We consider the problem of finding the distribution law for the output signal of an aperiodic link whose input is acted upon by a random jump signal in the form of a Markov chain with two states. It has been theoretically proved that the probability density of the output signal is described by the Pearson type I distribution; this is experimentally confirmed by the results of mathematical modeling. The results obtained are used to synthesize an adaptive recognition algorithm for unknown transition probabilities in a Markov chain.

The least upper bound for the probability that a random vector with fixed mean and covariance will be outside the ball is found. This probability bound is determined by solving a scalar equation and, in the case of identity covariance matrix, is given by an analytical expression, which is a multivariate generalization of the Selberg bound. It is shown that at low probability levels, it is more typical when the bound is given by the new expression if compared with the case when it coincides with the right-hand side of the well-known Markov inequality. The obtained result is applied to solving the problem of hypothesis testing by using an alternative with uncertain distribution.

We consider a time-invariant optimal control problem of a new linear-quadratic type on the positive time axis with an amplitude constraint on the control. Using sufficient optimality conditions, we find an optimal positional control with a discontinuity on a subspace of the state space. The motion of the closed-loop system on the subspace in the sliding mode is investigated. The exponential stability of the closed-loop system is shown. Examples are given.

The tasks of behavior planning and decision-making learning in a dynamic environment are usually divided and considered separately in control systems for intelligent agents. A new unified hierarchical formulation of the problem of simultaneous learning and planning (SLAP) is proposed in the context of object-oriented reinforcement learning, and an architecture of a cognitive agent that solves this problem is described. A new algorithm for learning actions in a partially observed external environment is proposed using a reward signal, an object-oriented subject description of the states of the external environment, and dynamically updated action plans. The main properties and advantages of the proposed algorithm are considered, including the lack of a fixed cognitive cycle necessitating the separation of planning and learning subsystems in earlier algorithms and the ability to construct and update the model of interaction with the environment, thus increasing the learning efficiency. A theoretical justification of some provisions of this approach is given, a model example is proposed, and the principle of operation of a SLAP agent when driving an unmanned vehicle is demonstrated.

In this work we study a navigation problem for a nonholonomic differential drive robot operating in the environment with static and dynamic obstacles. We present a multi-phase approach to solve it, which is based on heuristic search to tackle the trajectory planning problem and specific methods of the control theory to solve the path following problem. Results of the experimental evaluation show that the suggested controller is one order of magnitude faster than the widely used in robotics Model-Predictive Control (MPC) and is capable of accurately following the reference trajectory. On the planning side we show that the suggested planner is scalable and is able to plan in reasonable time.

The paper considers Bayesian multi-stage signaling games. Previously formulated for extensive-form games, concepts of sequential equilibrium, separating equilibrium, and pooling equilibrium are specified, and calculating methods for these equilibria are also discussed. A competitive collision model with signals indicating the rivals’ states is studied as a specific example. We determine conditions for the existence of separating and pooling equilibria with ordered competition in which the competition object goes to one of the rivals without a rigid encounter. The model parameters ranges of the equilibria existence are also determined.

We present an integrated mathematical model and its implementation in the form of a software and information complex developed for joint modeling of the spread of dangerous fire factors and spontaneous evacuation of people from premises of complex configuration. The propagation of fire, heat, and smoke is modeled based on the principle of cellular automata. To simulate the evacuation process, we have developed and used a multiagent model that takes into account the physical characteristics and behavior of people during collisions.

We consider the field of three-dimensional technical vision and in particular three-dimensional recognition. The problems of three-dimensional vision are singled out, and methods for obtaining and presenting three-dimensional data, as well as applications of three-dimensional vision, are reviewed. Deep learning methods in 3D recognition problems are surveyed. The main modern trends in this field are revealed. So far, quite a few neural network architectures, convolutional layers, sampling, pooling, and aggregation operations, and methods for representing and processing three-dimensional input data have been proposed. The field is under active development, with the greatest variety of methods being presented for point clouds.

We consider a set of autonomous systems in each of which a case that is nondegenerate for a periodic solution is realized. The problem of aggregating these systems into a model containing related subsystems is solved so as to construct a system with an attracting cycle. Aggregation methods are proposed, weak coupling controls are found, and adjusted systems are constructed.

We consider the problem of fault identification in nonlinear technical systems described by dynamic models and in their sensors in the presence of disturbances. A method based on sliding mode observers is used to solve the problem. A modification of this method is proposed which permits one to expand the family of systems for which the identification problem can be solved by weakening the constraints imposed on the original system. This modification has made it possible to reduce the complexity of diagnostic tools. The results are illustrated by practical examples.