Search

Network scanner is a key measuring instrument in network optimization for 5G mobile communication. This paper details the design of frequency source for 5G network scanner. The phase noise of frequency source was simulated with ADS. The frequency resource was developed with low phase noise, high frequency resolution, low spurious and small size. The output frequency range is 62.5MHz-32GHz with 1Hz frequency resolution. The phase noise of frequency source is -89.44dBc / Hz (1kHz offset) at the frequency of 28GHz.

In recent years, there have been a lot of interests in incorporating semantics into simultaneous localization and mapping (SLAM) systems. This paper presents an approach to generate an outdoor large-scale 3D dense semantic map based on binocular stereo vision. The inputs to system are stereo color images from a moving vehicle. First, dense 3D space around the vehicle is constructed, and the motion of camera is estimated by visual odometry. Meanwhile, semantic segmentation is performed through the deep learning technology online, and the semantic labels are also used to verify the feature matching in visual odometry. These three processes calculate the motion, depth and semantic label of every pixel in the input views. Then, a voxel conditional random field (CRF) inference is introduced to fuse semantic labels to voxel. After that, we present a method to remove the moving objects by incorporating the semantic labels, which improves the motion segmentation accuracy. The last is to generate the dense 3D semantic map of an urban environment from arbitrary long image sequence. We evaluate our approach on KITTI vision benchmark, and the results show that the proposed method is effective.

This paper presents real-time obstacles detection and their status classification method for collision warning in the vehicle active safety system. Specifically, stereo cameras and millimeter wave (mmw)-radar are fused to help the driving ego-vehicle to find “Danger” or “Potential Danger” in a timely way through combining with the vehicle kinematic model. The proposed method makes full use of the unique advantages of stereo cameras and mmw-radar to sense the environment through several modules. Cameras are mainly used to detect the near or lateral dynamic objects and to obtain the obstacles region of interest (ROI) considering its rich information and high sensitivity to the lateral displacement, while far or longitudinal relative dynamic objects are detected by mmw-radar according to its observational ability to make up for the disadvantage of cameras. In detail, a cameras detector utilizes ”error vectors” rather than the optical flow to obtain dynamic classes through two times clustering. Mmw-radar mainly detects relative dynamic objects, whose absolute speed can be computed according to the ego-vehicle’s state. Then, the detected objects of these two detectors are integrated in an obstacles ROI map, which is obtained through an UV-disparity obstacles detection algorithm to get the final dynamic and relative dynamic objects. Finally, they are classified by comparing them with a dangerous area that is acquired according to the vehicle kinematic model in a special vehicle coordinate system, which is fixed to the ground temporarily. This method is tested on our mobile platforms and the results prove that it can work effectively even though the ego-vehicle drives quickly.

Multiple object tracking (MOT) is an important yet challenging task in video understanding and analysis. Basically, MOT aims to associate detected objects into trajectories based on their temporal relationships. The occlusion among moving objects poses a major challenge towards robust modeling of these relationships. In this paper, we propose a novel Tracklet Siamese Network (TSN) for learning similarities between track-lets characterized by appearance information, achieving superior performance on two MOTChallenge benchmark datasets. Our framework constructs short tracklets from highly-related object detections by excluding inaccurate object detections. We also adopt a constrained clustering technique to piece tracklets together into long trajectories, thus recovering many missing detections caused by original detector or the detection removing in the previous step. Comparisons against state-of-the-art methods were reported while ablation studies further substantiate the viability of components in our approach.

Face recognition plays a critical role in surveillance and security systems. Due to the large appearance variation of human faces, the dissimilarity among faces for the same person may be quite large. This leads to unstable results. To improve the stability and reliability of face recognition, this paper proposes a novel deep-based approach by introducing an adaptively-weighted verification loss function. The proposed loss function can properly enlarge the margin between positive face pairs and negative face pairs from the global perspective, thus obtain a more reliable recognition model by minimizing the dissimilarity between same-person faces and maximizing the dissimilarity between different-person faces. Experiments on the benchmark LFW and YTF datasets demonstrate that the proposed approach can obtain the state-of-the-art performances for face recognition.

Face recognition has attracted much attention from researchers for past decades. Recently, with the development of deep learning, a deep neural network is adopted by face recognition system and better performance is obtained. Many works on metric learning have been done in the deep neural network. Meanwhile, there are several variation problems existing in face recognition, such as profile face image, low-resolution face image, different age of face image, face image wearing eyeglass, etc. In this paper, targeting at different kinds of variation problems, we proposed a novel network structure, called Extended Siamese Network. Another contribution is that a new loss function is proposed, to further take inter-class information into account based on the center loss function. The experiments show that recognition accuracy is improved in comparison with the other state-of-art methods.

Web of Things (WoT) makes it possible to connect tremendous embedded devices to web in Representational State Transfer (REST) style. Some lightweight RESTful protocols have been proposed for the WoT to replace the HTTP protocol running on embedded devices. However, they keep the principal characteristic of the REST style. In particular, they support one-to-one requests in the client–server mode by four standard RESTful methods (GET, PUT, POST, and DELETE). This characteristic is however inconsistent with the practical networks of embedded devices, which typically perform a group operation. In order to meet the requirement of group communication in the WoT, we propose a resource-oriented protocol called SeaHttp to extend the REST style by introducing two new methods, namely BRANCH and COMBINE respectively. SeaHttp supports parallel processing of group requests by means of splitting and merging them. In addition SeaHttp adds spatiotemporal attributes to the standard URI for naming a dynamic request group of physical resource. Experimental results show that SeaHttp can reduce average energy consumption of group communication in the WoT by 18.5%, compared with the Constrained Application Protocol (CoAP).

The optimal rate allocation problem with end-to-end delay constraints in wireless networks is highly challenging due to the dynamics incurred by delay related factors, such as traffic arrival pattern, queuing process, and wireless resource sharing mechanism. In this paper, we solve the problem through a utility maximization framework with two sets of constraints: (1) capacity and schedulability constraints and (2) end-to-end delay constraints. The end-to-end delay of a flow can be adjusted by controlling the per-link delays via a novel parameter called Virtual Link Capacity Margin (VLCM), which is the measurement of the gap between the schedulable link capacity and the maximum allowable flow rate over a link. This optimization problem can be solved via its dual decomposition through two prices derived with regard to the constraints: the link congestion price that reflects the relationship between the traffic load and the capacity of a link, and the flow delay price that reflects the margin between the average packet delay and the delay constraint of a flow. We prove that the algorithm converges to the optimal solution under the M/M/1 queuing model. We have also discussed using our algorithm under generalized traffic patterns, where we use link delay functions to formulate the relationship between VLCM and the average link delay. The algorithm is implemented distributedly via joint wireless link scheduling, VLCM adjustment and congestion control. Extensive experiments are conducted over diverse network topologies and traffic arrival models. The experimental results show that our algorithm is able to achieve optimal rate allocation while satisfying the end-to-end delay constraints.

Inspired by the mechanism of behaviour control of organisms, this paper attempts to build an underlying model of their physical and psychological activities in a dynamic environment with the purpose to achieve the self-adaption of robot behaviour in a complex, changeable environment. It designs an internal environment and emotion- & motive-generating models based upon the endocrine regulation mechanism (ERM). The robot relies upon the dynamic stability of the internal environment to accomplish its self-adaption to the external environment. The emotion-generating model is applied to in vivo and in vitro contingencies of various kinds; by releasing hormones, the emotion generated influences the robot’s in vivo and in vitro perception, and then its behavioural choice. The motive-generating model is employed to maintain the relative stability of the internal environment, and the motive generated leads to the corresponding behavioural choice directly. The experimental results prove the effectiveness of this ERM-based self-adaptive robot model.

A unique challenge in P2P network is that the peer dynamics (departure or failure) cause unavoidable disruption to the downstream peers. While many works have been dedicated to consider fault resilience in peer selection, little understanding is achieved regarding the solvability and solution complexity of this problem from the optimization perspective. To this end, we propose an optimization framework based on the generalized flow theory. Key concepts introduced by this framework include resilience factor, resilience index, and generalized throughput, which collectively model the peer resilience in a probabilistic measure. Under this framework, we divide the domain of optimal peer selection along several dimensions including network topology, overlay organization, and the definition of resilience factor and generalized flow. Within each sub-problem, we focus on studying the problem complexity and finding optimal solutions. Simulation study is also performed to evaluate the effectiveness of our model and performance of the proposed algorithms.

Route discovery in wireless mobile networks requires a considerable amount of resources due to the mobility of the hosts. Therefore, it would be wise to utilize the effort already invested in existing paths. This paper proposes an efficient way to reuse, whenever possible, existing paths when a new path is being established. In our proposed algorithm, called Route Reutilization Routing (RRR), the reusability is accomplished by the notion of the dynamic proactive zones (DPZ), through which nearby existing path information is disseminated. By utilizing the information stored in DPZs, RRR can achieve considerable savings over other on-demand routing algorithms that use flooding. The unique feature of the proposed algorithm is that DPZs are created and destroyed dynamically around the existing paths, whereas proactive zones are formed around the nodes throughout the network in other route finding algorithms. Even though using DPZs may not result in the shortest path between source and destination, simulation results show the considerable reduction in traffic needed to find a path and therefore increases the available bandwidth for data transmission.

Traditional wireless sensor networks (WSNs) can be integrated into Internet and be regarded as its sensing infrastructure, which supports development and running of multiple third-party applications simultaneously. Therefore, due to constrained resource of sensor nodes, it is necessary to establish a runtime framework to improve sensor sharing efficiency for concurrent third-party applications. This paper presents EasiCAE, a concurrent applications runtime framework, to enhance sensor sharing efficiency greatly by incorporating task allocation with redundancy elimination. In brief, EasiCAE decompose the applications into tasks and distributes tasks to the sensors which will bring the least energy to run them. EasiCAE has three salient features. Firstly, we define task-sensor correlation to indicate how many samplings of a sensor can be shared with the new task. Secondly, EasiCAE reduces energy consumption by assigning tasks to a sensor with higher task-sensor correlation. Finally, a light-weight merging algorithm is proposed to eliminate redundant samplings for the assigned sensors. Experimental results show that EasiCAE reduces energy consumption by 31% to 79% compared with existing methods, while introducing tolerable overheads. We also evaluate EasiCAE with various influencing parameters, showing that the performance of EasiCAE increases stably as the network scale and the number of concurrent applications increases.

Energy-efficient wireless reprogramming is key issues for long-lived sensor network. Most of wireless reprogramming approaches focus on the energy efficiency of the data transmission phase. However, the program rebuilding phase on target node is possibly as another significant part of the total reprogramming energy consumption, due to the high energy overhead of reading or writing operation on the energy-hungry nonvolatile memory. In this paper, we propose an energy-efficient reprogramming system—EasiLIR. The core of EasiLIR is to avoid r/w operations on nonvolatile memory as much as possible in two fold. Firstly, we design an in situ modification which creates a modified program equivalent to new one without rebuilding program. However, at the cost of no rebuilding program, the redundant binary codes existing in the modified program may break the program time constraint. Therefore, we also design a lightweight segmented rebuilding to directly create the new image in memory. Experiment results show that EasiLIR reduces the r/w operations on nonvolatile memory by approximately 88% and 81% compared to Deluge and R2, and its average reprogramming overhead is about 64.7% of R2.

Joint congestion control and scheduling has been widely applied to resource allocation in multi-hop wireless networks. Existing algorithms usually rely on seamless cooperation between network components to seek the optimal solution. However, this is unrealistic considering the presence of time delay and channel capacity perturbation in real wireless environments. The stability and efficiency of these cross-layer control algorithms face significant challenges. In this paper we present ROCS: a robust joint congestion control and scheduling algorithm for time-varying multihop wireless networks with time delay, to bridge the gap between the existing results and the reality of wireless networks. The fundamental idea behind ROCS is Capacity Space Mapping, which combines the slow time scale part of the original channel capacity and a margin estimated from the fast time scale part, to form a new capacity space. The cross-layer resource allocation problem is formulated into a utility maximization framework over the newly generated capacity space. The problem is solved by a control algorithm which contains two closely coupled components: Link Scheduling and Congestion Control. Link Scheduling coordinates wireless link utilization on a maximum weight basis, and Congestion Control allocates flow rates according to congestion feedback information. We further derive the sufficient conditions for the Lyapunov stability of the control algorithm with time delay, following the Razumikhin theorem. We demonstrate via extensive experiments conducted over simulated and real world wireless traces that ROCS substantially achieves robustness and efficiency.

This paper investigates the optimal rate allocation problem with end-to-end delay constraints in multi-hop wireless networks. We introduce Virtual Link Capacity Margin (VLCM), which is the gap between the schedulable link capacity and the maximum allowable flow rate over a link, for link delay control. We formulate the problem as a utility maximization framework with two sets of constraints: 1) capacity and schedulability constraints and 2) end-to-end delay constraints. By dual decomposition of the original optimization problem, we present a control algorithm that jointly tunes the flow rates and VLCMs, through a double-price scheme derived with regard to the constraints: the link congestion price reflecting the traffic load of a link, and the flow delay price reflecting the margin between the average packet delay and the delay requirement of a flow. We prove that the algorithm converges to a global optimum where the aggregate network utility defined over the flow rate set is maximized, while the delay constraints are satisfied. A key feature of our algorithm is it does not rely on a specific traffic model. The algorithm is implemented distributedly via joint wireless link scheduling and congestion control. Simulation results show that our algorithm outperforms heuristic rate allocation algorithms while satisfying the end-to-end delay constraints.

Integrating Peer-to-Peer streaming with wireless networks is a promising technique which promotes ubiquitous multimedia services. Rate allocation within this context is essential to the overall performance of a streaming system due to the limitation of wireless network resource. Optimizing rate allocation in wireless P2P streaming poses us the twofold challenge: the data flow rates are not only constrained by P2P overlay uplink capacity, but subject to the resource sharing mechanism of wireless networks. In this paper, we propose a utility-based rate allocation framework to optimize the flow rates in wireless P2P streaming systems. Our optimization framework involves both P2P overlay constraint and wireless channel sharing constraint. We propose a channel-aware approach to address the modeling of wireless channel sharing patterns. We then proceed to design a distributed double pricing algorithm to solve the optimization problem. Experimental results validate that our algorithm can achieve convergence of data flow rate allocation and optimality of network throughput.

In this paper we propose a quantitative model to study user satisfaction level in online video streaming, which features great variety across video access sessions under the impact of network conditions and user-specific factors. By applying survival analysis to video session duration ratio, which in our study is regarded to be a measure of user satisfaction level, we find user satisfaction is not only related to network QoS factors, such as buffer time and buffer count, but also susceptible to the users' inclination to the contents of videos. Based on Cox regression model, we analyze the quantitative relationship between user satisfaction level in video streaming and several predictor variables such as buffer count rate, buffer time rate and video popularity degree. Experimental results prove that the model can accurately evaluate user satisfaction levels of our test video access sessions with high probability.

Understanding user behavior in online video streaming is essential to designing streaming systems which provide user-oriented service. However, it is challenging to gain insightful knowledge of the characteristics of user behavior due to its high volatility. To this end, the paper provides an extensive analysis of user behavior in online video streaming, based on a large scale trace database of online streaming video access sessions. We categorize user behaviors into multiple patterns and probe the relationship between them. Our work puts emphasis on the statistical characteristics of user behavior patterns. Particularly, this study uncovers that the behavior of one individual user in a video streaming session is not only related to the popularity level of the video, but also has strong correlation with the user's behaviors in previous streaming sessions.

Efficiency has been concerned in the research of association rules mining. This paper presents an improved method called Direct-Fined-Remove (DFR) algorithm to mine a database consisting of remove and direct steps. When pruning the candidates itemsets, the algorithm eliminate non-frequent subset of candidates in the remove steps. In the direct steps, the algorithm directly generates the frequent itemsets by computing and comparing the frequency of frequent k-itemsts with k in the meantime. The contributions include: (1) proposes an algorithm to raise the probability of obtaining information in scanning database and reduce the potential scale of itemsets. (2) successfully applies the proposed algorithm to the Teaching Evaluation System to determine characteristics of good teaching effect. Experiments show that the probability which grade point of associate professor being more than 90 is 35% the support degree is 40% and the probability which grade point of bachelor being more than 90 is 16% the support degree is 83%.