№ 2 (2023)
Surveys
Methods for Solving Some Problems of Air Traffic Planning and Regulation. Part II: Application of Deep Reinforcement Learning
Аннотация
Following part I of the survey, this paper considers the problems of improving the safety and efficiency of air traffic flows. The main challenge in conflict detection and resolution by traditional optimization methods is computation time: tens and even hundreds of seconds are required. However, this is not so much for response in real situations. Deep reinforcement learning has recently become widespread due to solving high-dimensional decision problems with nonlinearity in an acceptable time. Research works on the use of deep reinforcement learning in air traffic management have appeared in the last few years. Part II focuses on the application of this promising approach to the following problems: detecting and resolving aircraft conflicts, reducing the complexity of air traffic at the national or continental level (a large-scale problem), and increasing the efficiency of airport runways through the improved planning of aircraft landings.
Control Sciences. 2023;(2):3-18
3-18
Analysis and Design of Control Systems
Interval Observer Design for Discrete Linear Time-Invariant Systems with Uncertainties
Аннотация
This paper considers the problem of constructing an interval observer for systems described by discrete-time linear models under uncertainties in the form of exogenous disturbances and measurement noise (unknown bounded functions). Such an observer is designed using the minimal-dimension model of the original system invariant with respect to the disturbances. The dynamic matrix of this model is defined in the identification canonical form. We present relations to design an interval observer of minimal complexity for estimating the set of admissible values of a given linear function of the state vector. If the observer invariant with respect to the disturbances does not exist, we suggest a method to construct an observer with minimal sensitivity to them based on the singular value decomposition of system matrices. Theoretical results are illustrated by an example.
Control Sciences. 2023;(2):19-27
19-27
Terminal Control of Moving Objects in the Classes Of Piecewise Constant and Piecewise Continuous Functions
Аннотация
This paper presents a terminal control problem with the separation of object’s state coordinates into two types: the slowly changing coordinates figuring in boundary conditions and the coordinates of the stabilization loop. A predictive model of the object is introduced to design the control action. A differential equation is derived for predicted mismatches in the boundary conditions. The original system is discretized in time based on this equation. This problem is solved step-by-step in the classes of piecewise constant and piecewise continuous control actions. As an illustrative example, the problem of controlling the fuel consumption of a stage of a liquid-propellant launch vehicle is considered. The class of control actions is extended from piecewise constant to piecewise continuous functions in order to cover additional requirements for the control process. The continuous functions on intervals between control jumps are chosen using the local boundary conditions obtained during the terminal control design in the class of piecewise constant functions.
Control Sciences. 2023;(2):28-36
28-36
Control in Social and Economic Systems
Models of Joint Dynamics of Opinions and Actions in Online Social Networks. Part I: Primary Data Analysis
Аннотация
Based on VKontakte data, we study the influence of various factors on the dynamics of opinions and actions both at the macro level (“public opinion”) and at the micro level (the opinions and actions of individual agents). Primary analysis results are presented for the dynamics of opinions and actions of agents in this social network. In particular, the growing polarization of opinions at the macro level is detected; changes in the opinions of agents over time are observed; socio-demographic characteristics of agents who changed their opinions are determined; a good consistency between the opinions and actions of agents is revealed; finally, an explicit relationship between the opinions and actions of agents is established.
Control Sciences. 2023;(2):37-53
37-53
Control of Technical Systems and Industrial Processes
Creating Feature Spaces and Autoregressive Models to Forecast Railway Track Deviations
Аннотация
Diagnosis of railway tracks reveals the deviations of rail parameters in the plan and profile from their nominal values. If the deviations approach the limit values, the speeds of trains must be reduced. Therefore, forecasting changes in the deviations is a topical problem. Despite the significant amount of diagnostic data collected, railway operators underuse machine learning methods to improve the quality of prediction. The proposed approach differs from known counterparts as follows. First, the dimensionality of the feature space is increased by calculating the variation of the amplitudes of deviations from the nominal values and two types of areas (the deviation length multiplied by the amplitude and the deviation length multiplied by the variation of the amplitude); subsequently, this space is represented in the 3D matrix form. Second, a set of control parameters is formed; it includes the time and space discretization step, the type of seasonal fluctuations, the number of trend change points, etc. Third, the deviations are predicted in groups differing in type and position along the track. Forecasting is based on minimizing the empirical risk criterion. As a result, a family of autoregressive models is obtained for each discretization interval along the length of the railway track.
Control Sciences. 2023;(2):54-64
54-64
Control of Moving Objects and Navigation
Inter-orbital Spacecraft Transfer: Trajectory Design by Iterating Parameter Values within a Data Grid
Аннотация
This paper considers the problem of designing an optimal inter-orbital spacecraft transfer. We present a computational algorithm and modeling results of the optimal transfer trajectory between near-Earth elliptical orbits for spacecraft with a chemical booster and fixed thrust. The trajectory design procedure includes four stages as follows: a) formation of the primary ranges of initial approximations for typical optimization problems; b) iterative integration to find the domains of convergence for a typical variational problem; c) determination of the optimal position for each problem statement within the accepted ranges and its implementation by calculating the terminal conditions residuals; d) analysis of the results obtained. We use numerical methods of mathematical analysis and mathematical programming. The risk of “overstepping” the potentially optimal result is minimized by varying the accuracy at different stages of calculations. Based on the results, we improve the primary solution of the reference problem statement, identify the domains of convergence of solutions, and obtain the sets of initial approximation vectors ensuring convergence of the considered problems for further analysis. The results of this study can be used to develop further and refine an algorithm for selecting optimal initial approximations for different optimization problems (including spacecraft trajectory optimization as a typical one).
Control Sciences. 2023;(2):65-74
65-74
Chronicle
K 110-letiyu so dnya rozhdeniya akademika Borisa Nikolaevicha Petrova
Control Sciences. 2023;(2):75-76
75-76