Recent progress of deep learning (DL) and computer vision (CV) have paved the way for the application of DL-based CV technologies in 6G wireless communications. DL-based CV is data-hungry, and thus it is important to collect a massive vision dataset designed for wireless applications. An aim of this paper is to propose a vision dataset called Vision Objects for Millimeter and Terahertz Communications (VOMTC) consisting of 20,232 pairs of RGB and depth images, each of which is manually annotated with three object classes (person, mobile, and laptop) and their corresponding boxes. To demonstrate the efficacy of VOMTC, we develop a CV-aided beamforming technique called VOMTC-based beam management (VBM). In VBM, the location of the mobile is extracted via the VOMTC-trained object detector and then a beam heading toward the extracted location is generated. Due to the use of this special object detector tailored for identifying mobiles, VBM enhances the chance of transmitting directional beams to the mobile location. Using the VOMTC test dataset, we show that VBM achieves 15% improvement in the data rate over the conventional CV-aided beam management.

Millimeter wave (mmWave) and terahertz (THz) communications have been considered as the key techniques to support extremely high data rates in the 6G system. One main limitation of the mmWave/THz communications is the severe path loss and low penetration power. For these reasons, it is expected that mmWave/THz communication will be mainly employed in the ultra-dense network (UDN) environment. In order to get the most out of the mmWave/THz UDN, a mobile should be associated to the base stations (BSs) providing a high quality-of-service (QoS). This task is challenging since the reliable path can be disappeared even with a small movement of a mobile. An aim of this paper is to propose a novel mobility management technique based on sensing and computer vision (CV). Our key idea is to predict the cell association from the visual sensing information and CV-based inference and decision. By extracting the geometric information of a mobile from the image and then using it for the downlink rate prediction, we preemptively switch the cell association in UDN. From the numerical evaluations on the realistic mmWave/THz UDN environments, we show that the proposed scheme achieves more than 30% throughput gain over the conventional mobility management techniques.

Recently, terahertz (THz) communication supported by the ultra-dense network (UDN) has received a great deal of attention as a means to satisfy stringent requirements in throughput, latency, and energy consumption in 6G. In the UDN supported by THz beamforming, handover, an action to change the base station (BS) serving the user, occurs frequently due to the small cell coverage and sudden line-of-sight (LoS) link blockage caused by the interruption of obstacles. To ensure the seamless connectivity in the THz UDN, mobility management, the process to identify the link deterioration and perform the handover, should be performed quickly and accurately. An aim of this paper is to propose a novel computer vision (CV)-aided framework to proactively control the mobility management in the THz communication regime. In our framework referred to as proactive computer vision-aided mobility management (P-CVMM), the position of the user (i.e., distance and angles) is extracted from the images via deep learning (DL)-based object detector and then exploited in predicting the optimal serving BS (S-BS). Since the sparse geometric channel parameters are immediately obtained without the complicated feedback process and the beam heading toward the user's future position can be generated without quantization, P-CVMM can predictively avoid radio link failure (RLF). Using the specially designed dataset called vision objects for mobility management (VOMM), we demonstrate that P-CVMM effectively avoids RLF and achieves more than 90% and 70% reduction in the localization error and handover interruption time (HIT) over the conventional schemes.

Millimeter-wave (mmWave) communications will play an important role in 5G and 6G communication systems as a means to support extremely high data rates. One main bottleneck of the mmWave communication is the severe signal attenuation caused by the foliage loss, atmospheric absorption, body and hand losses in the mmWave band. To compensate for the severe path loss, multiple-input-multiple-output (MIMO) antenna array-based beamforming has been widely used. Since the beams should be aligned with the signal propagation paths to get the most of beamforming gain, acquisition of accurate channel knowledge, i.e., channel estimation, is the key to the success of mmWave MIMO systems. In this paper, we propose a new type of deep learning (DL)-based parametric channel estimation technique. In our work, DL figures out the direct mapping between the received pilot signal and the sparse channel parameters characterizing the angular domain channel. By exploiting the long short-term memory (LSTM) as a main deep neural network (DNN) engine, we extract the temporally correlated features of time-varying channel parameters and make a fast yet accurate estimation with relatively small pilot overhead. From the numerical experiments, we show that the proposed scheme is effective in estimating the mmWave MIMO channel in various mmWave downlink environments.

Recently, massive machine-type communications (mMTCs) have become one of key use cases for 5G. In order to support massive users transmitting small data packets at low rates, grant-free (GF) access and nonorthogonal multiple access (NOMA) have been suggested. Since each device transmits information without scheduling in the GF-NOMA systems, the device identification process, called active user detection (AUD), is required at the base station (BS). For the NOMA-based systems, the channel estimation (CE), an operation after the AUD, is a challenging task since multiple devices’ transmit signals and channels are superimposed in the same wireless resources. In this article, we propose a deep learning (DL)-based AUD and CE in the GF-NOMA systems. In our work, DL figures out the direct mapping between the received NOMA signal and the indices of active devices and associated channels using the long short-term memory (LSTM). From numerical experiments, we show that the proposed scheme is effective in handling the AUD and CE in the mMTC environments.

As a means to support the access of massive machine-type communication devices, grant-free access and nonorthogonal multiple access (NOMA) have received a lot of attention recently. In the grant-free environment, each device transmits information without scheduling. Hence, the device identification process called active user detection (AUD) is indispensable at the base station. After the AUD process, the channel estimation for active devices is performed in the base station before detecting the data. These processes are challenging problems in the NOMA-based systems since it is difficult to detect the active devices and estimate the channel of those devices from the superimposed received signal. In this paper, we propose a deep neural network (DNN)-based joint AUD and CE scheme for the practical mMTC systems. Specifically, the proposed scheme consists of long short term memory (LSTM)-based AUD (L-AUD) and DNN-based CE (D-CE). In L-AUD, by feeding the training data in the designed network, the proposed LSTM network is trained to exploit the extracted features when mapping the received NOMA signal to the indices of active devices. After the AUD process, by using the deeply stacked hidden layers, D-CE extracts the channel features and the codebook features of the active devices to map the received NOMA signal to the corresponding channel. As a result, the trained DNN can jointly handle the whole AUD and CE processes, achieving an accurate detection of the active devices and the small channel estimation error.

As a means to support the access of massive machine-type communication devices, grant-free access and non-orthogonal multiple access (NOMA) have received great deal of attention in recent years. In the grant-free transmission, each device transmits information without the granting process so that the base station needs to identify the active devices among all potential devices. This process, called an active user detection (AUD), is a challenging problem in the NOMA-based systems since it is difficult to identify active devices from the superimposed received signal. An aim of this paper is to put forth a new type of AUD based on deep neural network (DNN). By feeding the training data in the properly designed DNN, the proposed AUD scheme learns the nonlinear mapping between the received NOMA signal and indices of active devices. As a result, the trained DNN can handle the whole AUD process, achieving an accurate detection of the active users. Numerical results demonstrate that the proposed AUD scheme outperforms the conventional approaches by a large margin in both AUD success probability and computational complexity.

Massive machine-type communication (mMTC), key component for internet of things (IoT), concerns the access of massive machine-type communication devices to the basestation. To support the massive connectivity, grant-free access and non-orthogonal multiple access (NOMA) have been recently introduced. In the grant-free transmission, each device transmits information without the granting process so that the basestation needs to identify the active devices among all potential devices. This process, called an active user detection (AUD), is a challenging problem in the NOMA-based systems since it is difficult to find out the active devices from the superimposed received signal. An aim of this paper is to propose a new type of AUD scheme suitable for the highly overloaded mMTC, referred to as dimension spreading deep neural network-based AUD (DSDNN-AUD). The key feature of DSDNN-AUD is to set the dimension of hidden layers being larger than the size of a transmit vector to improve the representation quality of the support. In doing so, the proposed scheme can better discriminate the supports generated from correlated structured environment. Numerical results demonstrate that the proposed AUD scheme outperforms the conventional approaches in both AUD success probability and throughput performance.