripts for loading and processing the dataset.

References

For more information of the dataset and results, see our publication. If you use our dataset for your research, please consider citing:

@inproceedings{s_icdf_dataset,
  author = {Christian Wielenberg and Lucas Heublein and Jonathan Ott and Alexander Mattick and Nisha L. Raichur and Jonas Pirkl and Lukas Schelenz and Tobias Feigl and George Yammine and Christopher Mutschler and Felix Ott},
  title = {{The S-ICDF Dataset: Sionna-Simulated Dynamic Interference Characterization and Direction Finding}},
  booktitle = {to be published soon},
  month = ?,
  year = 2026
}

Acknowledgment

This work has been carried out within the PaiL project, funding code 50NP2506, sponsored by the German Federal Ministry for Transport (BMV) and supported by the German Space Agency at DLR, the Bundesnetzagentur (BNetzA), and the Federal Agency for Cartography and Geodesy (BKG).

License

This work is licensed under a CC BY-NC-SA 4.0: Creative Commons Attribution-Noncommercial-ShareAlike, see https://creativecommons.org/licenses/by-nc-sa/4.0/

Kontakt

If you have any questions or tips to improve the datasets, contact us:

Christian Wielenberg: christian.wielenberg@iis.fraunhofer.de

Felix Ott: felix.ott@iis.fraunhofer.de

Nordostpark 84, 90411 Nürnberg, Germany, GoogleMaps

Fraunhofer Institute for Integrated Circuits IIS, Nürnberg, Germany