The rapid uptake of mobile devices and the rising popularity of mobile applications and services pose unprecedented demands on mobile and wireless networking infrastructure. Upcoming 5G systems are evolving to support exploding mobile traffic volumes, real-time extraction of fine-grained analytics, and agile management of network resources, so as to maximize user experience. Fulfilling these tasks is challenging, as mobile environments are increasingly complex, heterogeneous, and evolving. One potential solution is to resort to advanced machine learning techniques, in order to help manage the rise in data volumes and algorithm-driven applications. The recent success of deep learning underpins new and powerful tools that tackle problems in this space. In this paper, we bridge the gap between deep learning and mobile and wireless networking research, by presenting a comprehensive survey of the crossovers between the two areas. We first briefly introduce essential background and state-of-the-art in deep learning techniques with potential applications to networking. We then discuss several techniques and platforms that facilitate the efficient deployment of deep learning onto mobile systems. Subsequently, we provide an encyclopedic review of mobile and wireless networking research based on deep learning, which we categorize by different domains. Drawing from our experience, we discuss how to tailor deep learning to mobile environments. We complete this survey by pinpointing current challenges and open future directions for research.

/pdf/deep-learning-in-mobile-and-wireless-networking-a-survey-ct8cy2kuqi.pdf

Deep Learning in Mobile and Wireless Networking: A Survey

The rapid uptake of mobile devices and the rising popularity of mobile applications and services pose unprecedented demands on mobile and wireless networking infrastructure. Upcoming 5G systems are evolving to support exploding mobile traffic volumes, agile management of network resource to maximize user experience, and extraction of fine-grained real-time analytics. Fulfilling these tasks is challenging, as mobile environments are increasingly complex, heterogeneous, and evolving. One potential solution is to resort to advanced machine learning techniques to help managing the rise in data volumes and algorithm-driven applications. The recent success of deep learning underpins new and powerful tools that tackle problems in this space. 
In this paper we bridge the gap between deep learning and mobile and wireless networking research, by presenting a comprehensive survey of the crossovers between the two areas. We first briefly introduce essential background and state-of-the-art in deep learning techniques with potential applications to networking. We then discuss several techniques and platforms that facilitate the efficient deployment of deep learning onto mobile systems. Subsequently, we provide an encyclopedic review of mobile and wireless networking research based on deep learning, which we categorize by different domains. Drawing from our experience, we discuss how to tailor deep learning to mobile environments. We complete this survey by pinpointing current challenges and open future directions for research.

In fifth-generation (5G) communications, millimeter wave (mmWave) is one of the key technologies to increase the data rate. To overcome this technology's poor propagation characteristics, it is necessary to employ a number of antennas and form narrow beams. It becomes crucial then, especially for initial access, to attain fine beam alignment between a next generation NodeB (gNB) and a user equipment (UE). The current 5G New Radio (NR) standard, however, adopts an exhaustive search-based beam sweeping, which causes time overhead of a half frame for initial beam establishment. In this paper, we propose a deep learning-based beam selection, which is compatible with the 5G NR standard. To select a mmWave beam, we exploit sub-6 GHz channel information. We introduce a deep neural network (DNN) structure and explain how we estimate a power delay profile (PDP) of a sub-6 GHz channel, which is used as an input of the DNN. We then validate its performance with real environment-based 3D ray-tracing simulations and over-the-air experiments with a mmWave prototype. Evaluation results confirm that, with support from the sub-6 GHz connection, the proposed beam selection reduces the beam sweeping overhead by up to 79.3 %.

Deep Learning-Based mmWave Beam Selection for 5G NR/6G With Sub-6 GHz Channel Information: Algorithms and Prototype Validation

With vast mmWave spectrum and narrow beam antenna technology, precise position location is now possible in 5G and future mobile communication systems. In this article, we describe how centimeter-level localization accuracy can be achieved, particularly through the use of map-based techniques. We show how data fusion of parallel information streams, machine learning, and cooperative localization techniques further improve positioning accuracy.

Position Location for Futuristic Cellular Communications: 5G and Beyond

Existing fingerprint-based indoor localization uses either fine-grained channel state information (CSI) from the physical layer or coarse-grained received signal strength indicator (RSSI) measurements. In this paper, we propose to use a mid-grained intermediate-level channel measurement - spatial beam signal-to-noise ratios (SNRs) that are inherently available and defined in the IEEE 802.11ad/ay standards - to construct the fingerprinting database. These intermediate channel measurements are further utilized by a deep learning approach for multiple purposes: 1) location-only classification; 2) simultaneous location-and-orientation classification; and 3) direct coordinate estimation. Furthermore, the effectiveness of the framework is thoroughly validated by an in-house experimental platform consisting of 3 access points using commercial-off-the-shelf millimeter-wave WiFi routers. The results show a 100% accuracy if the location is only interested, about 99% for simultaneous location-and-orientations classification, and an averaged root mean-square error (RMSE) of 11.1 cm and an average median error of 9.5 cm for direct coordinate estimate, greater than 2-fold improvements over the RMSE of 28.7 cm and median error of 23.6 cm for RSSI-like single SNR-based localization.

/pdf/fingerprinting-based-indoor-localization-with-commercial-qb9vf6g67o.pdf

Fingerprinting-Based Indoor Localization With Commercial MMWave WiFi: A Deep Learning Approach

Over the last years positioning systems have become increasingly pervasive, covering most of the planet’s surface. Although they are accurate enough for a large number of uses, their precision, power consumption, and hardware requirements establish the limits for their adoption in mobile devices. In this paper, the energy consumption of a proposed deep learning-based millimeter wave positioning method is assessed, being subsequently compared to the state-of-the-art on accurate outdoor positioning systems. Requiring as low as 0.4 mJ per position fix, when compared to the most recent assisted-GPS implementations the proposed method has energy efficiency gains of $\mathbf {47\times }$ and $\mathbf {85\times }$ for continuous and sporadic position fixes (respectively), while also having slightly lower estimation errors. Therefore, the proposed method significantly reduces the energy required for precise positioning in the presence of millimeter wave networks, enabling the design of more efficient and accurate positioning-enabled mobile devices.

Dethroning GPS: Low-Power Accurate 5G Positioning Systems Using Machine Learning

The introduction of 5G’s millimeter wave transmissions brings a new paradigm to wireless communications. Whereas physical obstacles were mostly associated with signal attenuation, their presence now adds complex, non-linear phenomena, including reflections and scattering. The result is a multipath propagation environment, shaped by the obstacles encountered, indicating a strong presence of hidden spatial information within the received signal. To untangle said information into a mobile device position, this paper proposes the usage of neural networks over beamformed fingerprints, enabling a single-anchor positioning approach. Depending on the mobile device target application, positioning can also be enhanced with tracking techniques, which leverage short-term historical data. The main contributions of this paper are to discuss and evaluate typical neural network architectures suitable to the beamformed fingerprint positioning problem, including convolutional neural networks, hierarchy-based techniques, and sequence learning approaches. Using short sequences with temporal convolutional networks, simulation results show that stable average estimation errors of down to 1.78 m are obtained on realistic outdoor scenarios, containing mostly non-line-of-sight positions. These results establish a new state-of-the-art accuracy value for non-line-of-sight millimeter wave outdoor positioning, making the proposed methods very competitive and promising alternatives in the field.

Deep Learning Architectures for Accurate Millimeter Wave Positioning in 5G

With millimeter wave wireless communications, the resulting radiation reflects on most visible objects, creating rich multi path environments, namely in urban scenarios. The radiation captured by a listening device is thus shaped by the obstacles encountered, which carry latent information regarding their relative positions. In this paper, a system to convert the received millimeter wave radiation into the device's position is proposed, making use of the aforementioned hidden information. Using deep learning techniques and a pre-established codebook of beamforming patterns transmitted by a base station, the simulations show that average estimation errors below 10 meters are achievable in realistic outdoors scenarios that contain mostly non-line-of-sight positions, paving the way for new positioning systems.

Beamformed Fingerprint Learning for Accurate Millimeter Wave Positioning

Millimeter wave frequencies paired up with MIMO antennas employing beamforming are seen as critical enablers of next generation networks. However, selecting the most beneficial beamforming weights in a codebook-enabled downlink transmitter is a lengthy task, as the existing methods rely on some form of channel measurement. In fact, if the used codebook is too large, the traditional methods might fail to select an appropriate entry within the channel coherence time. In this paper, a new method to assist the beam selection is proposed, based on data obtained from previous connections. Through the continuous update of the set of optimal codebook entries for each position, the search space required for each connection can be greatly reduced if the user position is known. The simulations performed show that retrieving the sets of codebook entries in single user scenarios required less than 51 ns. For multi-user scenarios, results exceeding 10 simultaneous users using a 16 entry codebook were achieved, requiring less than $600\ \mu s$ . The obtained results show that the proposed method can greatly reduce the beam selection latency and energy requirements, opening the door to powerful millimeter wave networks.

Data-Aided Fast Beamforming Selection for 5G

High-end FPGAs are widely adopted as hardware accelerators, due to their power efficiency, flexibility, and high-performance computing ability. They are, therefore, extremely useful devices for a project with challenges and constraints such as the Square Kilometre Array (SKA). However, the traditional design methods require expert hardware knowledge and long development times for each of the SKA's target applications, making it difficult to make the most out of an array of FPGAs as a shared resource. High-level development approaches are positioned to overcome this issue. In this paper, we investigate the development efficiency and achievable performance of two popular high-level methods, Maxeler's MaxCompiler and OpenCL. They are evaluated by implementing a lengthy FIR filter with complex single precision floating-point (SPF) numbers, both in time and frequency domains (TDFIR and FDFIR, respectively). Our results show that high performance can be achieved with low development effort using such high-level methods, where OpenCL outperforms the MaxCompiler for TDFIR. OpenCL is flexible enough to develop and compare different filter approaches quickly, and as expected FDFIR based implementations clearly outperform TDFIR based ones. To demonstrate OpenCL portability and to compare performance with GPUs, the filters are also evaluated on a GPU platform. The evaluation shows that while the GPU performs better in TDFIR, is is outperformed by the FPGA in FDFIR.

Joao Gante

Papers

Dethroning GPS: Low-Power Accurate 5G Positioning Systems Using Machine Learning

Deep Learning Architectures for Accurate Millimeter Wave Positioning in 5G

Beamformed Fingerprint Learning for Accurate Millimeter Wave Positioning

Data-Aided Fast Beamforming Selection for 5G

High-Level Designs of Complex FIR Filters on FPGAs for the SKA