Search

Temporal sentence grounding aims to ground a query sentence into a specific segment of the video. Previous methods follow the common equally-spaced frame selection mechanism for appearance and motion modeling, which fails to consider redundant and distracting visual information. There is also no guarantee that all meaningful frames can be obtained. Moreover, this task needs to detect the location clues precisely from both spatial and temporal dimensions, but the relationship between spatial-temporal semantic information and query sentence is still unexplored in existing methods. Inspired by human thinking patterns, we propose a Coarse-to-Fine Spatial-Temporal Relationship Inference (CFSTRI) network to progressively localize fine-grained activity segments. Firstly, we present a coarse-grained crucial frame selection module, where the query-guided local difference context modeling from adjacent frames helps discriminate all the coarse boundary locations relevant to the sentence semantics, and the soft assignment vector of locally aggregated descriptors are employed to enhance the representation of selected frames. Then, we develop a fine-grained spatial-temporal relationship matching module to refine the coarse boundaries, which disentangles the spatial and temporal semantic information from query sentence to guide the excavation of visual grounding clues of corresponding dimensions. Furthermore, we devise a gated graph convolution network to incorporate the spatial-temporal semantic information by leveraging a gate operation to highlight frames referred to by the query sentence from spatial and temporal dimensions, and propagate fused information on the graph. Extensive experiments on two benchmark datasets demonstrate that our CFSTRI significantly outperforms most state-of-the-art methods.

We develop algorithms to control the scope of an infection spread on a network by allocating a fixed immunization budget to edges of the graph. We assume that the infection propagates according to an independent cascade model and interventions operate by reducing the propensities of edges to transmit the infection. We formulate this problem as a constrained min-max optimization problem with respect to the placements of interventions and the location of the worst-case seed nodes. However, the result is a challenging bilevel mixed integer optimization problem. Furthermore, gradients of the objective function with respect to the continuous variables are unavailable in closed form. We employ tools from derivative-free optimization and stochastic optimization to optimize the objective by iterating between the outer minimization and inner maximization problems. In the inner loop, we use a weighted degree discount (WDD) method to select the seed set of the influence maximization problem. In the outer loop, we utilize two methods: a sample-based simultaneous perturbation Nelder-Mead (SBSP-NM) algorithm and a simultaneous perturbation stochastic approximation (SPSA) algorithm. We perform simulations on synthetic graphs and three larger-scale real-world datasets and illustrate the computational feasibility of our algorithms and their efficacy at controlling epidemic spreads.

The integration of massive multiple-input and multiple-output (MIMO) enabled heterogeneous cellular networks (HCNs) and Internet of Things (IoT) is a promising treatment for future networking paradigms. In such networks, the human-to-human (H2H) devices and IoT devices have some distinct service requirements. Specifically, the former mainly concentrates on downlink throughput, but the latter pays more attention to uplink power consumption. Based on this, we design a device association mechanism (scheme) to achieve a tradeoff between these two performance metrics under devices' association requirements. Through some appropriate adjustments of weighting parameters, different types of devices can improve their performance metrics of interest. At last, such a scheme is formulated as a network-wide logarithmic utility maximization problem in a nonlinear and combinatorial form. To solve it, we develop two types of association algorithms, whose primary difference lies in the treatment of weighting parameters. Then, we show the corresponding convergence and computation complexity analyses for them. Finally, we investigate the impacts of weighting parameters, the BS power and the number of antennas on some certain performance metrics of designed algorithms and other existing one. The simulation results show that the designed algorithms can meet different devices' association requirements by properly adjusting weighting parameters.

Millimeter-wave (mmWave) cellular systems leverage the hybrid analog-digital structure to balance the costs and the capacity performance by using fewer Radio Frequency (RF) chains and the spatial sparsity of the channel. Besides, a more broadband system is the natural outgrowth of the gigantic frequency resource in mmWave band. However, it becomes precarious when the hybrid structure is extended to the wideband systems as the wideband phase shifters are used in the array. In this article, a multi-wideband planar array structure using the concept of Sierpinski carpet fractal is proposed, which exploits the beam squinting property instead of forming several independent beams through an existing hybrid structure. Moreover, multi-subarray can form a larger antenna array without additional costs of phase shifters for the wider range of values. First, we derive several properties of the steering beams from a wideband linear array. Then, we exploit the beam squinting property of the planar arrays employing wideband ideal phase shifters to form multi-beam. The functionality of the phase shifters in an array is simplified by a set of wideband cells connected in series. The maximum cell number needed in the array is derived, and during this derivation, we find a flaw of the beam squinting method. Finally, we present numerical results on the performance of the proposed arrays with the proposed algorithm, based on a cellular scenario of one base station (BS) and multi-user, where the BS has four subarrays and each user has one subarray.

Satellite communication plays an important role in future wireless network. In this article, we investigate the outage performance of the integrated satellite-terrestrial relay network (ISTRN) with relay selection based adaptive decode-and-forward (RSADF), where the relay selection is based on outdated channel state information (CSI), and the multiple-antenna relays conduct maximum-ratio receiving and transmitting for decoding and forwarding, respectively. Particularly, we first obtain a novel closed-form expression for the outage probability of the considered ISTRN with RSADF. Furthermore, to get further insights of the effect of key system parameters, e.g., the number of relays, the number of antennas, the CSI outdating level, and the channel fading characteristics, on the system performance, we conduct asymptotic analysis for three typical high average signal-to-noise-ratio (SNR) scenarios, i.e., 1) all satellite and terrestrial links are with high average SNR, 2) the satellite-to-relay link and the relay-to-user link are with high average SNR, and 3) only the relay-to-user link is with high average SNR. The derived high SNR approximations of outage probability explicitly reveal the impact of key system parameters in the form of diversity gain and coding gain. Finally, numerical simulations are presented to validate the analytical results.

In this paper, we investigate a secure transmission algorithm exploiting the vertical domain for simultaneous wireless information and power transfer (SWIPT) in a massive multiple-input multiple-output system over Rician fading channels. Each user adopts the power splitting (PS) technique to separate information and energy, which is regarded as a potential eavesdropper (ED) for the intended user. We consider the scenario, where at least one ED has the same angle of departure with the intended user. The base station (BS) applies artificial noise (AN) and deploys 3-D directional antennas to enhance the secure rate and energy transfer by adjusting the antenna tilt adaptively. We first derive the approximate ergodic achievable secrecy rate and harvested power. Based on these approximations, the optimization problem is formulated to minimize the total transmit power with individual secrecy rate and harvested power targets. We propose an iterative algorithm to jointly optimizing the BS antenna tilt, AN covariance, power allocation, and PS ratios. The numerical results verify that the approximate secrecy rate is very close to the actual values and show that the performance of the proposal approaches that of the optimal solutions from the brute-force search and outperforms the 2-D and conventional 3-D schemes, which proves the significant role of tilt adaptation in the secure SWIPT.

According to the limited communication distance and the low transmission efficiency for massive electronic tags/sensors of traditional or bistatic backscatter technology, the BackFi transmission protocol design with energy harvesting of massive smart multi-hop electronic tags/sensors was investigated in the ambient backscatter communication systems. This paper proposed the large-scale tags/sensors multi-hop communication protocols with energy harvesting under Wi-Fi architecture, where the “CTS-to-self” frame can be used to set the traditional AP/STA to the sensing status without sending, and then, the new defined frame “Multi-RTS” may send the trigger signal to massive tags or sensors with its frame field as “M-RTS” strategy, compressed bitmap, tags/sensors number, and other messages. The tags/sensors shall respond to the sounding signal and give the backscatter signal to AP using energy harvesting. It can be mentioned that the Wi-Fi standard in this paper is based on the IEEE 802.11ah. Finally, AP can collect and modulate the backscatter signal after massive tags/sensors response to the reader. It was worth noting that the multi-hop relay adopts the power splitting way to implement the communication process, where the object function was energy efficiency maximization with the parameters optimization of transmitting power and relay scaling factor. In order to solve the non-convex problem, this paper derived the approximate expression in the high SNR case and achieved the optimal solution via the Lambert W function. The simulation and analysis verified the effectiveness of the proposed schemes and the correctness of the derivations.

In this paper, for the heterogeneous cellular networks (HCNs) with non-orthogonal multiple access (NOMA), we mainly concentrate on the design of green base station (BS) assignment integrating with power allocation. Note that the power allocation is mainly used for mitigating network interference and reducing energy consumption. To attain the goal of green communications, we take account of an optimization problem with maximizing the whole energy efficiency under users' signal-to-interference-plus-noise ratio (SINR) constraints, where such efficiency represents the ratio of the total throughput of users to the total power consumption of BSs selected by them. To solve the finally formulated problem in a nonconvex and nonlinear form, we first introduce an auxiliary parameter to convert it into a tractable one whose optimal solution can be found by employing an effective iterative algorithm. After that, the whole problem, being in a coupling form, can be broken into the BS assignment and power allocation subproblems that can be optimized alternately. To solve the former, a type of highly effective algorithm is designed according to a coalitional game. In addition, an approach based on a two-sided scalable (2.s.s.) function is developed for the latter. Then, we reveal the properties of these proposed algorithms, including computation complexity and convergence. In the simulation, we mainly investigate the impacts of densities of small BSs and users on the different performance indices of the designed and existing algorithms.

This paper investigates a novel architecture of coordinated multi-point (CoMP) transmission in the cognitive satellite-terrestrial network associated with an unmanned aerial vehicle (UAV). We consider the downlink communication where the UAV and base station (BS) cooperatively serve the terrestrial user by sharing the licensed satellite network spectrum. The goal of this paper is to maximize the achievable rate of the terrestrial user by jointly optimizing BS/UAV transmit power allocation and UAV trajectory, subjecting to the interference temperature threshold imposed on satellite network as well as UAV mobility constraint. The formulated problem is shown in a complicated non-convex form, which is hard to tackle. To this end, we decompose the original problem into two sub-problems, and the block coordinate descent method is employed to solve the two sub-problems alternately. Specifically, for the first sub-problem, the optimal BS/UAV transmit power is obtained with a given UAV trajectory by using the Lagrangian dual method. For the second sub-problem, the UAV trajectory is attained with a given BS/UAV power allocation by using the successive convex approximation technique. Simulation results show that our proposed joint CoMP transmission scheme significantly improves the terrestrial network throughput compared with other benchmark schemes.

To fully utilize the system resources and enhance the user experience, two types of offloading mechanisms are designed for massive MIMO (multiple-input-multiple-out) enabled heterogeneous cellular networks (HCNs). Such mechanisms can achieve a tradeoff between energy efficiency (EE) and spectral efficiency (SE) experiences, but also have a difference on whether or not a power coordination measure is mentioned. At last, they are mathematically written as the network-wide utility maximization problems that are closely related to the SEs and EEs of associated users. In them, we introduce a crucial parameter α to adjust the EE-SE preference. It is noteworthy that the finally formulated problems are in some relatively complicated forms. To solve them, some necessary changes should be made at first, and then we can design some feasible algorithms. Specifically, we try to design a distributed algorithm for a mere offloading problem using dual decomposition (DD), and develop a two-layer iterative algorithm for the joint power coordination and offloading problem using DD and two-sided scalable (2.s.s.) function update. Regarding such these algorithms, we show the corresponding computation complexity and convergence analyses. In the simulation, we mainly investigate different network parameters on an EE-SE tradeoff for our advocated mechanisms and another existing mechanism.

In recommendation systems (RSs), nowadays, not only the traditional user-item rating matrices but also more additional information like contents, contexts, trust friends and other auxiliary information are available to enhance the performance of RS, leading to content-aware, context-aware, trust-aware RS, etc. Thus, it provides much potential to take into consideration the additional information in RS. Hence, we focus on a general low-rank matrix factorization (LRMF) model with similarity constraints and propose a decentralized algorithm based on alternating direction method of multipliers (ADMM) to relieve the computation burden in each server while preserving privacy. What's more, we utilize low-complexity skills in numerical analysis to reduce the computational complexity, based on the exploitation of the special form of the problem. Finally, simulations are performed to validate the effectiveness of our algorithms.

With the development of commercialized autonomous vehicles (AVs), the interaction between AVs and human-driven vehicles has become increasingly important. Nevertheless, on the one hand, complex driver behaviors like distraction are hard to detect by AVs, which may lead to traffic accidents because of the late alert to the following vehicles. On the other hand, advanced techniques such as the real-time image or video processing and vehicle-to-vehicle (V2V) communications make it possible to let AVs receive monitoring signals from nearby vehicles, predict the latent risks, and make smart control to avoid the vehicles driven by distracted drivers. Hence, in this paper, we envisage a collaborative framework integrating human driver distraction monitoring, V2V communications, and AV velocity control. Then, we design the smart velocity control of AVs by taking into consideration the distraction behaviors of the drivers in the human-driven vehicles, and by formulating it as a feasible optimization problem based on model predictive control (MPC) strategies. Furthermore, we analyze the safety benefits that the collaborative framework could help improve on the condition of preserving traffic performance. Finally, we implement the contrast tests of real-time evaluation on driver distraction monitoring based on convolutional neural networks (CNNs) and perform simulations of smart velocity control strategies of the AV at avoiding the distracted driver and reducing rear-end collisions. Through the analysis and the simulations, we show our framework could increase the safety regions, reduce the rear-end collisions, and thus increase the safety of the whole transportation networks.

In this paper, we investigate the design of multiple panel arrays (MPAs) in line of sight (LoS) multiple-input-multiple-output (MIMO) communication systems. We utilize the spherical-wave model for the LoS channel component. The fundamental relations among key design parameters in MPAs LoS MIMO system are studied to obtain the maximum spatial multiplexing gain according to the effective degrees of freedom and channel capacity. We further analyze the performance of MPAs system when the array size and transmission distance are not matched and propose an algorithm to optimize the location of antennas. Numerical results indicate that the proposed algorithm can reduce the computational load compared with the exhaustive approach for the similar system performance. Moreover, the optimized panel antenna array is robust to the geometrical change, misplacement, and signal-to-noise-ratio.

Although the massive MIMO enabled heterogeneous networks (HetNets) can intensify the spectral efficiency benefits, the proper user association and resource allocation are both crucial to achieve desirable performance since the power consumption of base station (BS) scales with the large number of antennas. This paper aims to investigate joint design of user association and power allocation with loads constraint and transmit power constraint for the massive MIMO HetNets by considering proportional fairness about the spectral efficiency under imperfect channel state information. First, we derive a closed-form lower bound on the ergodic spectral efficiency with linear zero-forcing beamforming, based on which, a mixed-integer nonlinear programming problem is formulated. It is difficult to efficiently obtain an exact solution since it is non-convex and combinational. To solve this NP-hard problem, an effective algorithm with guaranteed convergence is proposed, where the original problem is decomposed into the corresponding subproblems, which can be solved by low complexity approaches, respectively. Numerical results show that how the number of antennas and the number of BSs affect the spectral efficiency, and our proposed algorithm outperforms other algorithms in terms of the spectral efficiency and load balancing.

The wideband underwater acoustic multipath channel can be modeled as a multi-scale multi-lag (MSML) channel because signals from different paths might experience different Doppler scales. This brings great challenge to channel parameter estimation. In this paper, we propose a novel algorithm for parameter estimation of MSML channels. This new algorithm is a modified particle swarm optimization (MPSO) algorithm, which can estimate the parameters of the Doppler scale, the time delay, and the amplitude simultaneously for each individual path. Comparing to PSO algorithm, MPSO algorithm uses a multipath list to record positions and fitness values of particles whose fitness values are selected as lbests, and uses these lbests to update particles' velocities at each iteration. As for training sequence, we employ the zero correlation zone sequence which has excellent correlation properties. Computer simulation is used to evaluate the proposed algorithm in comparison with the matching pursuit (MP)-based method and the fractional Fourier transform (FrFT)-based method. Simulation results confirm that the proposed MPSO algorithm outperforms both MP-based method and FrFT-based method in estimation accuracy as well as computation complexity.

To fully exploit both multiplexing gain and array gain of massive multiple-input-multipleoutput (MIMO), the channel state information must be obtained accurately at transmitter side (CSIT). However, conventional channel estimation solutions are not suitable for frequency-division duplexing (FDD) multiuser massive MIMO because of overwhelming pilot and feedback overhead. To reduce the pilot and feedback overhead of channel estimation in FDD systems, we propose a compressive channel estimation scheme for FDD massive MIMO systems in this paper, where the beam-blocked sparsity of massive MIMO channels in beamspace is leveraged. Particularly, we first propose a beam-blocked compressive channel estimation scheme, which can reduce the overhead for downlink training. Then, an optimal block orthogonal matching pursuit algorithm at the BS is proposed to acquire reliable CSIT from the limited number of pilots. Furthermore, an efficient algorithm for channel matrix recovery from separately quantized amplitude and phase of received signals is developed to efficiently decrease feedback load. Simulation results demonstrate that our proposed scheme outperforms conventional solutions.