China Mobile Research Institute

About: China Mobile Research Institute is a based out in . It is known for research contribution in the topics: MIMO & Wireless network. The organization has 579 authors who have published 542 publications receiving 13897 citations.

Mobile Data Traffic Prediction by Exploiting Time-Evolving User Mobility Patterns

Abstract: Understanding mobile data traffic and forecasting future traffic trend is beneficial to wireless carriers and service providers who need to perform resource allocation and energy saving management. However, predicting wireless traffic accurately at large-scale and fine-granularity is particularly challenging due to the following two factors: the spatial correlations between the network units (i.e., a cell tower or an access point) introduced by user arbitrary movements, and the time-evolving nature of user movements which frequently changes with time. In this paper, we use a time-evolving graph to formulate the time-evolving nature of user movements, and propose a model Graph-based Temporal Convolutional Network (GTCN) to predict the future traffic of each network unit in a wireless network. GTCN can bring significant benefits to two aspects. (1) GTCN can effectively learn intra- and inter-time spatial correlations between network units in a time-evolving graph through a node aggregation method. (2) GTCN can efficiently model the temporal dynamics of the mobile traffic trend from different network units through a temporal convolutional layer. Experimental results on two real-world datasets demonstrate the efficiency and efficacy of our method. Compared with state-of-the-art methods, the improvement of the prediction performance of our GTCN is 3.2\% to 10.2\% for different prediction horizons. GTCN also achieves 8.4 $\times$ faster on prediction time.

Cells Planning of VLC Networks using Non-Circular Symmetric Optical Beam

Abstract: In the typical indoor office environments, the existing site locations resource on ceiling for visible light communications (VLC) access points (AP) is relatively limited due to the originally illumination & fire-fighting oriented ceiling planning. For reducing the capital expenditure of VLC networking in most situations, the straight forward solution is making full use of the limited linear location of luminaries. In this work, for the first time, the non-circular symmetric optical beam is introduced to the cells planning of VLC networks. The coverage performance is compared between the typical non-circular symmetric optical beam pattern i.e. BBE LED pattern and the well investigated circular symmetric optical beam pattern i.e. Lambertian pattern. It is sufficiently demonstrated that the satisfying matching capability of this non-circular symmetric optical beam to indoor scenario with just 3 linear site locations. By appropriately placing the BBE LED pattern, more than 12 dB the signal-to-interference-plus-noise ratio (SINR) enhancement is provided for more than 70% receiver positions while the total optical transmitted power of all VLC AP remains unchanged. Moreover, the superior performance uniformity of this optical beam pattern is separately identified in cellular cells configuration case and in electrical frequency reuse case.

TD-LTE Network Evolution with In-Band and Out-Band Micro Cells Deployment

Abstract: the wireless network gradually changes from voice service to broadband data access services and faces unprecedented growth in data traffic. The way to meet future capacity demand is by combining macro site expansion and outdoor micro layer deployment. Thus it is important for the operator to understand the basic cost and performance of different combinations of technologies according to their own network. In this paper, year-by-year network performance is simulated for an example TD-LTE network by using four networks upgrade approaches, including carrier upgrade, macro densification, micro deployment in dedicated carrier and micro deployment re-using macro carrier. Downlink TD-LTE network capacity evolution is analyzed in terms of the needed network upgrades. That is, we examine the number of macro and micro site need to be upgraded or added in each year for different network evolution paths, while a given network key performance indication (KPI) is maintained. The network cost in terms of relative Total Cost of Ownership (TCO) is then compared among network evolution paths.

TD-LTE Network Deployment Evolution in a Metropolitan Scenario

Abstract: The wireless network gradually changes from voice service to broadband data access services and faces unprecedented growth in data traffic. The operator needs to prepare network upgrade road map in advance according to their own situation. TD-LTE is considered as providing the evolution path for TD-SCDMA. To reduce network CAPEX and OPEX, TD-LTE network is advisable to re-use TD-SCDMA site and share antennas with TD-SCDMA system, in which 8 element antennas are widely used in Marco Node B for beam forming. The scope of this paper is to demonstrate with simulations how the TD-LTE network configuration is evolving under the increase of traffic. An existing TD-SCDMA site and its base station configurations were used as the staring configuration for the network upgrade. Example results are presented with 2 traffic growth scenarios. The simulation results show that TD-LTE network upgrade configuration is greatly impacted by the traffic growth forecast. It also shows traditional network evolution technique such as spectrum upgrade is an efficient way to increase capacity in roll-out phase of TD-LTE network deployment.

Fixed-Sign Binary Neural Network: An Efficient Design of Neural Network for Internet-of-Things Devices

Abstract: High computational requirement and rigorous memory cost are the significant issues which limit Convolutional Neural Networks' deployability in resource-constrained environments typically found in edge devices of Internet-of-Things (IoT). To address the problem, binary and ternary networks have been proposed to constrain the weights to reduce computational and memory costs. However, owing to the binary or ternary values, the backward propagations are not as efficient as normal during training, which makes it tough to train in edge devices. In this paper, we find a different way to resolve the problem and propose Fixed-Sign Binary Neural Network (FSB), which decomposes convolution kernel into sign and scaling factor as the prior researches but only trains the scaling factors instead of both. By doing so, our FSB avoids the sign involved in backward propagations and makes models easy to be deployed and trained in the IoT devices. Meanwhile, the convolution-acceleration architecture which we design for our FSB results in a reduced computing burden while achieving the same performance. Thanks to the efficiency of our FSB, even though we randomly initialize the sign and fix it to be untrainable, our FSB still has remarkable performances.