This paper proposes a two-stage fingerprint localization framework to address the challenges of excessive reliance on massive real reference points and high computational complexity. The proposed framework employs distinct neural networks on its two stages. First, a generative adversarial network (GAN)-based dataset augmentation is implemented to synthesize virtual fingerprints, expanding radio map coverage and reducing reliance on massive reference points. Second, a hierarchical localization model based on deep learning is introduced, using a convolutional neural network (CNN) for rapid and coarse-grained localization with reduced complexity, followed by a probabilistic deep neural network (DNN) to achieve refined and accurate positioning. Simulation results show that the proposed framework achieves a Top-1 accuracy of 81.2% and a Top-2 accuracy of 98%, outperforming benchmark algorithms. Furthermore, training with a dataset of 57% GAN-generated fingerprints reduces localization error by 30.9%, demonstrating the effectiveness of dataset augmentation strategy.

In this paper, we introduce a simplified long-range (LoRa) backscatter system that enables a lightweight tag to communicate with a remote transceiver using chirp carrier and harmonic backscatter. The key idea is twofold: first, we delegate the chirp waveform generation from the tag to the transceiver, thereby maintaining lightweight tag design; second, we introduce a frequency shift on the tag during its carrier modulation to create harmonic backscatter, thereby mitigating self-interference. To refine the framework, we first detail the system model, including the design for harmonic backscatter. We then present an efficient method to address synchronization and detection issues at the transceiver. Finally, we prototype the proposed system and evaluate its performance. Experimental results demonstrate that in a corridor environment with a carrier power of 0 dBm, the tag can backscatter a signal from the transceiver at a maximum bit rate of 4 kb/s over a distance of 50 meters.

Due to its almost zero power consumption and low-hardware costs, backscatter technology has recently gained considerable attention. However, traditional backscatter systems employ basic binary modulations to transmit data at a relatively low rate. In this article, we introduce VITAS, a new backscatter system that enables the tag to communicate with the transceiver using versatile modulations at a megabit rate. First, we present the design of the tag, which supports both binary and higher order modulations. We also develop a series of assembly instructions for the tag controller to generate modulated symbols with minimal clock cycles, thereby enhancing the data rate for a given clock frequency. Then, we design the transceiver to coordinate the backscatter communications with the tag. The transceiver incorporates a specific symbol synchronization algorithm to address the accumulated timing errors resulting from the tag unstable clock during high-rate transmission. Furthermore, we fabricate the tag on a four-layer acrlong PCB and implement the transceiver on a acrlong USRP X310. Finally, we provide experimental results to show that the VITAS system is capable of transmitting acrshort QPSK symbols at a maximum rate of 3 Mbit/s while consuming only$9.63~\mu \text{W}$(3.21 pJ/bit).

We propose a new wireless paradigm for railways – transmissive metasurface (TMS) assisted communications. It compensates for the Doppler shift brought by high mobility and reduces signal degradation due to train carriages. Specifically, the elements of the TMS panel attached to train windows manipulate the links between the base station (BS) and the onboard users. One fundamental problem with it is: does the channel introduced by TMS outperform the traditional direct channel? To answer this, we compare the gains of direct and cascaded channels and introduce the cascaded-outperform-direct probability (CODP), which is the probability that the latter exceeds the former. We then derive two simplified closed-form CODP expressions by approximating the cascaded channel gain as mixed Gaussian and Gamma distributions. Moreover, BS-related parameters impact the CODP; we show that the CODP has (i) global maximum points concerning the azimuth angle of the path from the BS to the TMS, the distance from the BS to the railway, and the BS height, respectively, and (ii) a minimum point in terms of the azimuth angle of the path from the BS to the TMS. Finally, we provide numerical results to verify our analysis and derivations.

Piezo-Acoustic Backscatter (PAB), one type of near-zero power backscatter, is a new technology that realizes underwater communications. In this paper, we first describe the PAB node and the channel. Then we set up a signal transmission model and formulate the underwater channel estimation problem. Based on the sparse characteristics of the channel and compressed sensing theory, we use the gradient projection for sparse reconstruction (GPSR) algorithm to solve the underwater channel estimation problem and obtain the channel estimate. Simulation results prove that in the underwater PAB system, the GPSR algorithm offers better estimation accuracy than the traditional LS algorithm.

In this6study, the channel estimation problem is investigated for a wireless communication system assisted by a reconfigurable intelligent surface (RIS). The RIS thus creates an assistant channel, which has the features of positivity and dominance. Owing to these features, the channel estimation problem is formulated as a constrained residual sum of squares minimisation problem, which differs radically from the traditional channel estimation issue. An efficient Lagrange multiplier and dual ascent‐based estimation scheme is then designed to obtain an iterative solution for the estimator. Moreover, the Cramér–Rao lower bounds are deduced as a performance benchmark. Simulation results show that the authors' designed scheme improves the estimation accuracy up to 33%, compared with the conventional least‐square method in the low signal‐to‐noise ratio regime.

In this paper, we consider a cellular device-to-device (D2D) system which consists of one base station (BS) deployed with a mobile edge computing (MEC) server, and a number of users. By defining task execution cost as the weighted sum of execution latency and energy consumption, the joint computation offloading and resource allocation problem is formulated as a task execution cost minimization problem under the constraints of task requirement, computation offloading, resource allocation and task partition, etc. As the formulated optimization problem is a mixed integer nonlinear problem, which cannot be solved conveniently, we decompose it into two subproblems, i.e., computation offloading subproblem and resource allocation subproblem, and solve the two subproblems by applying Kuhn-Munkres algorithm and Lagrange dual method, respectively. Numerical results demonstrate the effectiveness of the proposed scheme.