This paper studies multistability and hidden oscillations in a four-dimensional Keldysh model of flutter suppression. The model is written in the Lurie form and is considered with discontinuous damper characteristics interpreted in the sense of Filippov. We analyze the stationary set, the sliding manifold, and the corresponding reduced sliding dynamics. To detect undesired periodic solutions, we apply the locus of a perturbed relay system (LPRS) method and extend it to a discontinuous piecewise-linear approximation of the original Keldysh nonlinearity. It allows to reveal symmetric and asymmetric stable periodic solutions, including periodic solutions with complicated switching configurations. Numerical continuation from the discontinuous piecewise-linear characteristic to the original discontinuous characteristic with a quadratic term detects five stable periodic solutions, four of which are hidden with respect to the rest segment. These results show that the four-dimensional Keldysh model has rich and complex nonsmooth dynamics and provide a constructive framework for estimating the boundary of global stability in discontinuous aircraft control systems.
A simulation technique is described that was applied to optimise the design and ensure the desired field quality of a superconducting solenoid. A three-section solenoid with a lengthwise split winding was chosen as a reference configuration. The design was optimized through MathCAD calculations with the Levenberg-Marquardt minimization. The calculated data enabled adjustment of the coil geometry so as to provide required field homogeneity with respect to manufacture limitations. The optimised design was validated in comparative simulations and measurements on a prototype. The study was performed within the R&D activities on construction of an original electron beam ion source (EBIS). The results of the study were reported in the project documents.
This paper presents semantic inversion of pretrained image classifiers as an experimental probe of the view that late neural representations behave as class- dependent error-correcting codewords. The method reconstructs ImageNet validation images from late classifier activations by optimizing pixels to match one tail-proximal feature distribution under a symmetric Kullback–Leibler (Jeffreys) objective. The protocol is not a generative-prior attack and not a data-free synthesis method: it uses only the frozen classifier, a smoothed optimization path, a feature-mean constraint, a total-variation prior, and a class-logit KL constraint. The experiment evaluates 21 Torchvision models spanning RegNet, ResNet, EfficientNet, ConvNeXt, and ViT families on 48 class-capped high-margin validation samples per model. Reconstruction quality is measured objectively with OpenCLIP ViT-L/14 image-image retrieval metrics, feature-KL reduction in bits, and top-1 classification accuracy of the reconstructed images. The results show strong semantic recovery for ViT-B/16 and EfficientNet models, moderate recovery for RegNet models and ConvNeXt-Tiny, weak retrieval for the larger ResNet and ConvNeXt variants, and adversarial nonsemantic failures for ResNet18 and ResNet34. Qualitative reconstructions further illustrate that successful recoveries preserve class-level form and color without copying exact pixel-level appearance.
The paper develops a stochastic discrete-time framework for practical fixed-time consensus. Constructive estimates are obtained for the practical consensus set and for the expected entrance time into this set. A comparison-based argument is also established to relate the stochastic dynamics to an ideal noise-free model and to quantify their finite-horizon closeness.
This paper investigates a nonlocal control problem of a linear non-stationary dynamic system described by an ordinary differential equation subject to an integral constraint imposed on the components of the state vector over a certain subinterval of the system’s operating time. Conditions for complete controllability of the dynamic system under the imposed integral constraints are formulated. Conditions ensuring the existence of a solution to the control problem with such a nonlocal constraint are obtained. A constructive approach to solving the control problem is proposed, and the corresponding solution is constructed. The continuity and non-uniqueness of the control function are demonstrated. As an illustration, a control problem for the oscillatory motion of a material point with classical boundary conditions and an integral condition is solved.
The problem of the equidistant deployment of a group of mobile agents on a line segment is studied. It is assumed that agent dynamics is modeled by double integrators, and each agent receives information on its velocity and distances to some of its neighbors. New nonlinear nonhomogeneous decentralized control algorithms are proposed providing agent convergence to the prescribed positions in a fixed time for any initial conditions. The theoretical results are illustrated via computer simulations.
Iterative learning control (ILC) applies to systems that operate in repetitive mode. The goal of control is to ensure that the system’s output follows a reference trajectory with a specified accuracy. In ILC literature, each repetition is called a trial, and its finite duration is known as the trial length (also, the terms “pass” and “iteration” are commonly used). With each repetition, ILC should reduce the difference between the reference trajectory and the output signal called trial-to-trial error , ideally bringing it to zero. In this paper, a new ILC design method is proposed for multi-agent discrete systems based on the application of gradient optimization within 2D models in combination with the method of vector Lyapunov functions for repetitive processes. The new method yields a simple ILC structure that accelerates the convergence of the trial-to-trial error. An example confirming this property is provided.
Edge multiagent systems call for operating-system-level agreement primitives that tolerate peer failures, lossy and unsynchronized messaging, and heterogeneous local state, rather than relying on a permanent timing master. Neighbor exchange is modeled as a time-delay multi-agent system under bounded communication delays. This article presents two complementary building blocks—monotone logical-time synchronization via max/SoftMax consensus with PI-style implementation, and consensus-based accelerated distributed simultaneous perturbation stochastic approximation (ADSPSA) for cooperative tracking under zeroth-order oracles— together with a multimodal composition principle: logical-time and spatial coordinates are stacked per agent, coupled through multimodal losses and a block preconditioner on the ADSPSA channel. Formal convergence analysis for the logical-time synchronization primitive is developed in companion works on monotonic logical time (an accepted ApPLIED 2026 contribution and a manuscript submitted to IEEE access); fully coupled multimodal validation—joint logical-time and tracking experiments under delays and losses—remains outside the scope of this paper. For the Ψ-preconditioned recursion we develop a majorization framework that yields an asymptotic upper bound on the tracking-error covariance under explicit smoothness, centering, and spectral-contraction assumptions. Numerical experiments on a heterogeneous quadratic surrogate—isolating the tracking block but using a time/tracking block structure for Ψ—illustrate how an empirically tuned preconditioner can lower the asymptotic error level relative to the identity map, sometimes at the cost of slower transient decay.
Currently, new electronic generators of neuron-like activity are being developed, which can be used both to build spike neural networks and, in the longer term, to solve the problems of neuroprosthetics and creation of artificial life. At the same time, the mathematical description of such artificial neurons is usually qualitative or fragmentary. Though writing equations from the first principles (e. g. from the Kirchhoff’s laws) is still the main tool, there is another useful approach — system identification (model reconstruction) from experimental series. Actually, a combination of both approaches can give even more than each of them separately. The purpose of this work is to provide a new specific approach for identification of models which can to some extent be described by a generalized van der Pol oscillator. Since we propose this approach for a specific experimental device (electronic neuron) and test using its time series, we focused on some practical factors. First, we abandoned theoretical formulae for dissipation and potential functions and reconstructed them in the most general form without any additional assumptions. Second, we switched to equations integrated in time to reduce the impact of the measurement noise. Then, we considered the possibility of hysteresis for the dissipation function. Finally, we tested the approach based on series from three different instances of the same generator. We showed that the proposed approach can be useful to obtain the equations of the setup which really match the observed data.
A sequence-based deep learning pipeline is proposed for automated processing of gated SPECT studies of the right ventricle. The model processes a temporal sequence of 3D tomographic volumes and predicts right-ventricular myocardial masks for each cardiac phase. Based on the predicted segmentation masks, clinical parameters are computed and perfusion polar maps of the “bull’s eye” type, as well as magnitude and phase planar maps, are constructed. In experiments, the recurrent model achieved a Dice score of up to 0.8119.
Binary phase-manipulated probe signals (BPM, or BPSK) are widely used for echo detection and pulse compression. Reliable detection of the full return time of a long probe requires a sharp matched-filter peak at the correct delay and near-zero responses at other delays. This requirement leads to the synthesis of long binary sequences whose aperiodic autocorrelation has a dominant zero-shift peak and very small sidelobes. We formulate the task as a discrete optimization problem over length N sequences taking values in the set {−1, +1} and propose a practical construction strategy that combines (i) exhaustive enumeration of near-optimal short blocks, (ii) symmetry augmentation (reversal and sign inversion), and (iii) greedy/beam splicing to build long sequences. The method is simple to implement, naturally parallelizable, and improves the normalized sidelobe-energy objective (phi), ISL, and PSL over an optimistic random baseline (best-of-200 trials), while a genetic-algorithm baseline can reach lower ISL at a substantially higher number of objective evaluations.
The paper proposes algorithms for processing dermoscopic images: a preprocessing algorithm aimed at detecting noise on the image (hair structures and immersion gel bubbles) and subsequent restoration of color characteristics of the noisy skin areas; and a region of interest extraction algorithm that takes into account the specifics of dermoscopic images (possible low contrast in RGB space, characteristic of some types of skin lesions, preservation of hair fragments after preprocessing). The proposed hair detection method is based on a combination of directional Gabor filters and Laplacian of Gaussian filters. This hybrid approach allows for the detection of both thick dark hair structures and thin light ones, demonstrating robustness to the properties of color, direction, and thickness of the detected objects. To minimize false positives at lesion boundaries, an additional geometric analysis of the mask using an elliptical test is proposed. This stage allows for an automatic decision on the need to apply the inpainting procedure, which helps preserve information about the texture of the skin lesion on weakly noisy images.
This paper presents an analysis of the applicability of a quantum amplitude redistribution algorithm to the data filtering problem and the results of modeling the algorithm's operation in comparison with a median filter.
We consider a planar circular array of ultrasound emitter-receiver elements. Due to manufacturing reasons, the actual positions of the elements slightly differ from the ideal equidistant positions exactly on the circle; also, there exist delays in emitting and receiving the signals. This leads to corrupted resulting ultrasound images, and these misplacements and delays are to be evaluated in order to obtain better images. It is assumed that the only available information that we possess is the noisy measurements of the times between the instant of activation of every sensor and the instant of registration of the signal at the receiver. Some of the existing approaches to this calibration problem have various drawbacks such as very high dimensions of the associated convex optimization problem or convergence to local minima in nonconvex formulations, etc. To solve this problem, we developed a very simple iterative procedure that requires solving moderately-sized systems of linear equations. At every iteration, we first optimize by a part of variables and then use their updated values to optimize over the rest of the variables. With simple tricks, both problems are converted into linear ones, thus making solution very fast. Preliminary experiments over synthetic data testify to a rather promising performance of the method.
Methods for analyzing the spatial motion of a satellite in geostationary orbit during long-term conservation with episodic activation of control have been developed. The computer simulation results of a geostationary communications satellite motion during the year with quarterly correction are presented.
This paper presents the development of an additional module for the mechatronic vibration stand SM-2M at the Institute of Mechanical Engineering of the Russian Academy of Sciences. The module is designed to conduct experimental studies of various technological processes, including sieving, material mixing, and granulation. To evaluate the dynamic characteristics of the system, a virtual model of the equipment was developed, which includes elements of technological tooling represented by a working table accounting for its own mass. The model was implemented using the ADAMS.View software package, providing capabilities for visualization and dynamic analysis of the system. Testing of the novel mechanical design was carried out via co-simulation methodology, integrating a proportional-integral control algorithm implemented in MATLAB/Simulink with the physical model of the mechanical part realized in ADAMS. Several virtual experiments addressing the control of rotational speed and angular displacement of the drive motors of the vibrational setup equipped with technological tools and bulk materials are presented. Graphical dependencies depicting variations in motor rotation frequencies and trajectories of motion of the center of mass of the working components based on virtual sensor readings are provided. The results facilitate implementation of modern control methods, enhancing technology reliability and efficiency.
Standard compressed sensing (CS) theory typically assumes that noise is bounded in ℓ2 -norm (e.g., Gaussian). In practice, noise can be unknown-but-bounded (for example, in low-light imaging or MRI artifacts). In this work a new CS recovery algorithm for parameter estimation under unknown-but-bounded noise is proposed. Experiments on images with various non-Gaussian noises demonstrate that proposed method outperforms classical ℓ2 -constrained recovery.
We continue to consider triangulation algorithms. To the subject areas in which triangulation is used, we can add, for example, the task of clustering sensor data by location (spatial clustering) is to group data points located in the same region based on their density. The goal is to identify clusters that are close to each other. In the previous paper, the aspects related to the possible parallelization of the heuristic algorithms proposed by us were mostly considered. The main purpose of this paper is different: the main heuristic considered here is to improve the already obtained preliminary location of a certain point in the “bad case”, i.e., when we obtain from further calculations, that such an arrangement is unsuccessful; this is backtracking. We provide a general description of the triangulation algorithm we use. However, the algorithms we are considering are not limited to one description of backtracking. We use several options for choosing the next point to consider: either we take an arbitrary one (usually the next in number), or the one closest to the geometric center of the points already located, or the one furthest from this center. A simple example of such an algorithm is given. However, the main ideas of backtracking are noticeable in such a small example. Since we need to consider a wide variety of noise variants superimposed on the geometric arrangement of points, we, as in previous publications, consider geometric distances and multiply them by another implementation of independent identically distributed random variables with a mathematical expectation equal to 1 and different variants of the standard deviation. These standard deviation variants in our constructions vary from 0 (so-called geometric version) to 0.14. The end of the paper is devoted to a description of the experimental results and their understanding.
We invent a continuous field model for epileptiform dynamics and focus on the spatial evolution of the hypersynchronized ictal phase in the form of a scalar field. We add a control term to its dynamical equation to study the possibility of suppressing the epileptiform regime at the mesoscopic scales. We reproduce the exact analytical solutions for the hypersynchronised phase in two forms: with the separation of variables and in the shape of traveling waves. Then we conclude our results and discuss the possible applications and further developments of our model.
The study investigates a model of the dynamics of a two-rotor vibrating system (VS) with rotors of different masses and a non-stationary elastically attached mass. It is assumed that the supporting platform of the VS moves in the vertical plane, taking into account the rotation angle. Using computer modeling, the influence of rotor non-identicality on the stability of the frequency-coordinate synchronization regime is analyzed under various loading conditions. A special synchronization control algorithm, synthesized using the speed-gradient method, is employed in the work to ensure a stable synchronous regime.
