Indoor localization is essential for navigation, tracking, and path planning applications. Traditional systems, based on sensors like inertial devices and LiDAR, offer high accuracy but are costly and prone to cumulative errors. We propose a cost-effective multi-sensor fusion system specifically tailored for two-dimensional localization of two-wheeled mobile robots, combining Wi-Fi channel state information (CSI) and odometry, using an extended Kalman filter (EKF) and an adaptive Monte Carlo localization (CSI-AMCL) algorithm to enhance accuracy. Our innovative 1D convolutional neural network (1D-CNN) based on residual networks effectively processes CSI data, improving adaptability in complex environments by addressing the vanishing gradient issue. Our approach increases accuracy by 56% compared to Wi-Fi fingerprinting. Tests show a 20.1% improvement over WIO-EKF and a 36.3% improvement over Fusion-dhl. This demonstrates the potential of our method for enhancing multi-sensor fusion systems.

Recently, IoT (Internet-of-Things) devices are very widely used in our daily lives and their design and manufacturing are often outsourced to third parties to make them at a low cost. Meanwhile, malfunctions may be inserted into them intentionally by malicious third parties. Utilizing power waveforms measured from IoT devices is one of the effective ways to detect its anomalous behaviors. Most IoT devices regularly consume steady-state power due to the operating system and/or hardware components and we have to remove it from the total power to detect anomalous behaviors. However, the existing methods manually or semi-manually remove the steady-state power and further they utilize the pre-determined features in the power waveform to detect anomalies. Hence, they cannot well detect them automatically. In this paper, we propose a method, called Gen-Power2, to detect anomalous behaviors in IoT devices utilizing the generative machine-learning model. The proposed method generates an application power waveform by inferring the steady-state power by machine-learning from the observed total power waveform. Then, the anomalous application behaviors are detected by automatically extracting the latent features from the generated application power waveform. Experimental evaluations show that Gen-Power2 detects anomalous application behaviors successfully, while the recent state-of-the-art method cannot detect them.

Cache tiling or recursive data layouts for two-dimensional (2-D) data access has been proposed to ameliorate the poor data locality caused by conventional layouts like row-major and column-major. However, cache tiling and recursive data layouts require non-conventional address computation, which involves bit-level manipulations that are not supported in current processors, there is also a significant overhead in execution time due to software-based tiling address calculation. In this paper, we design a cache memory with hardware-based tile/line accessibility support for 2-D data access and a tile-set-based tag comparison (TSTC) scheme to optimize overall hardware scale overhead. Our technique captures the benefits of locality of the sophisticated data layouts while avoiding the cost of software-based address computation. Simulation results show the proposed method improves the performance of matrix multiplication (MM) over conventional data layout and Z-Morton order layout by reducing L1 cache, L2 cache and Translation Lookaside Buffer (TLB) misses, especially at larger matrix sizes. We implement the proposed cache with a SIMD-based data path by using 40 nm Complementary Metal-Oxide-Semiconductor (CMOS) technology. The entire hardware overhead of the proposed TSTC method was reduced to only 10% of that required for a conventional cache without performance degradation.

This paper presents an iterative algorithm for computing an eigenvalue close to a user-specified value and its corresponding eigenvector of a nonlinear eigenvalue problem. This algorithm iterates two parts alternately. The first part is the existing algorithm called the successive approximation algorithm, where the Taylor expansion of a matrix is used to transform the nonlinear problem to the linear problem. By solving the linear problem, an approximate eigenvalue and an approximate eigenvector of the nonlinear problem are computed. The second part refines the approximate eigenvalue computed by the first part. To this end, we approximately compute the Rayleigh functional, which is the solution of the nonlinear equation defined by the approximate eigenvector, and use it as a new approximate eigenvalue. Experimental results show that a combination of the successive approximation algorithm and the Rayleigh functionals converges within fewer iterations and requires less computational time in comparison with the existing successive approximation algorithms.

For a regularized least squares estimation of discrete-valued signals, we propose a Linearly involved Generalized Moreau Enhanced (LiGME) regularizer, as a nonconvex regularizer, of designated isolated minimizers. The proposed regularizer is designed as a Generalized Moreau Enhancement (GME) of the so-called sum-of-absolute-values (SOAV) convex regularizer. Every candidate vector in the discrete-valued set is aimed to be assigned to an isolated local minimizer of the proposed regularizer while the overall convexity of the regularized least squares model is maintained. Moreover, a global minimizer of the proposed model can be approximated iteratively by using a variant of the constrained LiGME (cLiGME) algorithm. To enhance the accuracy of the proposed estimation, we also propose a pair of simple modifications, called respectively an iterative reweighting and a generalized superiorization. Numerical experiments demonstrate the effectiveness of the proposed model and algorithms in a scenario of multiple-input multiple-output (MIMO) signal detection.

Permutations on the vector spaces 𝔽ⁿ_q are few at present. Inspired by the work of Chi, Li and Qu [1], we construct two classes of permutations with 3-homogeneous structures in trivariate form over 𝔽³_2^m. To establish their permutation properties, we formulate a system of equations and analyze it using techniques such as resultants, multivariate methods, and the method of undetermined coefficients.

Symmetrical Z-complementary code sets (SZCCSs), as a novel type of training sequences, have demonstrated exceptional channel estimation performance when applied to generalized spatial modulation (GSM) systems. This paper addresses the current limitations in the design methods and parameter range of SZCCSs by proposing three schemes. Initially, by analyzing the complementary properties of Discrete Fourier Transform (DFT) matrix elements, we construct a class of optimal SZCCSs using the Kronecker product of unit modulus sequences and DFT matrices. Subsequently, to diversify the parameter of code quantity, we perform the Kronecker product with complete complementary codes (CCCs) and orthogonal matrices, resulting in a class of optimal SZCCSs. Also, if a sequence with low autocorrelation replace the orthogonal matrix, the resulting SZCCS exhibits low autocorrelation sidelobes outside the zero correlation zone and complete complementary cross correlation characteristics. The study offer a broad range of parameters not previously identified in the literature, significantly enriching the parameter space of SZCCSs. Simulation results validate that the proposed SZCCSs show superior resistance to multipath interference in GSM systems compared to traditional sequences, highlighting their potential advantages in channel estimation.

With the expansion of industrial applications and data volumes, building a user-edge-cloud collaborative network and efficiently distributing computing tasks has emerged as a critical solution to alleviate terminal computing burdens and ensure quality of service (QoS). However, existing studies often overlook extra delays in the offloading process, including queuing, propagation, and wired transmission delays, significantly impacting task delay and offloading strategies. To this end, this paper proposes a multi-slice task offloading and resource allocation scheme for user-edge-cloud networks, targeting delay minimization while considering various delay factors. This scheme jointly optimizes offloading mode, offloading ratio, user association, and resource allocation under task delay constraints. To address the problem’s non-convexity, an alternating optimization framework is employed to decompose the problem into offloading mode selection and resource allocation subproblems. Specifically, a deep reinforcement learning (DRL)-based algorithm is developed for offloading mode selection, while convex optimization techniques are applied to determine optimal offloading ratios and resource allocation. Additionally, a matching theory-based algorithm establishes optimal connections between users and base stations (BSs). Simulations validate the effectiveness of the proposed scheme, showing that the three-layer offloading mode, i.e., collaborative computing across user, edge, and cloud reduces latency compared to single-layer and two-layer modes for large-scale tasks.

There are some research projects on the analysis of traffic states by using mathematical models. In this article, we propose a new method based on a distributed constant circuit, namely an electrical circuit based on distributed elements. In this method the distributed constant circuit is regarded as a traffic circuit. The voltage on the distributed constant circuit is regarded as traffic pressure, like the previous works based on the lumped elements on the lumped constant circuit. However, these previous works have some difficulties on the definition of the traffic pressure. In virtue of the distributed elements composed of not Resistance, but Inductance and Capacitance, the input voltage does not augment unreasonably. The analogy between traffic circuit and electrical circuit becomes more reasonable. Moreover, the status of traffic light is also taken into account. The validity of our proposed method was confirmed by some simulations based on distributed constant circuits.

The skeleton modality provides an efficient representation of human pose. However, its lack of appearance information can lead to poor performance in tasks requiring such information. To address this, we propose a multimodal skeleton representation that integrates intermediate feature maps from a pose estimation network, called Pose Feature Map Enhanced Skeleton Representation (PFMESR). Specifically, we estimate the joint positions of the human body in the video and locate the local features related to each joint from the feature maps of the pose estimation network. These local features are then aligned and fused with the skeleton features in the action recognition network. We believe that the feature maps from the pose estimation network contain rich appearance information that complements the skeleton information. Experiments on multiple datasets demonstrate that this approach significantly improves action recognition performance and yields favorable results in the Action-Identity Recognition task, proving the effectiveness of incorporating appearance information from pose estimation feature maps. We also investigated the relationship between PFMESR’s performance and sampling depth and range to explore its effectiveness under different parameters. Additionally, we validated the generality of PFMESR by applying it to various skeleton-based methods. Our method surpasses the state-of-the-art on multiple skeleton-based action recognition benchmarks, achieving accuracies of 94.6% on the NTU RGB+D 60 cross-subject split, 97.7% on the NTU RGB+D 60 cross-view split, and 93.1% on the NTU RGB+D 120 cross-subject split.

Automatic Question Generation (AQG) aims to generate natural and relevant questions based on a given context and optional answers. It is a significant and challenging task in the field of natural language processing. However, existing AQG models often produce a single type of question with repetitive content, which hinders the diversity of the generated questions. In this paper, we introduce a Diversify Question Generation model based on the Diffusion Model (DQG-DM). Our model effectively incorporates latent variables and fine-grained question types to ensure both the relevance and diversity of the generated questions. Experiments conducted on two benchmark datasets demonstrate that our proposed model outperforms the state-of-the-art results.

We propose a probabilistic interpretation of the Gap Shortest Vector Problem by leveraging Södergren’s analysis of random lattices. By linking shortest vector norms with the Gaussian heuristic, our study characterizes GapSVP behavior in high dimensions and informs secure parameter choices in lattice-based cryptography.

In this letter, we construct a class of binary parity-check matrices with column weight 3 and show that these matrices have full-rank. Then we prove that the stopping distance of each binary parity-check matrix is equal to the minimum distance of the code specified by the parity-check matrix. Taken together, we obtain a new class of binary linear codes with optimal stopping redundancy.

For linear state estimation problems involving Brownian motion process noise, this paper proposes a novel adaptive Kalman filter that leverages online assessment of the power spectral density (PSD) for continuous-time dynamic noise. Unlike existing adaptive filters that estimate the entire noise covariance matrix, this work proposes to directly evaluate the noise PSD according to a analytical derivation for process noise covariance. As the key innovation, the proposed adaption scheme significantly reduces the number of scalar unknowns and results in enhanced accuracy for estimating the PSD of Brownian motion noise. As the resulted advantage, the new adaptive Kalman filter mitigates the crucial reliance on noise statistics without extra computation. Numerical examples of target tracking demonstrate the new adaptive Kalman filter’s filtering adaptability, accuracy, and simplicity.