Source-linked AI summary
LIVE-RIS: A Comprehensive In-Flight Dataset for UAV-Mounted RIS Channel Measurements
David Müller, Kevin Weinberger, Aydin Sezgin, Martin Mönnigmann
TL;DR
UAV-mounted RIS studies have largely relied on simulations, leaving real-flight disturbances underrepresented. This paper provides a dataset from multiple flight experiments with synchronized pose, configuration, and S21 measurements. Initial analyses show that deployment location, configuration approach, and small disturbances affect measured link performance.
Problem
Most UAV-mounted RIS studies rely on numerical simulations under idealized assumptions, although flight disturbances can degrade channel quality.
Source: §I · p. 1Method
The paper provides a multi-flight dataset containing synchronized UAV and RIS pose estimates, ground-truth measurements, RIS configurations, and S21 channel measurements.
Source: §I · p. 1, §VII · p. 6Results
Initial analyses show that RIS configuration approach and deployment location influence achievable link performance, while small disturbances can substantially affect measured S21 magnitudes.
Source: §VI · pp. 5–6, §VII · p. 6Takeaways & Limitations
The dataset provides a foundation for future UAV-mounted RIS research and contributes toward bridging theoretical investigations and practical deployments.
Source: §VII · p. 6
Abstract
from arXiv · showhide
The dynamic control of the wireless propagation environment enabled by reconfigurable intelligent surfaces (RISs) makes them a promising technology for sixth-generation (6G) wireless networks. Beyond conventional fixed-position deployments, integrating a RIS with an unmanned aerial vehicle (UAV) provides additional spatial flexibility, facilitating favorable Line-of-Sight (LoS) conditions and broadening the range of potential applications. Although UAV-mounted RIS systems have attracted significant research interest, existing studies rely predominantly on numerical simulations and often overlook the impact of practical constraints and real-world disturbances. To address this gap, the authors recently presented the first UAV-mounted RIS prototype and experimentally validated its feasibility through real-world experiments. By using the UAV's extended Kalman Filter (EKF) for real-time RIS reconfiguration, the proposed system effectively mitigates disturbance effects and preserves the performance gains of the RIS-enabled link. To further advance UAV-mounted RIS research and facilitate broader access to real-world experimental insights, this work provides a comprehensive dataset collected from multiple flight experiments. The dataset includes EKF state estimates, ground-truth measurements, corresponding RIS configurations, and S21 channel data.
I. INTRODUCTION
UAV-mounted RISs offer flexible deployment but face real-flight disturbances that simulations often neglect. This work provides a multi-flight dataset to support study of their operation.
I. INTRODUCTION
- RISs dynamically manipulate impinging wavefronts and can support network objectives such as resilience and new applications.Their configuration and deployment position matter because the Tx–RIS–Rx link experiences multiplicative path loss.Source: §I · p. 1
- Existing UAV-mounted RIS studies largely rely on idealized numerical simulations, despite flight disturbances that can degrade channel quality.The authors’ earlier prototype work [7] experimentally validated a real-time-capable UAV-mounted RIS.Source: §I · p. 1
- The dataset combines EKF state estimates, ground-truth measurements, optimized RIS configurations, and corresponding VNA-acquired S21 channel measurements across multiple deployment geometries.It includes multiple Tx, Rx, and UAV-mounted RIS locations for analyzing their relative geometries.Source: §I · p. 1
A. RIS Subsystem
- The UAV-mounted RIS prototype uses 120 elements in a 10 × 12 array spanning 20 × 16 cm, with binary 1-bit phase shifting.The designed phase shift is 180° at 5.385 GHz; phase shifting incurs 3 dB attenuation, and the maximum reconfiguration rate is 50 Hz.Source: A. RIS Subsystem · pp. 1–2
B. UAV Subsystem
A customized Holybro X500 carries the RIS, with indoor position and other navigation measurements supplied by a motion-capture system. A 50 Hz link sends recent UAV EKF state to the RIS controller.
B. UAV Subsystem
- The customized Holybro X500 uses three IMUs with different frequency responses, whose measurements are fused with other sensors by the EKF.For indoor autonomous flight, MCS position measurements replace GNSS; the MCS also replaces onboard barometer and magnetometer measurements.Source: B. UAV Subsystem · p. 2
- The RIS center is mounted 265 mm below the UAV body-frame origin, with no horizontal offset.Fig. 1 marks the UAV-frame origin and RIS center on the mounted prototype.Source: B. UAV Subsystem · p. 2, Figure 1 · p. 2
C. Communication between Subsystems
- The flight controller sends data to the RIS controller over a 500000 Bd serial interface.The following passage specifies MAVLINK 2 streaming at 50 Hz, matching RIS sampling, with only the most recent UAV EKF data transmitted.Source: C. Communication between Subsystems · p. 2
- Transmitted EKF state includes UAV position and roll, pitch, and yaw at each time step.Only the most recent estimate is sent to reduce latency and maintain consistent data transmission.Source: C. Communication between Subsystems · p. 2
III. RIS ACTUATION
RIS actuation first derives the RIS pose from the UAV pose, then uses that pose for the subsequent configuration optimization.
III. RIS ACTUATION
- RIS pose estimation is a prerequisite for the subsequent configuration optimization.The section introduces both the pose calculation and the optimization approach.Source: §III · p. 2, A. RIS Pose · p. 2
A. RIS Pose
- Because the UAV and RIS are rigidly connected, their attitudes are identical.The RIS pose is determined before the subsequent optimization.Source: A. RIS Pose · p. 2
- The RIS position is obtained by rotating its fixed offset from the UAV and adding that offset to the UAV position.The offset is represented as l_RIS, and the rotation is the product of the yaw, pitch, and roll rotation matrices.Source: A. RIS Pose · p. 2
B. System and Channel Model
The channel model assumes a single-antenna Tx and Rx with no direct link, so communication proceeds through the UAV-mounted RIS. It models the effective link from per-element cascaded channels and RIS-controlled reflections.
B. System and Channel Model
- The model assumes a single-antenna transmitter and receiver, with the direct Tx–Rx link blocked and the RIS establishing the link.Source: B. System and Channel Model · p. 2
- Each RIS element applies a tunable phase and phase-dependent amplitude to its reflected signal.The reflection coefficient is defined as θ_m = A_m(φ_m)e^jφ_m.Source: B. System and Channel Model · p. 3
- The cascaded Tx–RIS–Rx channel for each reflecting element is modeled from LoS geometry and the Tx-to-element and element-to-Rx distances.The model specifies the speed of light, RIS-designed carrier frequency, and corresponding wavelength.Source: B. System and Channel Model · p. 3
C. RIS Optimization
RIS optimization derives the Tx–RIS and RIS–Rx distances at each time step, then selects a configuration to maximize the effective-channel magnitude under hardware constraints. It searches a finite set of target phases and evaluates the resulting candidate configurations.
C. RIS Optimization
- The Tx–RIS and RIS–Rx distances are determined for each reflecting element and time step before configuration optimization.The optimal candidate phase can be found independently for each patch by comparing eight choices and rounding.Source: C. RIS Optimization · p. 3
- The blocked Tx–Rx path permits arbitrary adjustment of the effective-channel phase, with a finite set of target phases used to limit computational complexity.The paper uses eight phase choices as a trade-off for finding a good solution within the available time.Source: C. RIS Optimization · p. 3
- The optimization selects the best candidate configuration by comparing the channel quality of configurations built from per-element 0° or 180° phase states.The procedure uses fast rounding to meet real-time requirements.Source: C. RIS Optimization · p. 3
IV. EXPERIMENTAL SETUP
The experiment section describes data acquisition, processing, and synchronization, with all attitude and position information expressed in the NWU coordinate frame.
IV. EXPERIMENTAL SETUP
- All attitude and position data use the right-handed NWU frame, with x north, y west, and z upward.The section also covers experiment design, data acquisition, processing, and synchronization.Source: §IV · p. 3
A. Experiments
Experiments use controlled hover flights in a flight lab, varying deployment geometry and RIS configuration method. A synchronization procedure excludes designated intervals from the measurement dataset.
A. Experiments
- Each 40-second flight uses 5 seconds each for takeoff and landing, and the UAV is restarted and its IMUs calibrated before every flight.Flights use hover control and the same environmental conditions.Source: A. Experiments · p. 3
- The experiments vary Tx, Rx, and UAV-mounted RIS locations across three scenarios and compare dynamic, static, and passive reflection methods.The scenarios cover optimal and non-ideal relative geometries.Source: A. Experiments · p. 3
- Of the remaining 30 seconds, the first and last 10-second intervals alternate between optimal RIS configuration and off state for synchronization and are excluded from the dataset.The central 10 seconds collect measurements using the respective RIS configuration.Source: A. Experiments · p. 4
1) UAV Flight Geometries:
The experiments cover three UAV-mounted RIS geometries and compare adaptive, fixed, and deactivated configurations. Together, these conditions vary placement relative to the direct link and whether RIS settings respond to UAV pose changes.
1) UAV Flight Geometries:
- The three geometries place the RIS at the Tx–Rx midpoint, directly above the Rx, or off the direct path.The midpoint placement follows the specular reflection path; the off-axis placement tests geometry that does not.Source: 1) UAV Flight Geometries · p. 4
2) RIS Configuration Methods:
- The adaptive method updates the optimized RIS configuration at 50 Hz using the latest UAV EKF estimate, while the fixed method keeps a preflight configuration unchanged.A third method deactivates the RIS, providing a passive metallic-reflector baseline.Source: 2) RIS Configuration Methods · p. 4
B. Data Collection
The flight campaign records UAV and RIS state, reference measurements, RIS configurations, and S21 channel data, then synchronizes these streams for comparison. Measurement and reference systems are aligned at 50 Hz.
B. Data Collection
- EKF state estimates, RIS pose, configurations, and timestamps are logged at 50 Hz, matching the RIS update rate.The scenario (iii) setup figure also indicates where the motion-capture cameras are placed.Source: B. Data Collection · p. 4, Figure 3 · p. 4
- A motion-capture system provides UAV position and orientation references at 100 Hz, with reported tracking error below 0.2 mm.The system uses 10 Vicon Vantage V5 cameras and Tracker 3 software.Source: B. Data Collection · p. 4
- A time-gated VNA measures RIS-reflected S21 across 201 frequency points from 5.225 to 5.875 GHz, with measurements at 50 Hz.The 61–71 ns time gate suppresses undesired multipath components.Source: B. Data Collection · p. 4
C. Processing of data
- MCS and EKF measurements are transformed to NWU and aligned by interpolating MCS data to EKF timestamps; VNA data are matched to RIS configurations using logged switching phases.The resulting UAV and RIS pose data and channel measurements are temporally aligned for comparison.Source: C. Processing of data · pp. 4–5
V. DATASET FORMAT
The dataset provides synchronized, timestamped flight records and a binary RIS-configuration representation, alongside validation measurements showing that configuration, deployment location, and flight disturbances affect measured link performance.
V. DATASET FORMAT
- Each 10 s flight file contains 500 timestamped samples at 50 Hz in CSV format, organized by scenario, RIS configuration, and flight.Files include UAV and RIS pose estimates alongside ground-truth measurements; headers specify units.Source: §V · p. 5
- The RIS configuration is stored as a binary row vector: 0 means inactive and 1 applies a 180° phase shift; entries map column-wise from the top-left element.The indexing figure shows front and UAV top-down views of the downward-facing RIS, with its longer side aligned to the x-axis.Source: §V · p. 5, Figure 4 · p. 5
VI. DATA VALIDATION
- In scenario (i), static and off use identical RIS configurations, yet flight-to-flight trajectory differences produce magnitude differences up to 10 dB and approximately 3 dB on average.This illustrates the effect of motion dynamics on measured propagation characteristics.Source: §VI · p. 5
- Dynamic RIS (a) outperforms static (b) and off (c): its average gain is approximately 0.5 dB over static in scenario (i) and up to 2.5 dB in scenarios (ii)–(iii).Static performance is more sensitive to flight disturbances.Source: §VI · pp. 5–6, Figure 5 · p. 6
- Figure 5 plots S21 magnitudes for three configuration methods across scenarios and shows representative ground-truth RIS poses only for the dynamic-RIS flight.The authors describe that flight as representative because the UAV trajectory is not affected by RIS configuration.Source: §VI · p. 5, Figure 5 · p. 6
- In scenario (ii), static RIS is inferior to off, whereas in scenario (iii) it outperforms off; [9] identifies overhead placement as sensitive to positional uncertainty.The authors connect this sensitivity to pronounced effective-channel variation from small UAV position changes.Source: §VI · p. 6
- With the RIS midway between Tx and Rx, S21 reaches −48.26 dB; at other locations magnitudes decrease by more than 20 dB, while scenario (ii) also has faster attitude changes and less stable flight.The authors associate scenario (ii)'s low altitude and Rx position beneath the UAV with rotor downwash.Source: §VI · p. 6
VII. CONCLUSION
The work presents a real-flight dataset combining synchronized channel, pose, and RIS-configuration data across deployment scenarios and configuration approaches. Initial analyses show that both RIS configuration and deployment location influence link performance, while small disturbances can substantially affect measured S21 magnitudes.
VII. CONCLUSION
- The dataset combines synchronized S21 measurements, UAV and RIS poses from the EKF and motion-capture system, and RIS configurations across multiple deployment scenarios and configuration approaches.The authors present it as a basis for future UAV-mounted RIS research and a contribution toward bridging theory and real-world deployments.Source: §VII · p. 6
- Initial dataset analyses show that configuration approach and deployment location influence achievable link performance, and that small disturbances can substantially affect measured S21 magnitudes.The authors note these disturbance effects as a limitation of purely theoretical analyses.Source: §VII · p. 6