Path planning is essential for autonomous underwater vehicles (AUVs) to perform tasks. Many existing single-objective path planning methods rely on prior knowledge of the underwater environment. However, extracting prior knowledge is challenging because the underwater environment is influenced by ocean current, complex terrain, and other factors. This paper proposes a two-stage reinforcement learning (RL) algorithm based on trajectory exploration and sequence modeling, called the Soft Actor Critic and Online Decision Transformer (SAC-ODT), which operates without prior knowledge. This algorithm utilizes the SAC strategy exploration capability to generate training data for ODT and uses the ODT strategy optimization capability to plan a smooth, energy-efficient, and safe path. In the first stage, a Multi-Reward Strategy Embedding (MRSE) method is designed to facilitate trajectory exploration with multiple strategies, enabling further training of an ODT with comprehensive decision-making capability. During the second stage, a Condition Prioritized Buffer Update and Sampling Strategy (CP-BUSS) is proposed to enhance the sensitivity of ODT to the reward function, enabling adaptation to various tasks while accelerating high-quality path learning. Experimental results demonstrate that compared to existing RL-based benchmarks, SAC-ODT reduces the path time and energy consumption by 2.7% and 2.5%, respectively, while improving path smoothness by 91.96%.

Tropical cyclones (TCs) are destructive weather systems, and it is crucial to accurately locating their positions, estimating their intensity, and predicting their trajectories to reduce their potential damage. Due to their excellent ability to extract deep spatial features from satellite images, machine learningbased methods can fully mine the relationships between image features and TC positions, which have become the mainstream approach in the field of TC center location. However, the existing methods mainly rely on the eye features identified from infrared channels of satellite images, which can be an invalid approach during the TC generation or extinction phase as the eye structures in these two phases are blurred. To address this problem, this study examines the relationship between the TC center and the low geopotential height center and proposes a non-eyed TC center location method by fusing the features from satellite images and geopotential height data. In the proposed method, a water vapor attention module is used to focus on regions with water vapor distribution patterns closely related to the TC center. Moreover, a multi-scale attention module is designed to mine and fuse knowledge from different spatial scales, aiming to extract the TC structural features fully. The performance of the proposed model is verified through experiments using Gridsat, IBTrACS and ECMWF HRES datasets. The results show that compared with the existing methods, the mean distance error of the TC’s center location is reduced by at least 11.3% compared with state-of-the-art prediction models.

Efficient oceanic mesoscale eddy (ME) trajectory prediction can help us better understand energy transport in the ocean while also ensuring maritime safety. However, the complex, nonlinear motion of MEs presents significant challenges. Existing data-driven methods are primarily focused on single-trajectory prediction, ignoring interactions between multiple MEs, which are the most important factors influencing trajectory evolution. This study addresses two major research limitations: a lack of methodologies that explicitly include multi-ME interactions, as well as a lack of quantitative assessments of their impact on prediction accuracy. To fill these gaps, this study proposes a spatial projection inference network for ME trajectory prediction (Eddy-SPIN). Eddy-SPIN uses a specially designed projection method based on ME intrinsic features to transform multiple 1-dimensional time-series trajectory data into 2-dimensional Gaussian distribution maps. Then, a spatiotemporal prediction network is designed to generate future Gaussian maps, which are then inverse-projected onto trajectories. Extensive evaluations using data from the South China Sea show that Eddy-SPIN reduces the 7-day cumulative mean geodesic distance error by at least 31.9% and 53.3% when compared to the state-of-the-art single-trajectory method and numerical model, respectively. Further, additional experiments conducted across four dynamically distinct oceanic regions consistently confirm the generalizability and robustness of Eddy-SPIN. The proposed Eddy-SPIN verifies the importance of interactions in the ME trajectory prediction problem and presents a novel predictive model that differs from previous works.

Matting is a fundamental computer vision problem, which has wide applications from daily life to professional fields. To get more precise matting result, we propose a deep learning based algorithm. The network called very deep residual network (VDRN). The first stage is designed to capture entire foreground object by an improved encoder-decoder architecture. It consists of a deep residual encoder and a sophisticated decoder. The second stage is a fully residual convolutional network used for recovering fine structure like hair. Experimental results show our algorithm can tackle complicated foreground textures even it has similar color with background. We evaluate our algorithm on alphamatting.com online benchmark, and Composition-1k dataset. The results demonstrate our method outperforms previous methods, especially in tackling fine structure.

This paper studies power allocation optimization for D2D-underlay enabled cellular systems where only channel fading information is known at the eNB. Unlike instantaneous channel state information (CSI) assumed in many existing works, statistical (partial) CSI is more readily available from long term practical measurement. Within the partial CSI framework, we study the problem of maximizing ergodic weighted D2D sum rate under cellular outage constraints. While the ergodic sum rate maximization problem is NP-hard for interference channels, the outage constraints make the problem even more challenging. We propose a solution based on the framework of monotonic optimization (MO) by identifying the monotonicity in both rate and outage functions. We apply the branch-reduce-and-bound (BRB) algorithm to solve the MO problem and improve its numerical efficiency. We present simulation results to show the feasibility of proposed solution under various channel models.

3GPP has recently accelerated the standardization process for Device-to-Device communications. Sidelink, which refers to the radio link between two user equipments (UEs), is a major update in 3GPP Release 12. In this paper, we will first introduce sidelink related structures and procedures in Release 12. Based on basic LTE signaling schemes, we propose methods of using demodulation reference signal (DMRS), which is used as reference signal in both uplink and sidelink, for advance D2D neighbor discovery and resource scheduling of D2D transmissions. Our study shows that reference signal of normal LTE channels can be well exploited to facilitate the newly proposed sidelink communications.

In cellular communication systems with optional device-to-device (D2D) links, user equipments (UEs) can operate in either D2D mode or cellular mode for data transport. This work introduces mixed-mode D2D communication in which D2D links can operate in multiple modes through resource multiplexing. Within this framework, we study the problem of maximizing weighted D2D sum rate under cellular rate constraints by optimizing mixed-mode allocation and resource allocation in term of transmit power and subchannel assignment. Due to nonconvex cellular rate constraints and binary constraints of subchannel allocation, this problem is a nonconvex mixed-integer problem that is generally difficult to solve. We propose a two-step approach by introducing energy-splitting variables such that mixed-mode allocation and resource allocation can be decoupled and optimized independently. The resulting algorithm can be distributive, requires little signaling overhead, and has low computational complexity. We present extensive numerical results to demonstrate the practicality of our proposed algorithm with regard to various network parameters.

In Device-to-Device (D2D) enabled cellular networks, user pairs can select either direct mode or regular cellular mode for data transport. The optimum choice varies depending on power constraints, channel conditions, and QoS requirements. This work proposes a mixed-mode resource allocation strategy through which user pairs can multiplex different modes with disparate resource fraction and transmission power. We study the problem of maximizing D2D sum rate under cellular rate constraint. Successive convex approximation (SCA) algorithm is applied to solve the resulting non-convex joint optimization problem. Our theoretical analysis indicates that mixed-mode achieves superior performance over commonly adopted single mode selection. We further show the advantages of mixed-mode resource allocation in improving cellular rate constrained D2D link performance through extensive numerical tests.

This work studies constrained rate maximization problem for clustered device-to-device (D2D) communication in cellular networks. In the cluster, only one capable user equipment (cluster head) downloads data from the base station. D2D communication is used for local data sharing between the cluster head and other user equipments in the cluster. When sharing and reusing cellular spectrum resources, D2D links must control the potential impact of mutual interference. In a scenario that D2D is only allowed by eNB when the cochannel cellular QoS constraint can be achieved, we investigate joint precoding strategy of D2D and cochannel cellular transmission subject to the cellular outage probability constraint. Based on estimated cellular channel state information (CSI), we analyze the feasibility and optimality of the constrained rate maximization problem and provide optimized solutions for different scenarios. Our simulation demonstrates the advantage of the proposed precoding strategy in satisfying cellular QoS and improving D2D rate.

As more advanced multi-antenna user equipment becomes widespread, multi-input-multi-output (MIMO) Device-to-Device (D2D) communications can provide substantial rate improvement when properly designed and implemented. This work studies MIMO precoding schemes for D2D and cellular downlink when D2D links reuse the channel resources of cellular downlink. Due to the co-channel interference resulting from shared resources, the selection of D2D precoder and cellular DL precoder become highly inter-dependent. We propose various precoding schemes in the distributed system where users only have their own channel information. These schemes enable cooperative precoding via signaling in the cellular context. We provide extensive simulations for different precoding schemes and illustrate their performance and overhead tradeoffs.

A cognitive network refers to the one where two overlaid structures, called primary and secondary networks coexist. The primary network consists of primary nodes who are licensed spectrum users while the secondary network comprises unauthorized users that have to access the licensed spectrum opportunistically. In this paper, we study the percolation degree of the secondary network to achieve k-percolation in large scale cognitive radio networks. The percolation degree is defined as the number of nearest neighbors for each secondary user when there are at least k vertex-disjoint paths existing between any two secondary relays in the percolated cluster. The percolated cluster is formed when there are an infinite number of mutually connected secondary users spanning the whole network. Each user in the cluster is possibly connected to several neighbors, inducing more communication links between any two of them. Since nodes located near the boundary have fewer neighbors, the boundary effect becomes a bottleneck in determining the percolation degree. For cognitive networks, when the primary node density becomes considerably large, the boundary effect spreads inside the network. The transmission area of most secondary users who are located near the primary nodes decreases due to the restriction of the primary network. Therefore, to ensure k-connectivity in the percolated cluster, each secondary user must be connected to more neighbors, and the percolation degree of the secondary network yields a function of the primary node density. We specify the relationship into three regimes regarding the topology variation of the cognitive network. A closed-form expression of the percolation degree under different primary node densities is presented. The expression characterizes the connectivity strength in the secondary percolated cluster, therefore providing analytical insight on fault tolerance improvement in cognitive networks.

In this paper, we investigate the multicast capacity for static network with heterogeneous clusters. We study the effect of heterogeneities on the achievable capacity from two aspects, including heterogeneous cluster traffic (HCT) and heterogeneous cluster size (HCS). HCT means cluster clients are more likely to appear near the cluster head, instead of being uniformly distributed across the network and HCS means each cluster is also not equal in size as most prior literatures assume. Both of these two properties are commonly found in realistic networks. For this class of networks, we find that HCT increases network capacity for all the clusters and HCS does not influence the network capacity. Our work can generalize various results obtained under non-heterogeneous networks in the literature.