In this paper, we investigate the problem of minimizing the long-term energy consumption of mobile edge computing (MEC) systems combining reconfigurable intelligent surface (RIS) and non-orthogonal multiple access (NOMA) techniques, based on the dynamic scenarios with sequential task arrival and time-varying channels. We jointly optimize the transmission power, transmission time, offloading ratio, local computing frequency, MEC server’s computing resource allocation, and phase shift of RIS elements. The problem is formulated as a challenging time-dependent non-linear programming problem with a large number of variables. By theoretically deriving the optimal expression of server computing frequency and local computing frequency, we simplify the original problem and decompose it into two nested subproblems, namely, the joint offloading ratio and communication process optimization (JORCPO) subproblem and the RIS’s phase shift optimization (RPSO) subproblem. Then, we propose a two-level approach for the solution. In the outer level, we formulate the RPSO subproblem as a deep reinforcement learning (DRL) problem and adopt the proximal policy optimization (PPO) algorithm to obtain the real-time decision of RIS’s phase shift. In the inner level, based on the given RIS’s phase shift, we first simplify the JORCPO subproblem by deriving the closed-form expression of transmission power and then adopt the Lagrange dual (LD) method to obtain the near-optimal solution of transmission time and offloading ratio. Simulation results testify the performance advantage of the proposed solution.

This article proposes a robust peer-to-peer localization scheme using compact arrays in underwater time-varying channels. The method combines two-way ranging and the time difference of arrival (TDOA) positioning technique, utilizing three elements of the compact array to achieve accurate localization. To better adapt to time-varying channels caused by the inherent changes in the water medium, and improve localization accuracy, this article adopts a TDOA correction method based on channel feature matching using the correlation between the received signals of different elements of the compact array. This method combines line-of-sight with non-line-of-sight information of multipath signals to correct the TDOA. The specific localization steps are as follows. First, multipath features are extracted from the single-frame received signal of an array element through the Savitzky-Golay filter and dynamic time warping (DTW) algorithm. Then, based on the obtained multipath features of multiframe signals, hierarchical clustering is used to extract the channel features from the perspective of the current array element. Finally, the target position is estimated by combining two-way ranging with the TDOA values corrected by DTW from different array elements. The accuracy and stability of the proposed method are validated through simulations and pool experiments by comparing it with the traditional methods.

The converging trends of reinforcement learning (RL) control and cloud-fog automation in industrial cyber-physical systems impose multiple challenges for communications to cope with stringent requirements in latency, reliability, control effectiveness and bifurcating user demands. Progressive goal-oriented (GO) communication is a promising technology to tackle the above challenges. This paper takes a two-step approach to design the first progressive codec of GO communications tailored for RL control tasks. The first step is to design a variable-rate coding scheme that extends the boundaries of rate regimes. This step is achieved by empowering the hierarchical variational autoencoder (HVAE) framework with novel algorithms such as mutual information based soft state abstraction (MISA). The second step is to transform variable-rate encoding into progressive encoding. This is achieved by applying residual-based encoding techniques upon latent representations learned by deep neural networks. Experiments on the Cartpole Swingup task demonstrate that the proposed progressive codec can facilitate smooth transitions from the ultra-low rate regime to regular rate regime, while achieving the state-of-the-art performance in terms of rate-distortion-effectiveness tradeoff.

In this article, a dynamic multiuser multiaccess mobile-edge computing (MEC) network where users split their tasks into multiple parts and offload them concurrently to multiple MEC servers via NOMA is investigated. In order to reduce the long-term energy, a complicated time-dependent nonconvex problem with many deeply coupled variables is formulated. We employ a novel approach that integrates the theoretical derivation and the deep reinforcement learning (DRL) algorithm to solve the problem in real time. In particular, we decompose the considered energy minimization problem into a computational resource and transmission power allocation subproblem as the offloading decision and offloading time of each user are known, and an offloading decision and offloading time allocation subproblem. Then, the first subproblem is decomposed into many independent problems, and we theoretically solve them in parallel and derive the optimal solution. For the second subproblem, in order to obtain the real-time allocation of offloading time and offloading decision, we apply a DRL algorithm that adjusts the reward function with a penalty mechanism while the total offloading ratio constraint is unsatisfied or the transmission power and computational resource allocation subproblem does not have a solution. Via simulations, we verify the optimality of the proposed theoretical derivation-based solutions, and, demonstrate that the system performance can be significantly improved by the proposed approach.

In this article, we aim to minimize the longterm energy consumption of a partial offloading mobile-edge computing (MEC) system with noncompletely overlapping nonorthogonal multiple access (NCO-NOMA) when time-varying channels and continuous tasks arrival are considered. Different from the existing NCO-NOMA-assisted MEC studies, the task data transmission, offloading decision, and task computation in users and the MEC server are jointly optimized. We describe the considered energy minimization problem as a complex nonconvex problem with lots of tightly correlated variables. To solve it, we decompose it into a set of computational and communication resource optimization subproblems and a computational resource optimization subproblem at the MEC server. For the computational and communication resource optimization subproblem, we introduce new variables to transform it into a standard difference of convex functions (DC) programming and propose the concave–convex procedure (CCCP) algorithm to solve it. To find out the real-time solution for the computational resource optimization subproblem at the MEC server, we propose a deep reinforcement learning (DRL) algorithm by adding penalty mechanisms to the reward function. Simulations demonstrate that the proposed solution converges quickly and achieves satisfying performance.

While sensing imagery in space missions has broad applications, the growing image resolution and data volume have caused a major challenge due to limited deep space channel capacities. To address this challenge, semantics-aware image compression becomes a promising direction. This paper is motivated to explore lossy compression at the ultra-low rate regime, which is a deviation from the high-fidelity- oriented tradition. Specifically, we propose an ultra-low rate deep image compression (DIC) codec by synthesizing multiple neural computing techniques such as style generative adversarial network (GAN), inverse GAN mapping, and contrastive disentangled representation learning. In addition, a residual-based progressive encoding framework is proposed to enable smooth transitions from the ultra-low rate regime to near- lossless regime. Experiments on the FFHQ and DOTA dataset demonstrate that compared with existing DICs, the proposed DIC can push the minimum rate boundary by about one order of magnitude while preserving the semantic attributes and maintaining a high perception quality. We further elaborate the design considerations for cross-rate-regime progressive DIC. Our study confirm that a semantically disentangled DIC holds the promise to bridge multiple rate regimes.

In this article, we investigate the long-term energy minimization for nonorthogonal multiple access (NOMA)-based mobile edge computing (MEC) systems with user mobility, continuous tasks arrival, and time-varying channel when the reconfiguration energy cost caused by dynamic voltage frequency scaling (DVFS) technology and the effect of imperfect successive interference cancelation (SIC) decoding in NOMA transmission are taken into account. We formulate the considered problem as a nonconvex optimization problem. To solve it, we decompose it into a computation resource and transmit power optimization subproblem, and an offloading ratio and transmission time optimization subproblem. We first show that when the offloading ratio and transmission time are given, the optimal local CPU frequency, the optimal computation resource allocation in the base station, and the optimal transmit power can be theoretically derived. Then, based on the above theoretical derivation, a soft actor–critic (SAC)-based deep reinforcement learning (DRL) algorithm is proposed to learn the near-optimal offloading ratio and transmission time for users. Simulation results show that the proposed algorithms can significantly improve the system performance.

Progressive deep image compression (DIC) with hybrid contexts is an under-investigated problem that aims to jointly maximize the utility of a compressed image for multiple contexts or tasks under variable rates. In this paper, we consider the contexts of image reconstruction and classification. We propose a DIC framework, called residual-enhanced mask-based progressive generative coding (RMPGC), designed for explicit control of the performance within the rate-distortion-classification-perception (RDCP) trade-off. Three independent mechanisms are introduced to yield a semantically structured latent representation that can support parameterized control of rate and context adaptation. Experimental results show that the proposed RMPGC outperforms a benchmark DIC scheme using the same generative adversarial nets (GANs) backbone in all six metrics related to classification, distortion, and perception. Moreover, RMPGC is a flexible framework that can be applied to different neural network backbones. Some typical implementations are given and shown to outperform the classic BPG codec and four state-of-the-art DIC schemes in classification and perception metrics, with a slight degradation in distortion metrics. Our proposal of a nonlinear-neural-coded and richly structured latent space makes the proposed DIC scheme well suited for image compression in wireless communications, multi-user broadcasting, and multi-tasking applications.

Not only have deep networks become standard in machine learning, they are increasingly of interest in neuroscience as models of cortical computation that capture relationships between structural and functional properties. In addition they are a useful target of theoretical research into the properties of network computation. Deep networks typically have a serial or approximately serial organization across layers, and this is often mirrored in models that purport to represent computation in mammalian brains. There are, however, multiple examples of parallel pathways in mammalian brains. In some cases, such as the mouse, the entire visual system appears arranged in a largely parallel, rather than serial fashion. While these pathways may be formed by differing cost functions that drive different computations, here we present a new mathematical analysis of learning dynamics in networks that have parallel computational pathways driven by the same cost function. We use the approximation of deep linear networks with large hidden layer sizes to show that, as the depth of the parallel pathways increases, different features of the training set (defined by the singular values of the input-output correlation) will typically concentrate in one of the pathways. This result is derived analytically and demonstrated with numerical simulation. Thus, rather than sharing stimulus and task features across multiple pathways, parallel network architectures learn to produce sharply diversified representations with specialized and specific pathways, a mechanism which may hold important consequences for codes in both biological and artificial systems.

Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in mammals, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available.

In this paper, we investigate the resource allocation of mobile edge computing (MEC) in cognitive capacity harvesting networks (CCHNs) when non-orthogonal multiple-access (NOMA) technique is adopted. Different from traditional studies for NOMA-MEC networks, we aim at minimizing the total cost of CCHN while satisfying the quality-of-service (QoS) of secondary users (SUs). We adopt the mechanism of time division multiple access (TDMA) when several NOMA groups use the same spectrum, and consider both the waiting delay and transmission delay during data offloading with the optimization of transmission order of NOMA groups. We formulate the considered problem as a mixed integer non-linear programming (MINLP). We show that the transmit power and the allocated computing resource for each SU can be derived when the transmission time and transmission order of the NOMA groups are given. Based on this, the considered problem can be decomposed into a transmission time and order optimization subproblem, a cellular resource block (CRB) selection subproblem and a cognitive radio (CR) router selection subproblem. To solve the transmission time and order optimization subproblem, we first simplify the delay constraint via theoretic analysis, and then propose a binary segmentation (B-Seg) algorithm and a transmission order adjustment (TOA) algorithm to find the optimal transmission time and transmission order of NOMA groups, respectively. To solve the CRB selection subproblem and the CR router selection subproblem, a bigger requirement first (BRF) algorithm and a game-based iteration (GBI) algorithm are respectively proposed. Simulation results show that the proposed algorithms can significantly improve the system performance.

In this paper, we investigate the energy minimization problem for a cognitive capacity harvesting network (CCHN), where secondary users (SUs) without cognitive radio (CR) capability communicate with CR routers via device-to-device (D2D) transmissions, and CR routers connect with base stations (BSs) via CR links. Different from traditional D2D networks that D2D transmissions share the resource of cellular transmissions in the same cell, we consider the scenario that D2D transmissions share the uplink cellular frequency bands (CFBs) of neighbor cells. To ensure that the transmissions from SUs do not affect the transmissions for the cellular users (CUs) in the neighbor cells, an inter-cell handshake process is proposed. We formulate the energy minimization problem for SUs as a mixed integer non-linear programming (MINLP). To solve this problem, we decompose it into two nested subproblems: a transmit power optimization subproblem and a CR router and uplink CFB selection subproblem. For the first subproblem, it is proved to be convex, and thus can be efficiently solved. For the second subproblem, we propose a two-level nested game theoretic approach to finding its solution. Simulation results show that the proposed algorithms can significantly improve the performance. With the help of CR routers/the neighbor resource sharing, the energy consumption for SUs can be saved around 30%-37% on average.

Partially inspired by features of computation in visual cortex, deep neural networks compute hierarchical representations of their inputs. While these networks have been highly successful in machine learning, it is still unclear to what extent they can aid our understanding of cortical function. Several groups have developed metrics that provide a quantitative comparison between representations computed by networks and representations measured in cortex. At the same time, neuroscience is well into an unprecedented phase of large-scale data collection, as evidenced by projects such as the Allen Brain Observatory. Despite the magnitude of these efforts, in a given experiment only a fraction of units are recorded, limiting the information available about the cortical representation. Moreover, only a finite number of stimuli can be shown to an animal over the course of a realistic experiment. These limitations raise the question of how and whether metrics that compare representations of deep networks are meaningful on these data sets. Here, we empirically quantify the capabilities and limitations of these metrics due to limited image and neuron sample spaces. We find that the comparison procedure is robust to different choices of stimuli set and the level of sub-sampling that one might expect in a large scale brain survey with thousands of neurons. Using these results, we compare the representations measured in the Allen Brain Observatory in response to natural image presentations. We show that the visual cortical areas are relatively high order representations (in that they map to deeper layers of convolutional neural networks). Furthermore, we see evidence of a broad, more parallel organization rather than a sequential hierarchy, with the primary area VisP (V1) being lower order relative to the other areas.

On-demand streaming of high-quality video content is a widely anticipated vehicular infotainment service. How to reduce the cost of content streaming is a primary concern of the service providers, but is still an underinvestigated subject in the literature. This paper proposes an integrated mobile streaming and caching scheme that jointly leverages two communication service tiers and on-board caching resource for cost reduction. Algorithms are presented to achieve optimal buffering at the session level and optimal caching at the device level. An analytical framework is established to characterize the average cost as a function of the streaming rate in a large scale network. Numerical results demonstrate how the “cost-streaming rate” function changes with vehicle density, network congestion level, content length, and average packet transmission time. We learn an important insight that there is a minimum cost threshold even when the streaming rate approaches zero. We also show that the proposed protocol can effectively reduce the overall cost when the network is not congested. Our findings can provide useful guidelines for the business planning and operation of vehicular content streaming services.

Cellular communication networks are plagued with redundant capacity, which results in low utilization and cost-effectiveness of network capital investments. The redundant capacity can be exploited to deliver secondary traffic that is ultra-elastic and delay-tolerant. In this paper, we propose an analytical framework to study the capacity-delay tradeoff of elastic/secondary traffic in large scale cellular networks with spectrum aggregation. Our framework integrates stochastic geometry and queueing theory models and gives analytical insights into the capacity-delay performance in the interference limited regime. Closed-form results are obtained to characterize the mean delay and delay distribution as functions of per user throughput capacity. The impacts of spectrum aggregation, user and base station densities, traffic session payload, and primary traffic dynamics on the capacity-delay tradeoff relationship are investigated. The fundamental capacity limit is derived and its scaling behavior is revealed. Our analysis shows the feasibility of providing secondary communication services over cellular networks and highlights some critical design issues.

Recent literature has suggested the benefits of integrating licensed radio and cognitive radio into a hybrid cooperative communication system. The fundamental properties of such hybrid systems, however, have not been thoroughly investigated. This paper studies the hybrid cognitive Gaussian relay channel (HCGRC), which uses licensed radio resource (RR) and cognitive/unlicensed RR for forward and relay transmissions, respectively. HCGRC fundamentally differs from conventional relay channels in that the licensed and cognitive RRs are not subject to a total resource constraint and that the cognitive RR is opportunistic in nature. With respect to both the upper and lower bounds, we derive the optimal power-bandwidth allocation strategies for the cognitive relay to maximize the capacity, spectrum efficiency (SE), and energy efficiency (EE). The Pareto-optimal EE-SE tradeoff curve is also derived analytically. Our study leads to two key observations. First, the multi-objective power-bandwidth allocation problem is characterized by five regions, each representing a unique performance tradeoff. Second, the reliability of cognitive RR has no impact on the EE-SE tradeoff given unlimited bandwidth and power.

Abstract In this paper, we investigate the carrier sensing range (CSR) of a general 802.11 network with physical‐layer network coding (PNC). We aim to derive a sufficient CSR that can prevent the hidden‐node collisions in a general 802.11 PNC network. The analysis includes two steps. First, we analyze the six link‐to‐link interference cases in an 802.11 PNC network to show that the mutual interference will be most severe when each node in the network initiates a two‐hop end node link. Second, we consider the worst interference case that all concurrently transmitting links in the network are two‐hop end node links and placed in the densest manner and develop a closed‐form expression of a sufficient CSR that prevents the hidden‐node collisions in a PNC network. From the analysis results, we find that to prevent the hidden‐node collisions, the CSR in PNC network should be bigger than the one in traditional non‐network‐coding network. Furthermore, we carry out extensive simulations to find out the throughput gain of PNC scheme in a general wireless network when considering the impact of CSR. Simulation results show that compared with the non‐network‐coding scheme, PNC scheme has throughput gain when a large proportion (i.e., 90%) of links in the network are two‐hop links and the link density has little effect on the throughput gain of PNC scheme. Copyright © 2014 John Wiley & Sons, Ltd.

Motivated by the deployment of cognitive radio (CR) based relays in cellular networks, this paper studies the fundamental limits of heterogeneous cognitive Gaussian relay channels (HCGRCs). Unlike conventional relay channels, in HCGRC a source transmits to a relay in a licensed band, while the relay transmits to a destination in an unlicensed cognitive spectrum band. The licensed and unlicensed bands are characterized by different power, bandwidth and reliability constraints. Taking an information-theoretic perspective, the fundamental properties of the HCGRC are analyzed thoroughly in terms of capacity, spectral efficiency (SE), and energy efficiency (EE). With regard to each metric, we derive the optimal resource allocation strategy and discuss the impacts of CR spectrum reliability and relay location on the metric. We find that in HCGRC, improving the SE and EE are not necessarily conflicting objectives. Instead, both metrics can be optimized simultaneously with proper resource allocation.

Virtual multiple-input multiple-output (V-MIMO) technology promises significant performance enhancements to cellular systems in terms of spectral efficiency (SE) and energy efficiency (EE). How these two conflicting metrics scale up in large cellular V-MIMO networks is unclear. This paper studies the EE-SE trade-off of the uplink of a multi-user cellular V-MIMO system with decode-and-forward type protocols. We first express the trade-off in an implicit function and further derive closed-form formulas of the trade-off in low and high SE regimes. Unlike conventional MIMO systems, the EE-SE trade-off of the V-MIMO system is shown to be susceptible to many factors including protocol design (e.g., resource allocation) and scenario characteristics (e.g., user density). Focusing on the medium and high SE regimes, we propose a heuristic resource allocation algorithm to optimize the EE-SE trade-off. The fundamental performance limits of the optimized V-MIMO system are subsequently investigated and compared with conventional MIMO systems in different scenarios. Numerical results reveal a surprisingly chaotic behavior of V-MIMO systems when the user density scales up. Our analysis indicates that low frequency reuse factor, adaptive resource allocation, and user density control are critical to harness the full benefits of cellular V-MIMO systems.

Geometry based block partitioning (GBP) has been shown to achieve better performance than the tree structure based block partitioning (TSBP) of H.264. However, the residual blocks of GBP mode after motion compensation still present some non-vertical/non-horizontal orientations, and the conventional discrete cosine transform (DCT) may generate many high-frequency coefficients. To solve this problem, in this paper we propose a video coding approach by using GBP and reordering DCT (RDCT) techniques. In the proposed approach, GBP is first applied to partition the macroblocks. Then, before performing DCT, a reordering operation is used to adjust the pixel positions of the residual macroblocks based on the partition information. In this way, the partition information of GBP can be used to represent the reordering information of RDCT, and the bitrate can be reduced. Experimental results show that, compared to H.264/AVC, the proposed method achieves on average 6.38% and 5.69% bitrate reductions at low and high bitrates, respectively.