Search

Search Results

• REINFORCEMENT LEARNING METHODS IN ADAPTIVE CONTROL OF NONLINEAR DYNAMIC OBJECTS

  М.Y. Medvedev , V.K., А.R. Gaiduk , I.М. Medvedev , Е.Y. Kosenko

  2026-04-29

  Abstract ▼

  The relevance of the problem of adaptive control of nonlinear dynamic objects is ensured by the ever-increasing demands on the quality and operating conditions of technical systems. Automation and robotics lead to an increase in the complexity of the problems being solved and the need to adapt to structural and parametric uncertainty and external disturbances. In recent years, the use of reinforcement learning methods for the synthesis of adaptive control systems has gained popularity. These methods demonstrate effectiveness in controlling uncertain dynamic objects. However, there are two fundamental problems with the application of machine learning methods to control systems for dynamic objects. First, the need to ensure asymptotic stability of the desired trajectory of a closed-loop system limits the application of deep learning methods. Second, during the learning process, it is necessary that the intelligent controller does not generate controls that lead to state variables exceeding specified limits. The purpose of this article is to review and analyze recent advances in the application of reinforcement learning methods to the synthesis of adaptive control systems for nonlinear dynamic objects. Particular attention is given to Actor-Critic methods, which are structurally similar to adaptive control systems with self-adjusting parameters. Based on the analysis, the structure of an adaptive system with two-component control, including nominal and adaptive controllers, is proposed. An adaptive control algorithm is proposed, distinguished by the use of a modified Actor-Critic algorithm, distinguished by a new form of the delta error and the absence of a singularity in the neighborhood of zero. The proposed algorithm allows for a reduction in the number of adjustable parameters. The article presents conditions for Lyapunov stability of the zero-equilibrium position of a closed-loop system and an example of the synthesis and modeling of the proposed adaptive control algorithm

• APPLICATION OF THE MIXED PARAMETER ESTIMATION METHOD IN THE CONSTRUCTION OF A HOMOGENEOUS NESTED PIECEWISE LINEAR REGRESSION OF THE FIRST TYPE

  S.I. Noskov , А. P. Medvedev , I. D. Kirillov

  102-109

  2026-09-10

  Abstract ▼

  The paper is devoted to the development of an algorithm for identifying the parameters of a homogeneous nested piecewise linear regression of the first type – a model that is in demand when analyzing complex systems whose behavior cannot be adequately described by smooth functions. The review part of the work systematizes modern publications illustrating the use of piecewise linear forms in various subject areas: from modeling energy consumption and nonlinear control systems to image processing, reconstruction of genetic networks, and filtering of geophysical data. The novelty of the study lies in the fact that for the first time for this class of models an identification algorithm based on the mixed estimation method (MEM) is proposed, which allows flexible combination of two different quality criteria. It is shown that by introducing additional Boolean variables and auxiliary constraints, the original optimization problem is reduced to a standard linear Boolean programming problem, which makes it possible to use available numerical methods for its solution. The effectiveness of the developed algorithm is demonstrated on real data of the mining and metallurgical company Norilsk Nickel for 2010–2024. The dependent variable is revenue, and the predictors are the production volumes of nickel, palladium, copper and platinum. Two alternative models are constructed – using the classical least absolute deviations method and the proposed mixed estimation method. A comparative analysis shows that the second model has a slightly higher average percentage error, but significantly outperforms in the magnitude of the maximum error on the control subsample, which makes it preferable in conditions where outliers are critical. The results obtained confirm the practical value of MEM for constructing interpretable and robust regression dependencies

• THE METHOD OF ESTIMATION POSITIONS OF THE UAVS BY MEASURING THE DISTANCES BETWEEN ELEMENTS OF THE GROUP

  V.A. Kostjukov, E.Y. Kosenko, M.Y. Medvedev, V.K. Pshikhopov, M.V. Mamchenko

  2021-04-04

  Abstract ▼

  Important problems in the development of mobile robotics are the task of autonomous navigation,
  automatic movement control and providing a reliable communication channel. For navigation,
  an unmanned aerial vehicle can use its own inertial navigation system and a satellite navigation
  system. The purpose of this article is to develop a method for reducing errors in the operation
  of the inertial navigation system of UAVs caused by the presence of random and systematic errors.
  In this case, we consider the situation of a monotonous increase in the systematic error over time.
  Usually, navigation data obtained from the satellite does not contain a significant systematic error
  in determining the coordinates. However, the satellite signal may be lost for a time significantly
  longer than the period of transmission of navigation data from the satellite in normal mode.
  As a result, there is a problem of increasing the accuracy of the data received from the inertial
  navigation system. This problem is particularly relevant for group application of UAVs. When
  solving group control tasks, it becomes necessary to prevent vehicle collisions and possible collisions
  already at the stage of traffic planning. In addition, to solve a number of group tasks, such as
  monitoring the terrain, conducting rescue operations, searching for objects in a given area, and
  joint cargo transportation, individual objects of the group must move smoothly in space with great
  accuracy. This imposes more stringent restrictions on the accuracy of the inertial navigation systems
  processing and the frequency of information exchange. In this paper, we propose a method
  that allows, based on data obtained from local systems that measure the mutual distances between
  objects in a group. This information allows correct the estimates of their own coordinates in such
  a way as to reduce the standard deviation of the corrected set of points from the true positions of
  objects at a given time. The method also reduces the maximum value of the corresponding deviation
  in comparison with the original set of estimates obtained from the navigation data of the INS.
  The method is demonstrated by the example of increasing the accuracy of determining global coordinates
  in a group of UAVs.

• DEVELOPMENT AND RESEARCH OF THE METHOD OF VECTOR ANALYSIS OF EMG OF THE FOREARM FOR CONSTRUCTION OF HUMAN-MACHINE INTERFACES

  N. A. Budko, M. Y. Medvedev , A.Y. Budko

  2021-07-18

  Abstract ▼

  The paper deals with the problems of increasing the depth and increasing the long-term stability
  of communication channels in human-machine interfaces, built on the basis of data on the
  electrical activity of the forearm muscles. A possible solution is to use the method of analysis of
  electromyogram (EMG) signals, which combines vector and command control. In view of the possibility
  of random displacement of the position of the electrodes during operation, a mathematical
  model was built for vector analysis of EMG in spherical coordinates, which is invariant to the
  spatial arrangement of the electrodes on the forearm. Command control is based on gesture
  recognition by means of a pretrained artificial neural network (ANN). Vector control consists in
  solving the problem of calibrating the channels of EMG sensors according to the spatial arrangement
  of the electrodes and calculating the resulting vector of muscle forces used as an additional
  information channel to set the direction of movement of the operating point of the control object.
  The proposed method has been tested on actually recorded EMG signals. The influence of the
  duration of the processed signal fragments on the process of extracting information about the
  rotational movement of the hand was investigated. Since the change in the position of the electrodes
  between operating sessions is different, an algorithm for reassigning and calibrating the
  amplification of the EMG channels is presented, which makes it possible to use a once trained
  ANN for recognition and classification of gestures in the future. Practical application of the results
  of the work is possible in the development of algorithms for calibration, gesture recognition
  and control of technical objects based on electromyographic human-machine interfaces.

• COMPARATIVE ANALYSIS OF CENTRALIZED AND DECENTRALIZED ALGORITHMS FOR THE MOVEMENT OF MULTICOPTER-TYPE UAVS

  М.Y. Medvedev, V.K. Pshikhopov

  2022-04-21

  Abstract ▼

  The development of robotics makes their group application relevant for solving various
  tasks. The effectiveness of performing the tasks of detecting and determining the coordinates of
  objects by a group of robots significantly depends on the accuracy of maintaining a given formation.
  In this regard, the task of determining motion planning algorithms that ensure the greatest
  accuracy of maintaining a given formation is of practical interest. This article is devoted to the
  study of the accuracy of maintaining the formation of a multicopter-type UAV group using a centralized
  motion planning algorithm and a decentralized algorithm. The centralized algorithm uses
  a master UAV, which transmits its coordinates to the slave UAVs. Based on the coordinates obtained
  and the given framework of the formation, the guided UAVs plan their movement. In a decentralized
  system, neighboring UAV groups transmit their coordinates to each other, on the basis
  of which the movement of a separate UAV is planned. The accuracy of the control system is investigated
  depending on the errors of the navigation system and the frequency of updating data on the
  position of the leading or neighboring UAVs. It is assumed that the group's UAVs determine their coordinates in discrete moments of time using an external navigation system. Centralized and
  decentralized algorithms are worked out by the same motion control system. The algorithms are
  investigated in this article by numerical modeling methods. In the process of simulation, models of
  kinematics, dynamics and actuators are taken into account, as well as models for the formation of
  errors in the navigation system. It is shown that the de-centralized algorithm of group motion
  planning provides higher accuracy compared to the centralized algorithm. However, the technical
  implementation of a decentralized algorithm is more complicated from the point of view of organizing
  a group communication system. In a centralized system, data transmission from the master
  UAV to the slave should be implemented. In a decentralized system, it is required to implement
  network communication.

• SUBTRACTION OF BACKPROPAGATION INTERFERENCE BASED ON POLARIZATION IN UNDERWATER VISION SYSTEMS FOR OPERATION IN TURBID WATER

  N.А. Budko, А.Y. Budko, М.Y. Medvedev

  2022-08-09

  Abstract ▼

  The study of the sea depths in order to ensure safety, the effective use of underwater resources
  is an urgent task. The first part of the article briefly considers the physical phenomena and
  limitations that arise during the propagation of electromagnetic waves in the visible range in the
  underwater environment. It is shown that underwater vision systems (as a class of specialized
  technical vision systems - TVS) based on conventional CCD matrices face a number of fundamental
  limitations in terms of improving the efficiency of functioning in natural water of low transparency.
  In particular, the use of artificial light sources as part of underwater vision systems in turbid
  water leads to the occurrence of backpropagation interference (BPR), which leads to spurious
  illumination of the optical device matrix. As a promising direction in the development of underwater
  vision systems, it is proposed to use methods for subtracting POR based on information about
  the polarization of light. In the review part of the article, the latest achievements in this field are
  considered. The main part of the article presents the methodology for studying the proposed method
  for subtracting the POR based on a comparison of the results obtained by processing images
  with known methods for estimating the Stokes vector parameters DoLP and AoLP, which allow
  obtaining information about the degree of polarization and the prevailing polarization angles of
  the scene, respectively. The experimentally obtained results of processing an underwater scene in
  water of varying degrees of turbidity using the DoLP, AoLP algorithms and the proposed methods
  for subtracting the POR are presented. Distinctive features are the use of four rather than two
  polarization directions in calculations, as well as the original mathematical apparatus for processing
  signals from the machine vision camera matrix.

• STUDY OF PATH PLANNING METHODS IN TWO-DIMENSIONAL MAPPED ENVIRONMENTS

  М. Y. Medvedev, V.K. Pshikhopov, D.О. Brosalin, B.V. Gurenko, М.А. Vasileva, Hamdan Nizar

  2022-08-09

  Abstract ▼

  The article studies the problem of motion planning in two-dimensional mapped environments.
  The review and analysis of known planning algorithms based on Voronoi diagrams, probabilistic
  road maps, rapidly growing random trees, Dijkstra algorithms, A*, D* and their modifications, artificial
  potential fields and intelligent heuristics are carried out. Based on the analysis, it is concluded
  that classical methods in dynamic environments require significant costs in terms of calculation time
  and the amount of memory used. The conclusion is made about the relevance of the development of
  algorithms that increase the efficiency of known planning methods. In this regard, this article is devoted
  to the development of a modified algorithm of rapidly growing random trees and the study of its
  effectiveness in comparison with known methods. The article presents a modified algorithm for rapidly
  growing random trees, characterized in that when checking for a path to a new potential node of
  the tree, the path to some area near the specified node is checked. This reduces the number of nodes
  in the tree under construction. The developed algorithm is first compared with the traditional algorithm
  of fast-growing random trees. The comparison is made by the trajectory calculation time, the
  amount of memory required, the path length and the percentage of situations in which the trajectory
  to the target point was successfully found. Next, the developed algorithm is compared with the planning
  algorithms of other classes. The study uses representative samples of numerical experiments and
  various environments that differ in the density of obstacles and the presence of mazes. A study of
  planning algorithms using the results of experiments on a ground-based wheeled robot is also being
  conducted. Based on the results of numerical and real experiments, conclusions are drawn about the
  advantages and disadvantages of the developed algorithm of motion planning and the feasibility of its
  application in various environments.

• RESEARCH OF AN INTELLIGENT ADAPTIVE CONTROL ALGORITHM BASED ON THE REINFORCEMENT LEARNING METHOD

  А. N. Karapeev, Е.Y. Kosenko, М. Y. Medvedev, V. K. Pshikhopov

  2025-04-27

  Abstract ▼

  An algorithm for adaptive control of a DC motor based on the use of machine learning technology
  with reinforcement is proposed and investigated. An overview and brief analysis of the state of affairs in
  the field of intelligent motor control systems is given. A mathematical model of the DC motor is presented,
  and a structural scheme for training an intellectual agent is presented. An intelligent adaptive motor speed
  control system is proposed. The DC motor is represented as a black box with the limited input and output.
  The control system is based on a zero-order Q-learning algorithm. It is assumed that the output of the
  intelligent agent is a control applied to the motor input. The intelligent system uses a tabular approximation
  of the value of each of the control action. In this article, we study the effect of the discreteness of the
  representation of state, the set of control effects used, the applied rewards, and the parameters of the
  learning algorithm on the control error. The sensitivity of the control system to the parameters of the motor
  and an unmeasured moment is investigated. Based on the results of the study, a modified algorithm is
  proposed, which assumes the measurement or evaluation of the current of the motor stator. The control
  algorithm provides robustness to parameters and external disturbance. Additionally, the approximation of
  the control value function using polynomials and using a neural network are investigated

• CONTINUOUS CONTROL OF NONLINEAR NON-AFFFINE OBJECTS

  А.R. Gaiduk, V.K. Pshikhopov, М. Y. Medvedev, V.G. Giscov

  2024-04-15

  Abstract ▼

  The paper proposes a method for constructing continuous control of non-affine control objects
  with differentiable nonlinearities and a measurable state vector. The method is based on the use of
  quasilinear models of nonlinear objects, which are created on the basis of their equations in Cauchy form while maintaining the accuracy of the description. It is shown that control by state and influences
  exists if the nonlinear object is completely controllable by state and satisfies the criterion of
  output controllability. To determine the control, it is necessary to find a number of polynomials using
  the object model and solve polynomial and nonlinear algebraic equations. The method is analytical
  and allows us to provide some primary quality indicators. The region of attraction of the equilibrium
  position of a closed system is determined by the region of state space in which the controllability
  condition of the quasi-linear model of the object is satisfied. Depending on the nonlinearity properties
  of the object, control is defined either as a function of state and deviation variables, or is a numerical
  solution obtained by an iterative method. The required control is oriented towards implementation
  by a computing device. The article provides the formulation of the problem, the conditions for
  its solvability, as well as analytical expressions for finding the control action. A numerical example is
  given with the results of synthesis and modeling, which allows us to conclude that the above relations
  lead to finding continuous control of a non-affine object with differentiable nonlinearities and a
  measurable state vector, which ensures the required properties of a closed-loop control system

• INTEGRATION OF LOCAL AND GLOBAL SCHEDULER INTO A MOBILE ROBOT CONTROL SYSTEM

  D.O. Brosalin, B.V. Gurenko, М. Y. Medvedev

  2024-01-05

  Abstract ▼

  This paper investigates the problem of integrating local and global motion planning methods
  in a robot control system. The current level of technological development allows mobile robots
  not only to follow predetermined coordinates, but also to make real-time decisions independently
  of the operator, reacting to changes in the environment. However, the dynamic nature of the environment and the constraints on planning time, as well as the high speeds of mobile robots, complicate
  the problems solved by planning algorithms. In this paper, some motion planning methods
  based on cellular decomposition (such as A*, D* and Wavefront) and random search procedures
  on graphs (such as fast growing random RRT trees and probabilistic roadmaps PRM) integrated
  with a motion trajectory prediction algorithm (DWA) are reviewed. A study of the performance
  characteristics of each of the above algorithms has been conducted, as well as a series of numerical
  and in-situ experiments to analyze the effect of map topology on the execution time and
  memory usage of the algorithms. The effect of the speed of local and global planning under different
  configurations of the external environment was investigated. To confirm the effectiveness of the
  investigated algorithms in real conditions, software for a mobile robot based on a wheeled chassis
  has been created. The paper presents structural and functional schemes of interaction between the
  implemented modules of planning and motion control of the mobile robot and the environment.
  It also presents a mathematical model of a wheeled platform, for which, based on the considered
  methods, motion planning algorithms are developed. In this paper, quantitative measures including
  the computation time of the motion planning algorithm and the amount of memory used by the
  algorithms under different environment maps are evaluated. Both environments with randomly
  placed obstacles and different types of mazes are considered. The implementation of the developed
  algorithms in the ROS-2 environment is also described. It is shown that the implemented system
  provides real-time control and motion planning of the mobile robot.

• DESIGN OF HYBRID CONTROL SYSTEM FOR NONAFFINE OBJECTS

  А.R. Gaiduk, Ali El A. Kabalan, V.K. Pshikhopov, М. Y. Medvedev, V.G. Gistsov

  2023-04-10

  Abstract ▼

  In the theory of automatic control, an urgent problem is the development of design methods by
  nonaffine control systems. In such systems, the control affects the input of the plant nonlinearly, so it
  affects the state variables non-additively. The purpose of this article is to develop a design method that
  ensures the stability of the zero equilibrium position of a closed control system in a certain area.
  The object described by a nonlinear system of differential equations with one control is considered.
  A restriction is introduced, consisting in the differentiability of the right part of the differential equations
  for all state variables. The task of designing control in the form of a function of the reference signal, a
  vector of state variables and control values at previous points in time is set. This problem is solved using
  a quasilinear model of the control plant. This model of description allows you to preserve all the features
  of a nonlinear plant without simplifying them. In the quasilinear model, matrices and vectors are
  functions of the variables of the state of the control plant. The control is performed using an algebraic
  polynomial matrix method. This method allows you to find control when the control condition of the
  plant are met in the form of inequalities. This article presents the expressions for calculating the control
  according to the polynomial matrix method. Based on the given coefficients of the desired polynomial,
  as a result of solving an algebraic system of equations, coefficients are found that are a function of control
  and state variables. At the same time, the fulfillment of the controllability condition guarantees the
  existence of a solution of the specified algebraic system. An expression has been found that allows calculating
  the control by the coefficients found. The article also finds a condition for the possibility of
  providing a non-zero value of the output controlled quantity of a nonlinear Hurwitz system in a steadystate
  mode. Under this condition, a zero value of the static error for the setting effect can also be provided.
  Further, the transformation of the obtained continuous control into a discrete one is proposed, which
  is implemented in a digital computer. The article also provides a numerical example of the control design
  of a second-order nonlinear control and the results of modeling a closed nonaffine system.
  The given example confirms the theoretical results obtained. Thus, the proposed approach makes it possible
  to design stable Hurwitz control systems for nonaffine objects using the algebraic polynomial matrix
  method with sufficiently small sampling periods of variables of the control object and small modules of the
  roots of the characteristic polynomial of the matrix of a closed system in its quasilinear model.

• THE CONCEPT OF CREATING A PROMISING SYSTEM FOR RECHARGING AN AUTONOMOUS GROUP OF UAVS

  М.Y. Medvedev, V.А. Kostyukov, М.Y. Butenko, V.G. Gistsov, I.D. Evdokimov

  2023-02-17

  Abstract ▼

  Due to the accelerated growth in the use of groups of autonomously functioning unmanned aerial
  vehicles (UAVs) in various environments, solving the problem of optimizing the functioning of groups of
  such vehicles according to the criterion of the minimum energy consumed is an urgent scientific task. In
  this article, a new approach is being developed to ensure energy saving of a group of unmanned aerial
  vehicles (UAVs) by using UAV charging stations that provide the necessary versatility in servicing different
  types of vehicles. The most effective variants of landing systems, precise positioning, power generation
  at the station are considered, and a hybrid charging system combining contact and contactless methods is also justified. A generalized scheme of multi-stage interaction of the UAV with the charging
  station is given, which provides for the possibility of repeating one of the stages if it is not fully completed
  within certain time intervals, and also takes into account the peculiarities of interaction between
  agents through the communication channels available to them. The problem of optimal distribution
  according to the energy criterion of the group's UAVs between destinations (PN) combined with charging
  stations (SP) has been set and solved. Both the cruising mode and the maneuvering of the vehicle
  during takeoff and landing, when the power consumption increases, are taken into account. The concept
  of the effective distance to the destination is introduced, taking into account the estimated energy costs
  of each UAV to reach this destination, taking into account its arbitrary current position and the available
  queue of tasks at the moment. To study the developed approaches and algorithms for targeting and
  planning the movement of the group's UAVs, software based on the Unity environment was created and
  tested. The flexibility of the latter allows modeling various algorithms of information interactions
  of elements within a group of UAVs, a group of charging stations, as well as cross-interactions
  between UAVs and charging stations. In particular, the software allows you to determine at each
  discrete moment the degree of charge of each UAV, the queues of destinations for each UAV, its
  history of recharge at stations.

• THE METHOD OF ESTIMATION POSITIONS OF THE UAVS BY MEASURING THE DISTANCES BETWEEN ELEMENTS OF THE GROUP

  V.A. Kostjukov, M.Y. Medvedev, V.K. Pshikhopov, E.Y. Kosenko

  2021-04-04

  Abstract ▼

  Currently, the active use of groups of robots has begun to solve a number of tasks for civil
  and military purposes. In this regard, problems arise associated with group management, the organization
  of reliable communication channels and ensuring the effective functioning of the group
  with limited energy resources. When solving the problem of optimizing energy consumption, the
  problem of increasing the efficiency of interaction of the elements of the group with stationary
  charging stations arises. This problem can only be solved by considering an integrated system,
  which includes robots and charging stations. Centralized management of such a system is justified
  in the case of a small number of its elements. However, with an increase in the number of elements
  in a group, the complexity of management increases, so a combination of centralized and decentralized
  management methods becomes a higher priority solution. The complex of problems of
  decentralized management of such a group includes the task of organizing the optimal interaction
  of its elements in order to achieve the goal of its functioning. When organizing energy exchange
  between robots and charging stations, solving this problem plays a key role in optimizing energy
  consumption. In this article, the concept of the interaction of mobile and stationary objects is developed,
  which implies the possibility of each agent choosing an appropriate companion for interaction.
  This choice is made taking into account the current state of the system and the assessment
  of the history of interaction results. The developed concept is detailed for a system that includes
  UAVs and their recharging stations. An algorithm is proposed for the decentralized selection of
  pairs of interacting elements "UAV - charging station" based on two indicators - the energy efficiency
  of the charging process, and the time spent by the UAV to reach the target point. Both indicators
  are taken into account when choosing the weights assigned to each charging station as its
  degrees of efficiency. Also, these indicators are included in the optimized quality criterion. An
  optimization procedure has been developed, the result of which is the number of the charging station
  that is most suitable for a given mobile object for interaction.

• CONEPT OF A ROBOT GROUP CALCULATION

  V.Kh. Pshikhopov, A.R. Gaiduk, M.Y. Medvedev, D.N. Gontar, V.V. Solovjev, O.V. Martyanov

  2020-07-10

  Abstract ▼

  The problem of calculation of an autonomous robotic group in order to destroy the detected enemy group is considered. A group of robots must be formed in such a way that the task assigned to it to destroy the enemy group is performed with a high degree of probability. The task is solved as an assignment problem. The initial information for solving this problem are types and numberof objects of the detected enemy group, positions of the enemy objects, information about the war possibilities of the tools available in our group, the type of group being formed (robotic or mixed), the purpose of the operation, the actions of the group at the end of the operation. We propose a solution to the problem based on the evaluation of the effectiveness of individual robotic systems. The solution is formulated as a sequence of the four stages. At the first stage, the calculation of a priori effectiveness of each element of the detected enemy group is performed. At the second stage, based on expert assessments, the choice of efficiency coefficients for each of the available robotic systems against each element of the detected enemy group is made. At the third stage, a priori estimates of the effectiveness of the available robotic systems are corrected, taking into account the coefficients selected at the second stage. At the fourth stage, a group of robotic systems is formed in such a way that its total application efficiency exceeds the total application efficiency of the detected enemy by 2.0–2.5 times. The proposed method of forming a group allows you to cre-ate both quantitative and qualitative composition of the group. The article provides an example of the formation of a group whose goal is to neutralize an exposed enemy.