Integrated sensing and communication have drawn great research attention in recent years. Specifically, 5G mmWave has demonstrated its capabilities not only in high-speed communications but also in perceiving the physical environment. Apart from providing locationing services for user equipment (UE), 5G mmWave can also estimate the position of target objects that does not carry any equipment (i.e., device-free). Existing works of device-free wireless localization often employs a single monostatic radar or a few transceivers in fixed positions. In this work, we examine a cooperative sensing case, where multiple UEs cooperate with the infrastructure of transmit/receive points (TRPs) to jointly locate device-free objects. This new setting brings in new challenges for existing locationing algorithms as the number and the locations of the UEs and sensing targets are all dynamic. Our work proposes a novel procedure that uses visualization methods to jointly represent the information in the mmWave channel impulse responses and the locations of UEs and TRPs. We then introduce an end-to-end deep learning transformer architecture inspired by popular models in the computer vision domain to estimate the target objects’ locations from the visualizations. On a dataset generated using 3D ray-tracing simulations, our system can locate multiple device-free objects with an average error of 0.47 meters within a 20 meterby-40 meter experiment area.

Ray tracing is one of the de-facto standard method-ologies for radio channel modelling, given the geographical map of the layout. However, the channel generated by ray-tracing cannot be adapted to incorporate knowledge from real-world channel measurements. Several recent papers have proposed training a deep neural network (DNN) to compute the radio map for a given input layout. Such techniques typically require a large number of measurements, transmitters and receivers to generate the dataset needed for training the DNN, and hence can only be trained on simulated data from ray tracing. We propose an extension to these techniques, whereby we first train our DNN on simulated data, and then use a small number of measurements from a given setting to predict the path loss at all locations of interest, borrowing from a generative modelling technique called Deep Image Prior. Our simulations show that Radio DIP can achieve a RMSE of 5 dB in predicting the path loss of 50k outdoor locations, given less than 100 measurements.

Estimating path loss for a transmitter-receiver location is key to many use-cases including network planning and handover. Machine learning has become a popular tool to predict wireless channel properties based on map data. In this work, we present a transformer-based neural network architecture that enables predicting link-level properties from maps of various dimensions and from sparse measurements. The map contains information about buildings and foliage. The transformer model attends to the regions that are relevant for path loss prediction and, therefore, scales efficiently to maps of different size. Further, our approach works with continuous transmitter and receiver coordinates without relying on discretization. In experiments, we show that the proposed model is able to efficiently learn dominant path losses from sparse training data and generalizes well when tested on novel maps.

The evolution of massive multiple-input multiple-output (MIMO), such as holographic MIMO and reconfigurable intelligent surface (RIS), is one of the hottest topics in the sixth generation (6G) mobile communication systems. In this area, the network node may be equipped with much more antenna elements and/or transmit-receive units (TxRUs) than its presence in current deployment. To obtain accurate channel estimation for these nodes, using high-density pilot may result in tremendous downlink overhead, so pilot reduction is a key area to investigate. In this work, we design a low-density frequency-aware antenna selection pattern and a transformer (TF)-based channel estimator that is deployed at the user equipment (UE) side to recover the channel. Compared to the two-sided model where the pilot transmission is based on a neural network (NN) that is jointly optimized with the UE side channel estimator, the proposed design is more convenient for practical implementation. The simulation results suggest that our proposed pattern can achieve similar channel estimation performance as the two-sided model and can save 50% pilot overhead compared to the conventional method. Moreover, we demonstrate that our proposed design is robust against different antenna selection patterns.

We present our study results on the neural network based approaches to the transmitter/receiver design over nonlinear channel. In the first part, we discuss the transmitter design for the peak-to-average power (PAPR) reduction. The algorithm unrolling is applied to a well-known PAPR reduction algorithm. The learning-based approach achieves the performance on par with the classical approach, but at the smaller computational complexity. The second part discusses the joint transmitter/receiver design for the mitigation of the transmitter nonlinearity. It is shown that by combining the neural network with the unfolded algorithm for PAPR reduction at the transmitter and employing the receiver neural network, sizable gain is achieved when the emission criterion is relaxed.

Multicarrier direct-sequence spread-spectrum multiple-access systems in frequency-selective fading channels are investigated. A consistent channel model is used for each system. First, the tapped delay line (TDL) channel model with uniformly spaced, uncorrelated taps is investigated to model a complex Gaussian, wide-sense stationary, uncorrelated scattering channel. An approximate average bit-error-rate expression is obtained for the receiver with RAKE fingers on each subcarrier branch, whose outputs are combined according to the maximum signal-to-noise ratio criterion. Various system configurations are examined for the same TDL channel model, data rate, total system bandwidth, and excess bandwidth of the chip waveform. All the systems compared employ the same number of correlators. The numerical results show that, given a contiguous spectrum, the systems with more fractionally-spaced RAKE fingers per subchannel are more robust than the systems with more subcarriers.

A subspace-based linear minimum mean-squared error (MMSE) multiuser detection scheme is proposed for a multicarrier direct-sequence code-division multiple-access (MC-DS-CDMA) system. Typically, a MC-DS-CDMA system employs a band-limited chip waveform. The band-limited nature of the chip waveform causes problem in applying standard subspace techniques because no nonnull noise subspace can be formed. It is shown that channel and timing information needed for the construction of the linear MMSE detector can be identified by a multiple-signal-classification-like algorithm based on a finite-length truncation approximation of the chip waveform. In practice, since perturbed versions of the subspaces assumed in the finite-length truncation approximation are actually observed, and because of the band-limited property of the chip waveform, the accuracy of the channel estimation and, hence, the performance of the MMSE detector are degraded. This effect is investigated in this paper.

We consider subspace-based MMSE multiuser detection for multicarrier code division multiple access (MC-CDMA) in Rayleigh fading channels. The knowledge of the fading coefficients for all the carriers and the timing delay of the desired user is needed to construct the MMSE detector. A subspace-based channel estimation technique is developed to obtain the required information. Numerical results show that the proposed algorithm is robust to fading and moderate near-far situations.