Search for: All records

Large-scale channel prediction, i.e., estimation of the pathloss from geographical/morphological/building maps, is an essential component of wireless network planning. Ray tracing (RT)-based methods have been widely used for many years, but they require significant computational effort that may become prohibitive with the increased network densification and/or use of higher frequencies in B5G/6G systems. In this paper, we propose a data-driven, model-free pathloss map prediction (PMP) method, called PMNet. PMNet uses a supervised learning approach: it is trained on a limited amount of RT data and map data. Once trained, PMNet can predict pathloss over location with high accuracy (an RMSE level of 10−2 ) in a few milliseconds. We further extend PMNet by employing transfer learning (TL). TL allows PMNet to learn a new network scenario quickly ( ×5.6 faster training) and efficiently (using ×4.5 less data) by transferring knowledge from a pre-trained model, while retaining accuracy. Our results demonstrate that PMNet is a scalable and generalizable ML-based PMP method, showing its potential to be used in several network optimization applications. 

Since the terahertz frequency band (0.1–1 THz) has attracted considerable attention for the upcoming sixth-generation (6G) wireless communication systems, accurate models for multipath propagation in this frequency range need to be established. Such models advantageously use the fact that multi-path components (MPCs) occur typically in clusters, i.e., groups of MPCs that have similar delays and angles. In this paper, we first analyze the limitations of a widely used clustering algorithm, Kernel-Power-Density (KPD), in evaluating an extensive THz outdoor measurement campaign at 145–146 GHz, particularly its inability to detect small clusters. We introduce a modified version, which we term multi-level KPD (ML-KPD), iteratively applying KPD to detect whether a cluster determined in the previous round is made up of multiple clusters. We first apply the method to synthetic channels to demonstrate its efficacy and select suitable values for the adaptive hyperparameters. Then, multi-level KPD is applied to our channel measurements in line-of-sight (LOS) and non-line-of-sight (NLOS) environments to determine statistics for the number of clusters and the cluster spreads. 

Pathloss is one of the essential characteristics of wireless propagation channels. It is usually captured from channel measurements with (quasi)isotropic antennas. To characterize the wireless channels at high frequencies, beamforming or directional antennas are commonly used, in which case a method for estimating the isotropic pathloss is needed. The method should account for the possible spatial overlap of the different directional measurements while including the received signal from all the multipath components in the channel. In this letter, we propose an efficient method that uses a weighted sum of the powers received from the directional measurements. The weights can be calculated using matrix inversion. We verify the solution using synthetic data and demonstrate the usage with measurements at sub-THz frequencies. 

The THz band has attracted considerable attention for next-generation wireless communications due to the large amount of available bandwidth that may be key to meet the rapidly increasing data rate requirements. Before deploying a system in this band, a detailed wireless channel analysis is required as the basis for proper design and testing of system implementations. One of the most important deployment scenarios of this band is the outdoor microcellular environment, where the Transmitter (Tx) and the Receiver (Rx) have a significant height difference (typically ≥10 m). In this paper, we present double-directional (i.e., directionally resolved at both link ends) channel measurements in such a microcellular scenario encompassing street canyons and an open square. Measurements are done for a 1 GHz bandwidth between 145–146 GHz and an antenna beamwidth of 13 degree; distances between Tx and Rx are up to 85 m and the Tx is at a height of 11.5 m from the ground. The measurements are analyzed to estimate path loss, shadowing, delay spread, angular spread, and multipath component (MPC) power distribution. These results allow the development of more realistic and detailed THz system performance assessment. 

Pathloss prediction is an essential component of wireless network planning. While ray tracing based methods have been successfully used for many years, they require significant computational effort that may become prohibitive with the increased network densification and/or use of higher frequencies in 5G/B5G (beyond 5G) systems. In this paper, we propose and evaluate a data-driven and model-free pathloss prediction method, dubbed PMNet. This method uses a supervised learning approach: training a neural network (NN) with a limited amount of ray tracing (or channel measurement) data and map data and then predicting the pathloss over location with no ray tracing data with a high level of accuracy. Our proposed pathloss map prediction-oriented NN architecture, which is empowered by state-of-the-art computer vision techniques, outperforms other architectures that have been previously proposed (e.g., UNet, RadioUNet) in terms of accuracy while showing generalization capability. Moreover, PMNet trained on a 4-fold smaller dataset surpasses the other baselines (trained on a 4-fold larger dataset), corroborating the potential of PMNet.1 

This paper describes our pathloss prediction system submitted to the ICASSP 2023 First Pathloss Radio Map Prediction Challenge. We describe the architecture of PMNet, a neural network we specifically designed for pathloss prediction. Moreover, to enhance the prediction performance, we apply several machine learning techniques, including data augmentation, fine-tuning, and optimization of the network architecture. Our system achieves an RMSE of 0.02569 on the provided RadioMap3Dseer dataset, and 0.0383 on the challenge test set, placing it in the 1st rank of the challenge. 

The availability of large bandwidths in the terahertz (THz) band will be a crucial enabler of high data rate applications in next-generation wireless communication systems. The urban microcellular scenario is an essential deployment scenario where the base station (BS) is significantly higher than the user equipment (UE). Under practical operating conditions, moving objects (i.e., blockers) can intermittently obstruct various parts of the BSUE link. Therefore, in the current paper, we analyze the effect of such blockers. We assume a blockage of the strongest beam pair and investigate the availability and extent of angular diversity, i.e., alternative beampairs that can sustain communication when the strongest is blocked. The analysis uses double-directional channel measurements in urban microcellular scenarios for 145- 146 GHz with BS-UE distances between 18 to 83 m. We relate the communication-system quantities of beam diversity and capacity to the wireless propagation conditions. We show that the SNR loss due to blockage depends on the blocked angular range and the specific location, and we find mean blockage loss to be on the order of 10-20 dB in line-of-sight (LOS) and 5-12 dB in NLOS (non-LOS). This analysis can contribute to the design of intelligent algorithms or devices (e.g., beamforming, intelligent reflective surfaces) to overcome the impact of the blockage. 

Multi-path components (MPCs) in wireless channels generally occur in clusters, i.e., groups of MPCs that have similar delay/angle characteristics. However, when those clusters are widely separated and have significantly different power, highresolution parameter extraction (HRPE) algorithms based on serial interference cancellation, such as CLEAN, can miss some of the weaker clusters because they concentrate the path search in the strongest cluster. This effect can occur particularly in the presence of calibration error and/or diffuse scattering. To solve this problem, we propose a heuristic modification, Regional CLEAN (R-CLEAN), that employs cluster identification in the Fourier domain and limits the number of MPCs per cluster. We first demonstrate the effect, and the effectiveness of our proposed algorithm, on synthetic channels with calibration error or diffuse scattering. We then demonstrate it with a THz Multiple-Input- Multiple-Output (MIMO) measurement at 145 - 146 GHz. The proposed optimization and algorithm can thus be an essential step towards evaluating channels with multiple clusters.