Search

Smart surfaces as one of the promising enabling technologies for sixth-generation wireless communications (6G) are paving their way. In this article, aiming at radio frequency electromagnetic field compliance assessments of them, two approximate approaches, namely the array-theory approach and the equivalent-current approach, are investigated. Smart surfaces with different beams, sizes, and shapes are designed, assessed, and analyzed. The results are compared with those of a reference approach, which is based on full-wave simulations, in terms of assessment accuracy. On one hand, both the approximate approaches provide significant improvement in assessment time, which is quite useful for compliance assessments in exposure scenarios with large numbers of possible beam states. On the other hand, it is found that the assessment accuracy for the two approximate approaches depends intrinsically on system configurations. Specifically, for a circle-shaped smart surface with a diameter equaling 21 times of the element periodicity, except for extreme situations where scanning angle is far from broadside, overestimated front compliance distances with relative errors no more than 30% and 24% are obtained for the array-theory approach and the equivalent-current approach, respectively. [ABSTRACT FROM AUTHOR]

A novel wideband, multifunctional space-fed planar array antenna is proposed in this article to independently control forward/backward beams. A novel four-layer polarization-dependent unit cell is introduced as the phasing element. A generalized-scattering-matrix (GSM)-based network model is developed to evaluate the element performance and optimize the element structure. Then, a system-level analysis in terms of source feed performance, antenna configuration, and efficiency factors are demonstrated. To verify the design concept, a $13\times 13$ -element ($195 \text {mm}\times 195 \text {mm}$) antenna prototype is designed, fabricated, and tested. The measured realized gains of the reflectarray functionality and transmitarray functionality at 10 GHz are 24 dBi with 18.6% 1 dB fractional gain bandwidth and 23.5 dBi with 22.6% 1 dB fractional gain bandwidth, respectively. Moreover, an additional experiment is implemented to verify the simultaneously bidirectional capability of the proposed antenna. Measured results match well with those of simulation, revealing potential for bidirectional wireless communication applications. [ABSTRACT FROM AUTHOR]

A wideband transmitarray antenna exploiting 1 bit polarization-rotation unit cell is proposed in this article. A printed antipodal fermi tapered slot antenna is introduced as the source feed. Considering the radiation characteristics of the source feed, system-level performances in terms of spillover, illumination, and aperture efficiencies are analyzed, revealing the inherent restriction of wideband operation. In order to obtain wideband performance, a novel three-layer polarization-rotation 1 bit unit cell is developed to introduce phase errors over the aperture, which act as one kind of degree-of-freedom during the system design. An optimization method is employed to further balance the overall phase errors experienced over a wide frequency band, resulting in a flat gain response. To verify the design concept, a $13\times 17$ -element ($195\,\, \text {mm}\times 255 \,\,\text {mm}$) antenna prototype is fabricated and measured with a realized gain of 22.5 dBi at 10 GHz and 1 dB fractional gain bandwidth of 37% achieved, demonstrating great potentials in many communication applications. [ABSTRACT FROM AUTHOR]

This paper presents a numerical method to simulate the particle trajectory in the head–disk interface (HDI). The simulator solves the reduced Navier–Stokes equation with the second-slip boundary conditions to get the air velocity distribution, then solves the motion equations of particles and obtains the trajectories of the particles in the HDI using the binary Lagrange interpolation and the fourth-order Runge–Kutta method. The calculated trajectories of particles follow well with the air streamlines in the recessed regions. The particle density, diameter, and initial entering velocity have different effects on particle movement. Furthermore, particle velocity behaviors in the recessed regions are investigated for the first time. The particle is accelerated or decelerated by the action of the drag force and becomes close to the velocity of local air flow. Such effects are more obvious for small particles than large particles. The particle velocity becomes very small when it is close to the leading edge of the trailing pad. The numerical results also show that the particle goes up when it crosses the transition regions, and the particle moves upward more sharply with higher initial entrance height due to the larger vertical velocity of the local air flow. [ABSTRACT FROM AUTHOR]

With the ongoing development and improvement of deep learning technology, autonomous vehicles (AVs) have made tremendous progress in recent years. Despite its great potential, AV supported by deep learning technology still faces numerous security threats, which prevent AV from being putting into large-scale practice. Aiming at this challenging situation, in this paper, we would like to exploit two attacks against deep learning algorithms in traffic sign recognition system by leveraging particle swarm optimization. Specifically, we first exploit the PAPSO (poisoning attack with particle swarm optimization) which focuses on the training process of the deep learning algorithms in the traffic sign recognition system, i.e., the attacker injects crafted samples into the training dataset, causing a reduction in classification accuracy of the traffic sign recognition system. Then, we also explore the EAPSO (evasion attack with particle swarm optimization) which on the other hand focuses on the interference process of the deep learning algorithms, i.e., the attacker adds some hardly perceptible perturbations to the targeted test sample, leading to a misclassification on it. Extensive experiments are conducted to shed light on the effectiveness of our attacks, and some corresponding defense strategies are also presented. [ABSTRACT FROM AUTHOR]

With the advancement of mobile crowd sensing systems and vehicular ad hoc networks, the human-carried mobile devices (e.g., smartphones, smart navigators, and smart tablets) equipped with a variety of sensors (such as GPS, accelerometer, and compass) can work together to collect sensory data consequently delivered to the cloud for processing purposes, which supports a wide range of promising applications such as traffic monitoring, path planning, and real-time navigation. To ensure the authenticity and privacy of data, privacy-preserving truth discovery has attracted much attention since it can find reliable information among uneven quality of data collected from mobile users, while protecting both the confidentiality of users’ raw sensory data and reliability. However, these methods always incur tremendous overhead and require all participants to keep online for interacting frequently with the cloud server. In this paper, we design an efficient and privacy-preserving truth discovery (EPTD) approach in mobile crowd sensing systems, which can tolerate users offline at any stage, while guaranteeing practical efficiency and accuracy under working process. More notably, our EPTD is the first solution to resolve the problem that users must be online all times during the truth discovery under a single cloud server setting. Moreover, we design a double-masking protocol to ensure the strong security of users’ privacy even if the cloud server colludes with multiple users. Extensive experiments conducted on real-world mobile crowd sensing systems also demonstrate the high performance of our proposed scheme compared with existing models. [ABSTRACT FROM AUTHOR]

The RRAM-based accelerators have become very popular candidates for neural network acceleration due to they perform matrix-vector multiplication in-memory with high storage density and low latency. Many related works have used fixed-precision quantization to achieve the model compression and enhance the tolerance of process variation, but these methods still suffer a large accuracy degradation and poor robustness to nonideal effects. In this work, we propose a crossbar-aware mixed-precision quantization scheme, which enables to search for the optimal precision of each part of the network as a way to improve the accuracy and robustness to noise. First, we introduce a group quantization strategy that can flexibly adjust the group size dynamically according to the crossbar size. Then, we propose a detailed mixed-precision search flow to search for the optimal precision set of the network. Finally, we give a noise injection adaption training method to enhance the tolerance of noise. Experimental results show that our proposed method can improve inference accuracy by at least 2.04% compared to the fixed-precision quantization under the same resource cost. The searched architecture with the highest accuracy most accurate (MA) could achieve an accuracy of 92.39% and a resource saving of 93.30% compared to the full precision model. The searched architecture with the biggest resource savings most efficient (ME) could achieve an accuracy of 91.11% and a resource saving of 95.57% compared to the full precision model. And, the average precision of the ME mixed-precision architecture is only 1.4 bits. Besides, the results show the mixed-precision network with noise adaption training is more robust to noise than the fixed-precision network with noise adaption training. [ABSTRACT FROM AUTHOR]

The scalability issue hinders the deployment of Software-Defined Networking (SDN) in the Wide Area Networks (WANs). Existing solutions have two issues: (1) network performance relies on complicated controller synchronization, which increases the complexity of network control; (2) fine-grained flow processing enables flexible flow control at the cost of high processing load on the controllers and high flow table occupancy on switches. In this paper, we propose a scalable routing solution named HybridFlow, which achieves a good load balancing performance using a single controller with low control overhead (i.e., flow routing and rerouting overhead). HybridFlow mainly employs two techniques: hybrid routing and crucial flow rerouting. Hybrid routing enabled by commercial SDN switches gives us opportunities to reduce the processing load of the controller by routing flows with the hybrid OpenFlow/OSPF mode. Thus, the majority of flows can be routed by OSPF without involving the controller. Crucial flow rerouting realizes load balancing by dynamically identifying crucial flows based on a new metric called Variation Slope and rerouting these flows with the hybrid OpenFlow/OSPF mode. The simulation based on the real traffic traces and network typologies shows that compared with the optimal solution, HybridFlow can achieve 87% of the optimal load balancing performance by rerouting 36% less flows on average. [ABSTRACT FROM AUTHOR]

Image-to-image translation tasks have been widely investigated with Generative Adversarial Networks (GANs). However, existing approaches are mostly designed in an unsupervised manner, while little attention has been paid to domain information within unpaired data. In this article, we treat domain information as explicit supervision and design an unpaired image-to-image translation framework, Domain-supervised GAN (DosGAN), which takes the first step towards the exploration of explicit domain supervision. In contrast to representing domain characteristics using different generators or domain codes, we pre-train a classification network to explicitly classify the domain of an image. After pre-training, this network is used to extract the domain-specific features of each image. Such features, together with the domain-independent features extracted by another encoder (shared across different domains), are used to generate image in target domain. Extensive experiments on multiple facial attribute translation, multiple identity translation, multiple season translation and conditional edges-to-shoes/handbags demonstrate the effectiveness of our method. In addition, we can transfer the domain-specific feature extractor obtained on the Facescrub dataset with domain supervision information to unseen domains, such as faces in the CelebA dataset. We also succeed in achieving conditional translation with any two images in CelebA, while previous models like StarGAN cannot handle this task. [ABSTRACT FROM AUTHOR]

Vehicle type recognition (VTR) is a quite common requirement and one of the key challenges in real surveillance scenarios, such as intelligent traffic and unmanned driving. Usually coarse-grained and fine-grained VTRs are applied in different applications, and the challenge from multiple viewpoints is critical for both cases. In this paper, we propose a feedback-enhancement multi-branch CNN (FM-CNN) to solve the challenge in these two cases. The proposed FM-CNN takes three derivatives of an image as input and leverages the advantages of hierarchical details, feedback enhancement, model average, and stronger robustness to translation and mirroring. A single global cross-entropy loss is insufficient to train such a complex CNN and so we add extra branch losses to enhance feedbacks to each branch. Though reusing pre-trained parameters, we propose a novel parameter update method to adapt FM-CNN to task-specific local visual patterns and global information in new datasets. To test the effectiveness of FM-CNN, we create our own multi-view VTR (MVVTR) data set since there are no such data sets available. And, for fine-grained VTR, we use the CompCars data set. Compared with state-of-the-art classification solutions without special preprocessing, the proposed FM-CNN demonstrates better performance in both coarse-grained and fine-grained scenarios. For coarse-grained VTR, it achieves 94.9% Top-1 accuracy on the MVVTR data set. For fine-grained VTR, it achieves 91.0% Top-1 and 97.8% Top-5 accuracies on the CompCars data set. [ABSTRACT FROM AUTHOR]

Particle intrusion into the head–disk interface (HDI) is becoming a concern in hard disk drives. Numerical studies of particle movement in HDI are of great importance for reducing particle-induced damage. In this investigation, we study the particle rebound mechanism in HDI. Not only the particle–slider collision but also particle–particle collision models are developed, which are divided into three types, including the collisions of particle and surface (P–S), particle and edge (P–E), and particle and particle (P–P). The subsequent response of particles after the collision, which includes rebound and adhesion, is calculated considering the collisional energy loss. The proposed adhesion/rebound model can more accurately present the particle trajectories in HDI and accumulation on the air bearing surface (ABS). We find that particles are likely to accumulate on the leading rail surfaces of the slider due to the air flow drag and Saffman lift. Considering the adhesion and rebound, particles with a larger diameter are more likely to fly away from the slider after the collision. The existence of particle rebound, to a certain degree, reduces contamination on ABS. [ABSTRACT FROM AUTHOR]

A numerical method is presented to study the air-flow pattern and particle trajectory in the head–disk interface (HDI) in this paper. It first solves the generalized steady-state Reynolds equation with slip boundary conditions to get the slider attitude. Then, it solves the reduced Navier–Stokes equation to get the air flow velocity distribution in the HDI. The motion equations of particles are solved to obtain the particle trajectory in the HDI by using the fourth-order Runge–Kutta method. Air flow characteristics and particle behaviors are investigated to illustrate the relationship between the air flow and particle trajectory. The effect of the particle release height is also studied. The simulation results show that particle trajectories basically overlap air flow streamlines in the horizontal plane. The particle moves a longer horizontal distance at a lower particle release height. The difference between the horizontal velocity of the particle and the air flow indicates that the Saffman force points to the disk in the transition region close to the leading edges of leading pads. Moreover, the peak values of the air flow vertical velocity in the previously mention location increase with the reduction of the particle release height. And the air flow vertical velocity along the particle trajectory is almost two times larger than that of the particle at the leading edges of leading pads. The results also show that the drag force of the air flow mainly drives particles up toward the air bearing surface at the leading edges of leading pads, but not the Saffman force. [ABSTRACT FROM PUBLISHER]

The presence of particles, which can intrude into the air bearing, is one of the most common factors in the failure of hard disk drives (HDDs). Previous studies have investigated particles trajectories with the assumption of ideal trapping or reflecting boundary conditions in air-filled drives. However, only the colliding particle with insufficient energy to escape the potential well will be trapped by the surface. In this paper, considering the particle-surface energy during the collision, the trapping criterion of the incident normal critical velocity ( V\text {ni}^{\mathrm {\ast }}) for Al2O3 particles is developed as the boundary conditions for different colliding surfaces inside a 2.5 in drive. Then, trapping status for Al2O3 particles and particles trajectories inside the drive are simulated by using the commercial computational fluid dynamics solver FLUENT with user-defined functions. The results reveal that the particles will travel longer distances until trapped by HDD components when considering the trapping criterion. In addition, smaller particles will more likely degrade the head–disk interface reliability, since they easily stick on the disk surface. [ABSTRACT FROM PUBLISHER]