Two-Way Channel Frequency Response Dataset Obtained with MCPD technique

Overview The dataset contains complex-valued two-way channel frequency response (CFR) measurements obtained using the phase-based ranging (PBR) technique. The dataset was created with the method developed in the paper [1], where the multi-carrier phase difference (MCPD) technique was integrated into the IEEE 802.15.4 time slotted channel hopping (TSCH) protocol, enabling distance estimation alongside communication. The dataset includes measurements from three setups under controlled indoor office conditions and provides both raw and interfered versions of the data. Measurement Campaign Description Measurements were performed in an office environment with two large magnetic boards providing strong reflective surfaces. The room contained typical office furniture and was unoccupied during the measurements. Four Wi-Fi access points were active in the 2.4 GHz band throughout the experiments. Two devices were placed in the center of the room on cardboard stands 50 cm above the table. Three LoS configurations were measured: Setup LoS Distance Description Setup 1 2 m Devices in direct LoS at 2 meters apart Setup 2 3 m Devices in direct LoS at 3 meters apart Setup 3 4 m Devices in direct LoS at 4 meters apart Hardware and Firmware The experiments used two VESNA devices equipped with AT86RF215 radios, which support IQ sampling. Devices were running Contiki-NG OS with a modified protocol stack enabling phase-based ranging as described in [1]. Measurement Method In each setup, the devices exchanged data packets while simultaneously measuring the wireless channel using the MCPD technique. One device acted as the initiator (`init`), and the other as the reflector (`ref`). After each packet exchange, devices sampled the in-phase and quadrature (IQ) components of the continuous-wave (CW) tones exchanged across the 2.4 GHz ISM band. The resulting MCPD configuration: - Frequency range: 2401.0 MHz to 2480.5 MHz - Frequency step: 0.5 MHz - Total tones: 160- Total bandwidth: 80 MHz Calibration Because low-cost radio hardware is used, the devices exhibit differences between their transmit and receive carrier frequencies, known as carrier frequency offset (CFO). Before the measurements, the devices were calibrated to minimize CFO, ensuring negligible frequency mismatch during operation. Each distance estimate also includes a fixed hardware-induced bias resulting from additional propagation delays through the antennas, cables, and circuit board traces. In the setup, the bias was measured at **1.8 m**. Post-processing calibration is required to remove it from the final results. Random Initial Phase Offsets Standalone CFR measurements from `init` or `ref` contain random phase offsets because the phase-locked loop (PLL) of the radio initializes with an unknown phase at each frequency step. To eliminate the offsets, the one-way CFRs from both nodes are combined through element-wise multiplication. The operation cancels the random phase components across frequencies, providing phase-coherent measurements and enabling distance estimation through inverse Fourier transform analysis (or others). The detailed derivation of this method is presented in [1] and [2]. The used hardware inverts the reflector measurements [2], therefore they have to be inverted prior the usage. One can simply use the Python with NumPy library, that provides conjugate function Two-way CFR drawbacks The multiplication of two one-way CFRs in the frequency domain corresponds to the convolution of their respective one-way CIRs in the time domain, which introduces artefact peaks. As a result, the two-way CIR exhibits apparent multipath components that make the LoS peak estimation challenging. As presented with the picture below, the one-way CIR peaks from 3.7 m (true LoS component) and peak at 8.4 m (multipath reflection) created an artefact at 5.9 m in the two-way CIR. Thus the LoS is no longer the highest peak in the two-way CIR. (FIGURE IS DISPLAYED AS PREVIEW OF THE DATASET) The problem can be addressed by converting the TWCFR back into the OWCFR through channel reconstruction. Details about channel reconstruction process can be found in [3]. References [1] G. Morano, K. Guan, A. Hrovat, and T. Javornik, *"Phase-Based Distance Estimation Integrated With IEEE 802.15.4 TSCH Communication"*, IEEE Internet of Things Journal, vol. 11, no. 7, pp. 11 460–11 470, Apr. 2024, issn: 2327-4662. doi: 10.1109/JIOT.2023.3330933. [2] G. Morano, A. Simončič, T. Kocevska, T. Javornik, and A. Hrovat, *"Distance- and Angle-Based Hybrid Localization Integrated in the IEEE 802.15.4 TSCH Communication Protocol"*, Sensors, vol. 24, no. 12, p. 3925, Jan. 2024, issn: 1424-8220. doi: 10.3390/s24123925. [3] G. Morano, T. Javornik, A. Simončič, A. Švigelj and A. Hrovat, *"Interference Mitigation in One-Way Channel Reconstruction for Robust Phase-Based Ranging"*, IEEE Access, Early Access, 2025, issn: 2169-3536. doi: 10.1109/ACCESS.2025.3649293### 数据集概述 本数据集包含基于相位测距（Phase-based Ranging, PBR）技术采集的复值双向信道频率响应（Channel Frequency Response, CFR）测量数据。本数据集采用文献[1]中提出的方法构建，该方法将多载波相位差（Multi-carrier Phase Difference, MCPD）技术集成至IEEE 802.15.4时分跳信道（Time Slotted Channel Hopping, TSCH）协议，可在通信的同时实现距离估计。数据集包含受控室内办公环境下三种配置的测量数据，并提供原始数据与受干扰数据两个版本。 ### 测量活动描述 本次测量在配备两块大型磁性板（提供强反射表面）的办公环境中开展，室内摆放常规办公家具，测量期间房间无人员在场。实验全程有4个工作在2.4 GHz频段的Wi-Fi接入点处于活跃状态。两台设备被放置在房间中央的纸板支架上，距桌面高度为50 cm。 本次测量包含三种视距（Line of Sight, LoS）配置： | 配置编号 | 视距间距 | 描述 | | ---- | ---- | ---- | | 配置1 | 2 m | 两台设备间距2 m，处于直接视距环境 | | 配置2 | 3 m | 两台设备间距3 m，处于直接视距环境 | | 配置3 | 4 m | 两台设备间距4 m，处于直接视距环境 | ### 硬件与固件 本次实验使用两台搭载AT86RF215无线电模块的VESNA设备，该模块支持同相正交（In-phase and Quadrature, IQ）采样。设备运行Contiki-NG操作系统，并通过修改后的协议栈实现了文献[1]中所述的相位测距功能。 ### 测量方法 在每种配置下，两台设备在交换数据包的同时，通过MCPD技术对无线信道进行测量。其中一台设备作为发起端（`init`），另一台作为反射端（`ref`）。每次数据包交换完成后，设备对2.4 GHz工业、科学和医疗（Industrial, Scientific, Medical, ISM）频段内交换的连续波（Continuous Wave, CW）音调的同相及正交分量进行采样。本次实验采用的MCPD配置参数如下： - 频率范围：2401.0 MHz 至 2480.5 MHz - 频率步长：0.5 MHz - 总音调数：160 - 总带宽：80 MHz ### 校准 由于本次实验采用低成本无线电硬件，设备的发射与接收载波频率存在差异，该差异被称为载波频率偏移（Carrier Frequency Offset, CFO）。测量前，已对设备进行校准以最小化CFO，确保运行过程中的频率失配可忽略不计。 此外，每次距离估计结果均包含由天线、线缆及电路板走线带来的额外传播延迟所导致的固定硬件偏差。本次实验中该偏差经测量为**1.8 m**，后续后处理阶段需将其从最终结果中移除。 ### 随机初始相位偏移 单独采集的发起端或反射端的信道频率响应数据包含随机相位偏移，原因在于无线电的锁相环（Phase-Locked Loop, PLL）在每个频率步长下均以未知相位初始化。为消除该偏移，将两个节点的单向信道频率响应进行逐元素相乘操作，该操作可抵消各频率下的随机相位分量，得到相位相干的测量数据，并可通过逆傅里叶变换分析（或其他方法）实现距离估计。该方法的详细推导过程可参见文献[1]与[2]。 由于本次使用的硬件会对反射端的测量数据进行反转处理，因此在使用前需先对其执行反转操作。用户可借助搭载NumPy库的Python环境完成该操作，NumPy库提供了共轭函数（conjugate function）以实现相关运算。 ### 双向信道频率响应的局限性 在频域中将两个单向信道频率响应相乘，等价于其时域中各自对应的单向信道冲激响应（Channel Impulse Response, CIR）进行卷积运算，该过程会引入伪影峰值。因此，双向信道冲激响应会呈现出虚假的多径分量，使得视距峰值的估计难度增加。 如下方图示所示，来自3.7 m（真实视距分量）的单向信道冲激响应峰值与8.4 m处的多径反射峰值，会在双向信道冲激响应中产生一个位于5.9 m处的伪影峰值。此时，视距分量不再是双向信道冲激响应中的最高峰值。 （数据集预览中展示了该图示） 该问题可通过将双向信道频率响应重构为单向信道频率响应的方式解决。关于信道重构过程的详细细节可参见文献[3]。 ### 参考文献 [1] G. Morano, K. Guan, A. Hrovat, and T. Javornik, *Phase-Based Distance Estimation Integrated With IEEE 802.15.4 TSCH Communication*, IEEE Internet of Things Journal, vol. 11, no. 7, pp. 11460–11470, Apr. 2024, ISSN: 2327-4662, DOI: 10.1109/JIOT.2023.3330933. [2] G. Morano, A. Simončič, T. Kocevska, T. Javornik, and A. Hrovat, *Distance- and Angle-Based Hybrid Localization Integrated in the IEEE 802.15.4 TSCH Communication Protocol*, Sensors, vol. 24, no. 12, p. 3925, Jan. 2024, ISSN: 1424-8220, DOI: 10.3390/s24123925. [3] G. Morano, T. Javornik, A. Simončič, A. Švigelj and A. Hrovat, *Interference Mitigation in One-Way Channel Reconstruction for Robust Phase-Based Ranging*, IEEE Access, Early Access, 2025, ISSN: 2169-3536, DOI: 10.1109/ACCESS.2025.3649293.