Energy harvesting (EH) is of prime importance for enabling the Internet of Things (IoT) networks. Although, energy harvesting relays have been considered in the literature, most of the studies do not account for the processing costs, such as the decoding cost in a decode-and-forward (DF) relay. However, it is known that the decoding cost amounts to a significant fraction of the circuit power required for receiving a codeword. Hence, in this work, we are motivated to consider an EH-DF relay with the decoding cost and maximize the average number of bits relayed by it with a time-switching architecture. To achieve this, we first propose a frame structure consisting of three phases: (i) an energy harvesting phase, (ii) a reception phase and, (iii) a transmission phase. We obtain optimal length of each of the above phases and communication rates that maximize the average number of bits relayed. We consider the radio frequency (RF) energy to be harvested by the relay is from the dedicated transmitter and the multiple block case when energy is allowed from flow among the blocks, different from the single block case when energy is not allowed to flow among the blocks. By exploiting the convexity of the optimization problem, we derive analytical optimum solutions under the EH scenario. One of the optimal receiving rate for the relay is the same as in single block case. We also provide numerical simulations for verifying our theoretical analysis.

With the rapid development of smart grids, significant research has been devoted to the methodologies for short-term load forecasting (STLF) due to its significance in forecasting demand on electric power. In this paper an enhanced LSTM model is proposed to upgrade the state-of-the-art LSTM network by exploiting the long periodic information of load, which is missed by the standard LSTM model due to its constraint on input length. In order to distill information from long load sequence and keep the input sequence short enough for LSTM, the long load sequence is reshaped into two-dimension matrix whose dimension accords to the periodicity of load. Accordingly, two LSTM networks are paralleled: one takes the rows as input to extract the temporal pattern of load in short time, while the other one takes the columns as input to distill the periodicity information. A multi-layer perception combines the two outputs for more accurate load forecasting. This model can exploit more information from much longer load sequence with only linear growth in complexity, and the experiment results verify its considerable improvement in accuracy over the standard LSTM model.

With the emergence and development of the Internet of Vehicles (IoV), higher demands are placed on the response speed and ultra-low delay of the vehicle. Cloud computing services are not friendly to reducing latency and response time. Mobile Edge Computing (MEC) is a promising solution to this problem. In this paper, we introduce MEC into the IoV to propose a specific vehicle edge resource management framework, which consists of fog nodes (FN), data service agents (DSA), and cars. We proposed a dynamic service area partitioning algorithm that enables the DSA to adjust the service area and provide a more efficient service for the vehicle. A resource allocation framework based on Stackelberg game model is proposed to analyze the pricing problem of FN and data resource strategy of DSA. We use the distributed iterative algorithm to solve the problem of game equilibrium. Our proposed resource management framework is finally verified by numerical results, which show that the allocation efficiency of FN resources among the cars is ensured, and we also get a subgame perfect nash equilibrium.

With the advent of the information age in contemporary society, images are everywhere, no matter in military use or in daily life. Therefore, as a medium for people to obtain information, images have become more and more important. With the fast development of deep convolution neural networks (DCNNs), Single-Image Super-Resolution (SISR) becomes one of the techniques that have made great breakthroughs in recent years. In this paper, we give a brief survey on the task of SISR. In general, we introduce the SR problem, some recent SR methods, public benchmark datasets and evaluation metrics. Finally, we conclude by denoting some points that could be further improved in the future.

With the development of artificial intelligence technology, the application of mobile robots is more and more extensive. How to solve the control problem of mobile robots in complex network environment is one of the core problems that plague the promotion and application of AGV clusters. In view of the above problems, this paper studies the control technology in the cloud big data environment, and realizes the decision, planning and control of AGV in the cloud environment. Firstly, cloud-side data sharing and cross-domain collaboration are used to realize intelligent adaptive association of heterogeneous data, then establish a collaborative hierarchical information cloud processing model, and design a new AGV sensing structure based on various devices such as laser radar and ultrasonic sensors. Finally, a set of AGV motion control methods in cloud environment is proposed. The experimental results show that the efficient coordination among network nodes in the heterogeneous AGV system in the cloud environment is stable overall and has a lower delay rate, which will greatly promote the application of AGV in various complex network environments.

As the number of cloud services and user interest data soars, it’s hard for users to find suitable could services within a short time. A suitable cloud service automatic recommendation system can effectively solve this problem. In this work, we propose KGCF, a novel method to recommend users cloud services that meet their needs. We model user-item and item-item bipartite relations in a knowledge graph, and study property-specific user-item relation features from it, which are fed to a collaborative filtering algorithm for Top-N item recommendation. We evaluate the proposed method in terms of Top-N recommendation on the MovieLens 1M dataset, and prove it outperforms numbers of state-of-the-art recommendation systems. In addition, we prove it has well performance in term of long tail recommendation, which means that more kinds cloud services can be recommended to users instead of only hot items.

A semi-supervised classification method for hyperspectral remote sensing images based on convolutional neural network (CNN) and modified tri-training is proposed. The abstract features are captured by training a CNN model with the pixels’ vectors as inputs. Based on the extracted high-level features, different classifiers will perform different outputs under the same training set, due to the different types of classifiers take on diverse characteristics. Thus, taking multiple classifiers’ results into consideration can integrate different prediction labels synthetically from a high level and can perform more credible results. At the meantime, the number of training samples of hyperspectral images is limited, which will hinder the classification effect. Illuminated by tri-training algorithm, we utilize triple different classifiers to classify the hyperspectral images based on the extracted high-level features in the semi-supervised mode. By utilizing triple classifiers jointly to train and update the training samples set when the number of labeled samples is limited. At the meantime, we pick the confident samples via randomize and majority vote into the training set for data editing during the iterative updating process. Experiments performed on two real hyperspectral images reveal that our method performs very well in terms of classification accuracy and effect.

Machine learning is very popular right now. We can apply the knowledge of machine learning to deal with some problems in our daily life. Taxi service provides a convenient way of transportation, especially for those who travel to an unfamiliar place. But there can be a risk that the passenger gets overcharged on the unnecessary mileages. To help the passenger to determine whether the taxi driver has made a detour, we propose a solution which is a cloud-based system and applies machine learning algorithms to detect anomaly taxi trajectory for the passenger. This paper briefly describes the research on several state-of-art detection methods. It also demonstrates the system architecture design in detail and gives the reader a big picture on what parts of the application have been implemented.

The Mobile Edge Computing (MEC) is a new paradigm that can meet the growing computing needs of mobile applications. Terminal devices can transfer tasks to MEC servers nearby to improve the quality of computing. In this paper, we investigate the multi-user computation offloading problem for mobile-edge computing. We study two different computation models, local computing and edge computing. First, we drive the expressions for time delay and energy consumption for local and edge computing. Then, we propose a server partitioning algorithm based on clustering. We propose a task scheduling and offloading algorithm in a multi-users MEC system. We formulate the tasks offloading decision problem as a multi-user game, which always has a Nash equilibrium. Our proposed algorithms are finally verified by numerical results, which show that the scheduling strategy based on clustering can significantly reduce the energy consumption and overhead.

With the development of technology and application of industrial internet of things, a large amount of data is generated in the research and development (R&D) processes in manufacturing domain, including manufacturing procedures, enterprise management, and product transactions. However, these data usually maintained in different departments, which result in information isolation and data with relations cannot be synchronized. This issue leads to the waste of storage space for redundant data and human resources for coordinating essential information. Aiming these problems, we proposed a cloud-based data management platform architecture to collect and maintain the data from isolated domains and distributed departments. A graph database is employed to store the data emphasizing the relations between entities and Master Data Management is deployed to link the entities cross standalone databases. The efficiency of inspecting, managing and updating information across databases shall be improved by the features of the proposed platform.