Pareto local search (PLS) serves as an important component in multi-objective combinatorial optimization. Nevertheless, achieving a balance between convergence and diversity remains a challenge, as few studies have leveraged knowledge from the search archive to effectively guide the PLS process. This paper proposes an archive entropy-guided Pareto local search algorithm (AEG-PLS). In the proposed method, the objective space is partitioned into subregions using a set of reference vectors. The archive entropy is then computed for each subregion to assess population diversity. To enhance diversity in less explored areas, a PLS is initiated using a well-performing solution selected from the subregion with the lowest entropy, thus indicating the weakest diversity. This approach promotes a more balanced trade-off between convergence and diversity throughout the optimization process. Experimental results on 25 multi-objective combinatorial optimization benchmark instances demonstrate that the proposed AEG-PLS achieves competitive performance in terms of both Inverted Generational Distance and Hypervolume metrics when compared to nine state-of-the-art multi-objective evolutionary algorithms.

This study presents an innovative approach to designing filament-wound conical pressure vessels loaded by internal pressure, combining analytical computations based on classical lamination theory with data-driven evolutionary algorithms. The most critical area of the whole cone is the balanced layer at the lower base, in which the safety level is evaluated using the failure index based on Hoffman’s strength criterion. The designed mathematical model related the geometrical parameters of the cone to the magnitude of internal pressure, representing the performance of the structure and safety expressed by the failure index. The optimization loop runs on two modules –Evolutionary Deep Neural Networks EvoDN2 module for machine learning and creating a lighter surrogate model, together with Constrained Reference Vectors Evolutionary Algorithms (cRVEA) for multicriteria optimization. This was applied to three different material configurations: E-glass/DA4518U (glass-epoxy), T300/N5208 (carbon-epoxy) and AS4/3501-6 (carbon-epoxy). The results show that the critical area is invariant to the material configuration. In addition, it was shown that the material configurations with higher anisotropy achieve a more uniform distribution of the load in the layers of the conical three-layered wall.

The paper addresses the problem of improving the fault tolerance of the inertial sensor block in strapdown attitude systems.The proposed solution is based on the use of a redundant number of sensors with independent sensitivity-axis orientations.Three minimally redundant configurations of sensor sensitivity axes are considered: “3 orthogonal + 1”, “cone-4”, and “cone-3 with axis”.Geometric schemes are presented, along with the necessary relationships between the unit vectors of sensor sensitivity axes and those of the body-fixed (instrument) coordinate system for forming the measurement matrix. Under normal operation of all sensors, formulas are derived for transforming the angular quasi-coordinates of the complete sensor set into the body-fixed coordinate frame. In the case when a faulty sensor is identified, transformation equations are given for converting the angular quasi-coordinates obtained from sensor triads into the body-fixed axes for subsequent use in the attitude determination algorithm. As a reference vector for fault identification and exclusion of the faulty sensor’s data from the measurement matrix, a fictitious apparent rotation vector, formed from the sensors’ angular quasi-coordinates, is proposed. An exponential approximation of the apparent rotation vector modulus is performed over a limited interval of inertial data sampling, and a discrete model of this modulus is constructed based on its values over the previous four sampling intervals. Subsequent use of this discrete model as an extrapolation function allows estimation of the apparent rotation vector modulus for the current sampling step. By comparing the modulus of the apparent rotation vector computed from various sensor triads with the extrapolated value, the faulty sensor can be identified. Numerical implementation of the proposed fault detection and sensor identification method, using a formal kinematic model of rigid-body conical motion under angular quasi-coordinate quantization, has demonstrated the feasibility and effectiveness of the proposed approach.

To construct a vector field-oriented control system in the rotor circuit of a doubly fed machine (DFM) it is necessary to determine the spatial position of the reference vector of the stator flux linkage. Recently instead of sensors flux linkage observers have been used for these purposes, which are systems of differential equations of electromagnetic processes of an asynchronous machine (AM), which are solved in real time relative to the projections of current vectors and flux linkages. The accuracy of identification is affected by the change in the AM parameters during its operation and, first of all, by the change in the active resistances of the phase windings of the stator and rotor of the DFM. To increase the robustness of the observer to the change in internal parameters it is possible to apply negative feedback on the deviations of actual currents from their measured values, which leads to signal compensation of parametric disturbances. However an approach is possible consisting in the implementation of parametric feedback on the parameters that are most susceptible to change. In this case the equations of the disturbed motion of the observer are written down and on the trajectories of his motion the total derivative with respect to time is taken from a positively defined quadratic form, which claims to be a Lyapunov function. If the negative definiteness of the total derivative with respect to time is ensured, then the algorithms for identifying parameters are obtained that are asymptotically stable in all modes of operation of the DFM. If the negative definiteness of the total derivative with respect to time is not achieved then due to the sufficiency of the stability conditions formulated by Lyapunov's second theorem on the stability of motion, it is necessary to conduct additional studies. These studies consist in determining the initial conditions and ranges of change of the state variables of the DFM under which the parameter identification algorithms retain local stability as a whole without the property of asymptotic stability. It is this case that is studied in this paper, where the areas of local stability of the algorithms of identification of active resistances of the windings of the DFM were determined, as parameters included in the stator flux linkage observer.

Facial expression recognition based on multi-domain norm-referenced encoding.

Purpose of reseach . To develop a self-learning algorithm for a two-layer neural network model that is as efficient as possible, given the current technical implementation of intelligent systems, including those designed for solving pattern recognition problems. This algorithm will be based on increasing the number of neurons and varying the weight coefficients of synaptic connections, with the possibility of extending it to a high-order multiconnected neural network with an internal product of vectors. Methods . To solve this problem, this paper proposes an approach to synthesizing a high-order multiconnected neural network model with an internal product of vectors, as well as a self-learning algorithm for such a neural network. This algorithm provides for the rapid correction of the elements of the reference matrix, instead of the traditional variation of the weight coefficients of synaptic connections, in order to reduce the resource intensity of the performed operations. Results . The proposed method was implemented as a software application linked to the self-training of a high-order neural network using speech sound types represented in raster format, pre-segmented from the general stream and transformed into polar coordinates for ease of processing and storing the resulting images as a training set. Conclusion . The developed algorithm, during software implementation testing, demonstrated relatively high efficiency by eliminating the resource-intensive operation of varying weight coefficients and replacing it with direct correction of the reference matrix. This algorithm demonstrated relatively high efficiency, convergence in a finite number of steps due to the limited number of first-approximation codes of reference vectors, and noticeable performance compared to known analogs.

The article presents the development of an intelligent system for classifying user-generated texts in social networks under conditions of linguistic uncertainty. It addresses the growing need for automated identification of professional interests among social media users, which is especially relevant for recruitment, education, and professional orientation. Traditional methods for identifying target audiences are time-consuming, costly, and inefficient when dealing with unstructured data. The proposed system utilizes and compares modern text vectorization methods — TF-IDF, FastText, and BERT — to improve classification accuracy. The system architecture includes modules for collecting user data from VKontakte via API, preprocessing text (normalization, lemmatization, noise reduction), and thematic classification using a predefined set of IT-related terms. The classification decision is based on the scalar product between text vectors and reference vectors of key terms. The system also supports an ensemble decision mechanism combining multiple models to increase reliability. The study provides a comparative analysis of the effectiveness of the selected vectorization methods in binary classification tasks (IT-related or not) using real user data. Experiments demonstrate the superiority of BERT in terms of accuracy, followed by FastText. TF-IDF showed lower sensitivity to thematic content in short and informal mes­sages. A web-based interface was developed to automate user classification. It allows the input of a VK community ID, retrieves and analyzes user data, and generates reports on IT interest distribution. The system can be applied in career guidance, educational analytics, and HR processes to identify suitable candidates based on their digital footprint.

This study develops, dynamically simulates, and optimizes an integrated tri‑generation system for year-round electricity, heating, and cooling supply under the hot-dry climatic conditions of Baghdad, Iraq. The proposed configuration couples a low‑concentration hybrid PV–compound parabolic concentrator (LCPV-CPC) with dual small‑scale gas turbines, high- and low-grade water-source heat pumps, and an ammonia-water absorption chiller, coordinated through a following-electric-load (FEL) strategy. The primary objectives are to maximize primary energy savings, annual cost reduction, CO2 emissions mitigation, and exergy efficiency by exploiting multi‑grade thermal integration and dispatch optimization. A methodological novelty lies in applying a Reference Vector Guided Evolutionary Algorithm (RVEA) with entropy‑weighted VIKOR analysis to achieve balanced trade‑offs among energy, economic, environmental, and thermodynamic criteria. Dynamic co‑simulation through Aspen HYSYS-MATLAB, validated against high‑quality experimental data, ensures predictive reliability. Results confirm substantial performance gains compared with a separate production facility: primary energy savings up to ~ 33%, annual cost savings exceeding 10% at favorable solar conditions, and CO2 emission reduction approaching 50%. Parametric analysis shows that increased solar irradiance significantly improves environmental and economic outcomes, with economic feasibility achieved beyond ~ 472 W/m2 average radiation. Exergy efficiency remains stable or slightly declines at high irradiance due to intensified off‑design irreversibilities. Optimal inlet water temperatures to the LCPV-CPC further enhance renewable contribution without notable thermodynamic penalties. The findings demonstrate a technically and economically viable pathway for sustainable tri‑generation in climates with strong solar resources and high cooling demand, offering a transferable optimization framework for future hybrid renewable–fossil energy applications in urban buildings.

A key direction of the development of modern power systems is the application of a continuously increasing number of grid-forming power converters to provide various system services. One of the possible strategies for the implementation of grid-forming control is a control algorithm based on a virtual synchronous generator (VSG). However, at present, the problem of VSG operation under abnormal conditions associated with an increase in output current remains unsolved. Existing current saturation algorithms (CSAs) lead to the degradation of grid-forming properties during overcurrent limiting or reduce the possible range of current output. In this regard, this paper proposes to use the structure of modified current-controlled VSG (CC-VSG) instead of traditional voltage-controlled VSG. A current vector amplitude limiter is used to limit the output current in the CC-VSG structure. At the same time, the angle of the current reference vector continues to be regulated based on the emerging operating conditions due to the voltage feedback in the used VSG equations. The presented simulation results have shown that it was possible to achieve a wide operating range for the current phase from 0° to 180° in comparison with a traditional VSG algorithm. At the same time, the properties of the grid-forming inverter, such as power synchronization without phase-locked loop controller, voltage, and frequency control, are preserved. In addition, in order to avoid saturation of the voltage controller, it is proposed to use a simple algorithm of blocking and switching the reference signal from the setpoint to the current voltage level. Due to this structure, it was possible to prevent saturation of integrators in the control loops and to provide a guaranteed exit from the limiting mode. The results of adding this structure showed a five-second reduction in the overvoltage that occurs when it is absent. A comparison with conditional integration also showed that it prevented lock-up in the limiting mode. The results of experimental verification of the developed prototype of the inverter with CC-VSG control and CSA are also given, including a comparison with the serial model of the hybrid inverter. The results obtained showed that the developed algorithm excludes both the dead time and the load current loss when the external grid is disconnected. In addition, there is no tripping during overload, unlike a hybrid inverter.

Reference Vector-guided Evolutionary Algorithm for cluster analysis of single-cell transcriptomes.

This paper presents a novel methodology for diagnosing hardware systems based on the use of the digital twin by using a vector-based representation of both input and output values. The method involves comparing the output vector of a real electronic device with a corresponding reference vector derived from a mathematical model (digital twin). Deviations between these vectors are analyzed using a specially designed table of discrete qualitative correspondences, which enables identification of the likely causes of incorrect behavior in specific components without requiring intrusive measurements or additional sensors.The approach models the behavior of an electronic system as a transformation from an input vector to an output vector, where each deviation in input causes a deterministic change in output. This vector-based abstraction allows for the construction of decision rules and classification zones using relative relationships between components rather than absolute values, making the system robust to noise, parasitic effects, and measurement uncertainties. The proposed method is computationally efficient, allowing deployment on low-power embedded platforms, and does not rely on deep physical modeling or extensive machine learning training datasets. The methodology introduces the concept of a “gray zone” in the multidimensional output space, where small deviations are treated as acceptable tolerances, and only significant directional changes trigger diagnostic logic. Additionally, a normal-ization step and the introduction of threshold vectors allow for scalable precision and adaptation to non-linear dependencies in hardware behavior.Advantages of this approach include simplicity of implementation, human readability of diagnostic outputs (e.g., “+”, “++”, “–”, “--”), low computational requirements, and high resilience in noisy or constrained environments. Moreover, the system enables real-time monitoring and provides a foundation for future extensions such as remaining useful life (RUL) pre-diction and adaptive table generation through machine learning.The methodology is particularly suited for power electronics and embedded systems, but can be generalized to any ap-plication where the device can be abstracted as a system with measurable input/output behavior. This work provides a flexible, scalable, and interpretable diagnostic framework that relies on vector mappings within a model-based digital twin approach, enabling robust and low-cost fault detection without the need for high-precision measurements, extensive sensor networks, or machine learning.

The low-cost Strapdown Inertial Navigation System (SINS)/Global Navigation Satellite System (GNSS) is widely used in autonomous vehicles for positioning and navigation. Initial alignment is a critical stage for SINS operations, and the alignment time and accuracy directly affect the SINS navigation performance. To address the issue that low-cost SINS/GNSS cannot effectively achieve rapid and high-accuracy alignment in complex environments that contain noise and external interference, an adaptive multiple backtracking robust alignment method is proposed. The sliding window that constructs observation and reference vectors is established, which effectively avoids the accumulation of sensor errors during the full integration process. A new observation vector based on the magnitude matching is then constructed to effectively reduce the effect of outliers on the alignment process. An adaptive multiple backtracking method is designed in which the window size can be dynamically adjusted based on the innovation gradient; thus, the alignment time can be significantly shortened. Furthermore, the modified variational Bayesian Kalman filter (VBKF) that accurately adjusts the measurement noise covariance matrix is proposed, and the Expectation–Maximization (EM) algorithm is employed to refine the prior parameter of the predicted error covariance matrix. Simulation and experimental results demonstrate that the proposed method significantly reduces alignment time and improves alignment accuracy. Taking heading error as the critical evaluation indicator, the proposed method achieves rapid alignment within 120 s and maintains a stable error below 1.2° after 80 s, yielding an improvement of over 63% compared to the backtracking-based Kalman filter (BKF) method and over 57% compared to the fuzzy adaptive KF (FAKF) method.

Abstract Ground segmentation-based point cloud processing methods have gained significant attention in simultaneous localization and mapping (SLAM) for LiDAR-inertial odometry applications, particularly in autonomous vehicles. However, existing methods often lack robustness in complex environments, such as uneven terrains or obstructed areas. To address these challenges, we propose a novel point cloud segmentation algorithm that combines vector iteration processing and ground expansion techniques. This method iteratively updates the reference vector using the normal vectors of successfully fitted planes, while expanding the ground points by incorporating neighboring planes. Furthermore, to mitigate redundant calculations in conventional approaches, we introduce a grid-based local plane fitting method. By partitioning the ground into smaller regions and employing a custom node structure, this technique updates plane parameters iteratively, improving both fitting accuracy and computational efficiency. For non-ground point clouds, we propose an iterative principal component analysis (iPCA) method combined with outlier filtering. Experimental evaluations were conducted using 20 sequences from four public datasets and a custom campus dataset. The proposed method achieved an average time savings of 26.7% compared to state-of-the-art approaches, while maintaining superior map accuracy. This work contributes to advancing robust and efficient point cloud processing for autonomous vehicle applications in challenging environments.

ABSTRACTWe propose a novel estimator for the principal component (PC) subspace tailored to the high‐dimension, low‐sample size (HDLSS) context. The method, termed adaptive reference‐guided (ARG) estimator, is designed for data exhibiting spiked covariance structures and seeks to improve upon the conventional sample PC subspace by leveraging auxiliary information from reference vectors, presumed to carry prior knowledge about the true PC subspace. The estimator is constructed by first identifying vectors asymptotically orthogonal to the true PC subspace within a signal subspace, the subspace spanned by the leading sample PC directions and the references, and then taking the orthogonal complement. The estimator is adaptive, as it automatically selects the subspace asymptotically closest to the true PC subspace inside the signal subspace, without requiring parameter tuning. We show that when the reference vectors carry nontrivial information, the proposed estimator asymptotically reduces all principal angles between the estimated and true PC subspaces compared to the naive sample‐based estimator. Interestingly, despite being derived from a completely different rationale, the ARG estimator is theoretically equivalent to an estimator based on James–Stein shrinkage. Our results thus establish a theoretical foundation that unifies these two distinct approaches.

The optimization and control of the wellbore trajectory is one of the important technologies to improve drilling efficiency, reduce drilling cost, and ensure drilling safety in the process of modern oil and gas exploration and development. In this paper, a multi-objective wellbore trajectory optimization mathematical model is established, which takes into account the five factors of wellbore trajectory length, friction, torque, trajectory complexity, and target accuracy. A DR-NSGA-III-MGA algorithm (dynamic reference NSGA-III with multi-granularity adaptation) is proposed. By introducing multi-granularity reference vector generation and an information entropy-guided search direction adaptation mechanism, the performance of the algorithm in the complex target space is improved, and the three-stage wellbore trajectory is optimized. Simulation experiments show that the DR-NSGA-III-MGA algorithm is stable in a variety of complex problems, while maintaining good convergence, and has good generalization ability and practical application value.

Abstract We study claims problems in which agents may also have reference points. We show first that many classical rules satisfy an egalitarian property in this setting; namely, the differences between each agents’ payoff and the corresponding reference value are as equal as possible. We also introduce a broad class of rules that satisfy a generalized condition, dubbed egalitarian-in-deviation relative to a reference system. For each problem, the system proposes a reference vector which is a function of the claims. We show that these rules allocate the nearest efficient point to the reference vector. Our findings generalize previous results in the literature, such as the one stating that the CEA rule minimizes the squared distance to the equal division point. Concede-and-divide, a focal rule to solve two-agent claims problems, does not satisfy the egalitarian-in-deviation condition relative to any reference system. But, under certain conditions, it can be reinterpreted as the limit of a weighted egalitarian-in-deviation rule. Finally, we explore the behavior of egalitarian-in-deviation rules with respect to the important notions of consistency and duality.

We study the modifications to decay amplitudes in heavy to heavy semileptonic decays with multiple hadrons in the final state due to intermediate heavy hadrons being off-shell or having a finite width. Combining Heavy Hadron Chiral Perturbation Theory (HHχPT) with a BCFW on-shell factorization formula, we show that these effects induce O(1/M) corrections to the standard results computed in the narrow-width approximation and therefore are important in extracting form factors from data. A combination of perturbative unitarity, analyticity, and reparameterization invariance fully determine these corrections in terms of known Isgur-Wise functions without the need to introduce new form factors. In doing so, we develop a novel technique to compute the boundary term at complex infinity in the BCFW formula for theories with derivatively coupled scalars. While we have used the B¯ → Dπℓν decay as an example, these techniques can generally be applied to effective field theories with (multiple) distinct reference vectors.

ABSTRACTThis paper proposes a novel model predictive control (MPC) for five‐phase fault‐tolerant permanent magnet motor (FTPMM) using an improved vector selection. Unlike conventional MPC approaches, this method does not rely on the deadbeat control principle. It effectively reduces the computational burden of the algorithm while enhancing the motor drive performance. The scheme determines the optimal output vector and calculates the corresponding action times based on the concept of synthesizing two adjacent virtual vectors (VVs). Firstly, a VVs control set is adopted to improve the current control accuracy. Then, an improved vector selection method is introduced to obtain the reference vector directly by table lookup method without computational derivation. This eliminates the dependence on the deadbeat control principle found in traditional methods. Then, a two‐step optimization method is used to obtain the optimal vector combination and to calculate the duty cycle. In the two‐step optimization process, constraints are introduced to avoid duty cycle overflow. This optimization process is also independent of the deadbeat control principle. The feasibility and superiority of the proposed MPC method are demonstrated through theoretical analysis and experimental validation.

Penelitian ini membahas penggunaan polinom Jacobi dan sifat t-design pada kode linear. Tujuan utama dari penelitian ini adalah untuk membangun program menggu nakan Python dan SageMath guna menghitung polinom Jacobi dari kode linear dengan berbagai reference vector. Metode yang diterapkan melibatkan analisis kode self-dual pada berbagai lapangan hingga mendapatkan polinom Jacobi untuk beberapa kondisi kode. Hasil menunjukkan bahwa kode yang diuji tidak memenuhi kriteria t-design karena variasi reference vector menghasilkan polinom Jacobi yang berbeda. Studi ini memberikan wawasan tentang hubungan antara kode linear dan polinom Jacobi serta potensi pengembangan lebih lanjut dalam eksplorasi kode yang lebih kompleks untuk memenuhi t-design.

Abstract This paper addresses the problem of algorithmic prediction of protein folding pathways, namely, the transient three-dimensional conformations of protein molecules during folding, under constrained rates of entropy change. We formulate the physics-based prediction of folding pathways as a control synthesis problem, where the control inputs guide the protein folding simulations. These folding control inputs are obtained from large-scale trust-region subproblems (TRS) utilizing a computationally efficient algorithm with no need for outer iterations. The proposed control synthesis approach, which leverages the solutions obtained from a special generalized eigenvalue problem, avoids potentially cumbersome and unpredictable iterative computations at each protein conformation. Moreover, the TRS-based control inputs align the closed-loop dynamics closely with the kinetostatic compliance method (KCM) reference vector field while satisfying ellipsoidal constraints on the folding control inputs. Finally, we provide conditions for existence and uniqueness of the resulting closed-loop solutions, which are the protein folding pathways under constraints on the rate of entropy change. Numerical simulations utilizing the KCM approach on protein backbones confirm the effectiveness of the proposed framework.