ABSTRACT Shadow detection helps reduce ambiguity in object detection and tracking. However, existing shadow detection methods tend to misidentify complex shadows and their similar patterns, such as soft shadow regions and shadow‐like regions, since they treat all cases equally, leading to an incomplete structure of the detected shadow regions. To alleviate this issue, we propose a structure‐aware transformer network (STNet) for robust shadow detection. Specifically, we first develop a transformer‐based shadow detection network to learn significant contextual information interactions. To this end, a context‐aware enhancement (CaE) block is also introduced into the backbone to expand the receptive field, thus enhancing semantic interaction. Then, we design an edge‐guided multi‐task learning framework to produce intermediate and main predictions with a rich structure. By fusing these two complementary predictions, we can obtain an edge‐preserving refined shadow map. Finally, we introduce an auxiliary semantic‐aware learning to overcome the interference from complex scenes, which facilitates the model to perceive shadow and non‐shadow regions using a semantic affinity loss. By doing these, we can predict high‐quality shadow maps in different scenarios. Experimental results demonstrate that our method reduces the balance error rate (BER) by 4.53%, 2.54%, and 3.49% compared to state‐of‐the‐art (SOTA) methods on the benchmark datasets SBU, ISTD, and UCF, respectively.

An important use case in fifth generation systems are vehicular applications, where, reliability and low latency are the main requirements. In order to determine if a vehicular application can be used one can apply machine learning (ML) tools to determine if these constraints are met, which open questions such as “which data is available”, “which features to use”, “the quality of this prediction”, etc. In this paper we address some aspects of predicting quality-of-service (QoS) in a cellular vehicular-to-everything scenario, where we employ supervised learning as well as the autoregressive integrated moving average filter to predict if a packet can be delivered within a desired latency window. Particularly, we are interested in the reliability of this prediction, including predicting if a packet generated some time ahead will be delivered in time. Such information is essential when asserting that a vehicular application can indeed be employed safely. We show via simulation results that ML can be used as a prediction tool in vehicular applications. For instance, QoS levels can be predicted two seconds ahead with 85 % reliability.

Heterogeneity among patients always leads to different progression patterns and may require different types of therapy in clinical diagnosis. Therefore, it is crucial to study patient subgrouping. Normally, patient subgrouping is an unsupervised work due to the lack of labeled data. Analysing patients with complex medical data is challenging because of the data multiformity and time irregularity. To handle these issues, we propose a time-sensitive hybrid learning model to subgroup patients. First, we divide the multiform clinical data into two parts: non-time series data and time series data. Then we utilize basic autoencoder (AE) which is a commonly used unsupervised algorithm to learn patients' representations from non-time series data, and we use a recurrent neural network (RNN) based AE to extract representations from time series data. To capture the time irregularity in time series data, we propose a time-sensitive RNN which utilizes the time intervals to control the decaying degree of history memories. Finally, we present a weighted k-means method to subgroup patients with the pairwise representations. Experiments on real world medical datasets demonstrate that our proposed model can effectively improve the validity of patient subgrouping.

Vehicle-to-everything (V2X) communications, enabled by cellular device-to-device (D2D) links, have recently drawn much attention due to its potential to improve traffic safety, efficiency, and comfort. In this context, however, intracell interference combined with demanding latency and reliability requirements of safety vehicular users (V-UEs) are challenging issues. In this paper, we study a resource allocation problem among safety V-UEs, non-safety V-UEs, and conventional cellular UEs (C-UEs). Firstly, the resource allocation problem is formulated as a three-dimensional matching problem, where the objective is to maximize the total throughput of non-safety V-UEs on condition of satisfying the requirements on C-UEs and on safety V-UEs. Due to its NP-hardness, we then exploit hypergraph theory and propose a local search based approximation algorithm to solve it. Through simulation results, we show that the proposed algorithm outperforms the existing scheme in terms of both throughput performance and computational complexity.

This paper studies the synchronization problem for mobile cellular device-to-device (D2D) networks. Depending on the number of devices that are in coverage of the base station, the D2D environment can be divided into three categories where the partial-coverage and out-of-coverage are challenging scenarios and thus are the focus of this paper. First, we discuss five main challenges imposed on the synchronization problem in mobile D2D networks. More specifically, there are different challenges in the two coverage scenarios, since they do not have exactly the same synchronization objectives. Second, we propose a low-complexity Adaptive distRibuted nEtwork Synchronization (ARES) algorithm to address the five challenges. The design principles and the theories behind the ARES scheme are also analyzed in detail. Finally, we provide comprehensive simulations to evaluate different synchronization schemes, where the proposed ARES mechanism shows very promising performance.

This paper investigates the power control problem for broadcast vehicular communications in the presence of adjacent carrier interference (ACI). First, we study the impact of ACI on vehicle-to-vehicle (V2V) communications and conclude that the ACI indeed significantly affects the reliability of V2V links. Second, we formulate a power control optimization problem for vehicles to reduce the negative influence of ACI, which is shown to be NP-hard. Furthermore, we propose two power control schemes where the first one solves the formulated problem by a branch and bound method and the second one considers a heuristic algorithm with much reduced complexity. Finally, simulations are presented which illustrate the necessity of power control when ACI exists and also show promising performance of the proposed algorithms.

Deploying direct device-to-device (D2D) links is a promising technology for vehicle-to-X (V2X) applications. However, intracell interference, along with stringent requirements on latency and reliability, are challenging issues. In this paper, we study the radio resource management problem for D2D-based safety-critical V2X communications. We first transform the V2X requirements into the constraints that are computable using slowly varying channel state information only. Secondly, we formulate an optimization problem, taking into account the requirements of both vehicular users (V-UEs) and cellular users (C-UEs), where resource sharing can take place not only between a V-UE and a C-UE but also among different V-UEs. The NP-hardness of the problem is rigorously proved. Moreover, a heuristic algorithm, called Cluster-based Resource block sharing and pOWer allocatioN (CROWN), is proposed to solve this problem. Finally, simulation results indicate promising performance of the CROWN scheme.

Clock synchronization is a crucial issue for mobile ad hoc networks due to the dynamic and distributed nature of these networks. In this paper, employing affine models for local clocks, a random broadcast based distributed consensus clock synchronization algorithm is proposed. In the absence of transmission delays, we theoretically prove the convergence of the proposed scheme, which is further illustrated by numerical results. In addition, it is concluded from simulations that the proposed scheme is scalable and robust to transmission delays as well as different accuracy requirements.

This paper investigates the problem of clock synchronization of nodes in a wireless sensor network based on the two-way timing message exchange mechanism with an unknown deterministic transmission delay and random exponential transmission delays. Without knowing the fixed delay, a novel synchronization scheme is proposed for the linear clock model, which works well in both symmetric and asymmetric links. In the proposed algorithm, the clock skew and offset are estimated by directly utilizing mean square error as the metric to be optimized. This consideration results in significant performance improvements compared to the existing synchronization methods, especially for a small number of observations and large standard deviation of the random delays.

Clock synchronization is a crucial issue in the operation of wireless sensor networks. Although the existing synchronization algorithms under linear clock model assumptions perform well for short periods, they will become problematic for applications with long-term requirements. In this paper, we consider a more realistic and flexible relationship model for two clocks and exploit a Taylor expansion to approximate the relationship. Based on this model and a two-way time message exchange procedure, an estimation algorithm is proposed to recover the relationship and then achieve the synchronization. Finally, simulation results demonstrate that the proposed algorithm improves the accuracy of synchronization as compared to existing algorithms in many scenarios, and is also robust to different distributions of random delays.

In this paper, joint optimization of subcarrier pairing, relay selection and power allocation is formulated through a capacity maximization problem in orthogonal frequency division multiplexing (OFDM) relay systems, with fairness considerations and individual power constraint on each node. Because achieving the optimal solution is computationally demanding, which is resulted from the excessive possibilities of joint resource allocation, we propose a sub-optimal allocation method. The proposed scheme decouples the allocation procedures into two steps, which reduces the complexity significantly by avoiding the search for all possible subcarrier pairs. Moreover, simulation results reveal the effectiveness of the proposed algorithm in both Amplify-and-Forward (AF) and Decode-and-Forward (DF) relay systems.

In this paper, we evaluate the existing transmission schemes in multi-cell environment, such as single-cell transmission with Inter-cell interference (ICI) as well as Collaborative Multi-Point transmission/reception (CoMP) technique in LTE-A, and propose an more effective solution to deal with the inter-cell interference. The proposed transmission structure takes advantage of the cooperation between base stations (BS), and combines both closed-loop and open-loop techniques. With the concept of Effective Channel, the classic Space Frequency Block Coding (SFBC) decode scheme can be used directly in our receiver for single layer transmission. Thus, we prove that the scenario of multiple base stations can employ SFBC conveniently while the performance is guaranteed. Analysis and simulation results show that our method outperforms conventional single cell transmission. Moreover, the proposed transmission scheme can obtain elegant performance enhancement compared with local precoding and global precoding of CoMP in terms of capacity, and doubly reduce the feedback overhead at the same time.

In orthogonal frequency division multiplexing (OFDM) systems, some subcarriers are often retained as null subcarriers for easier filter implementation. In this case, the performance of conventional Discrete Fourier Transform (DFT)-based channel estimation will be degraded dramatically due to dispersive distortion of actual channel impulse response (CIR). To solve this problem, in the present paper, we analyze the effect of null subcarriers on channel estimation and propose a novel channel estimation scheme. The proposed scheme, whose basic principle is the joint iteration of correlation operation and interference cancellation in time domain, can hold most energy of actual CIR and suppress noise effectively. Finally, simulation results show that the proposed method outperforms Least Square (LS) channel estimator, conventional DFT-based channel estimator and IDFT-Zero-Forcing(ZF) channel estimatior, especially in low signal-to-noise ratio (SNR).