This paper addresses the problem of reliable object detection for robotic depalletizing in real-world industrial environments, where factors such as occlusions, dust, and lighting variability significantly affect perception systems. Building on previous work based on segmentation and geometric post-processing, we propose a simplified, task-aligned approach using YOLO26-OBB for direct oriented object detection. The proposed method replaces the multi-stage approach with a single-stage detector capable of predicting oriented bounding boxes, enabling direct estimation of object position and rotation required for robotic grasping. To evaluate the impact of annotation strategies, three dataset variants were introduced: a multi-class dataset including occlusions and pallets, a single-class dataset of all visible bricks, and a task-aligned dataset containing only top-layer (pickable) bricks. Experimental results demonstrate strong detection performance across all variants, with the task-aligned setting providing the most practically relevant and most stable results for the nano and small model variants. The results indicate that aligning dataset annotations with actionable robotic constraints can improve practical usability and reduce unsafe detections. Compared to the previous segmentation-based approach, the proposed method simplifies the perception process while maintaining high detection quality. The findings suggest that task-oriented detection formulations are crucial for the deployment of reliable robotic perception systems in industrial environments.

The paper focuses on education in the field of process control and automation. The controller design based on the principles of predictive control - MPC (Model Predictive Control) is shown. The proposed discrete-time controller with a state observer is used to control a 4th order electric system and the influence of selected parameters on the behaviour of the closed loop is shown. For comparison, control by a discrete-time LQ controller with asymptotic set-point tracking is presented. The control is implemented using the laboratory system RCDue, which allows the implementation of the proposed controllers in the MATLAB environment. The paper is related to previous papers devoted to the description of the laboratory system RCDue and its use for the identification of the controlled system model and the implementation of the discrete-time state observer and the LQ controller.

The paper is devoted to education in the field of process control and automation. It shows the design of an LQ controller, which includes a state observer for a general discrete-time state-space model. The proposed controller with state observer is applied to the control of a 4th order electric system and the influence of selected parameters on the behaviour of the closed loop is shown. For comparison, the control is presented by a discrete-time PID controller. The laboratory system RCDue is used to implement the control, allowing the implementation of the proposed controllers in MATLAB environment. The paper is related to previous papers devoted to the description of the laboratory system RCDue and its use for the experimental identification of the mathematical model of the controlled electric system.

The simultaneous optimization of neural network topology and training remains an underexplored research direction’ despite its potential to improve model efficiency and performance dynamically. This paper introduces a control-based framework for jointly adjusting the structure and training process of fully connected neural networks. The methodology formulates the training and pruning process as a multivariable dynamic system with two input variables-training process parameters and network architecture adjustments-and two output variables-model performance and computational complexity. A discrete two-dimensional Proportional- Integral- Derivative (PID) controller is employed to regulate these inputs, ensuring a balanced trade-off between accuracy and computational efficiency. The control system is tested on a function approximation task, where a fully connected network is initially set with redundant capacity and gradually optimized according to predefined reference trajectories of performance and complexity. Experimental results demonstrate the effectiveness of the proposed approach, revealing the dynamic interaction between topology and training in real-time network adaptation. The findings highlight the feasibility of integrating control strategies into neural network optimization and pave the way for future research on more advanced control-based learning architectures.

Nowadays, predictive control methods are very popular and are often used to control various systems. Linear predictive control techniques are relatively well discovered and standardised, but this is not true for predictive control methods used for nonlinear systems, which are far more frequent in nature. There is no universal approach to handle it. Nonlinear model predictive control (NMPC) is an extension of the linear model predictive control (MPC) method, which is widely used for solving nonlinear control problems. In this paper, a real-time adaptation of linear MPC for nonlinear systems is presented. This method uses nonlinear model dynamics with the advantage to provide a precise prediction of the system response to initial conditions, known future and estimated progress, and an inside-one-step evolving linearised discretised nonlinear model for predicting response to estimated control input deviance used for optimisation. The functionality of the proposed method is shown using MATLAB simulation and demonstrated by experiment on a fast nonlinear magnetic levitation plant. The proposed controller is compared with a simple control loop with PID with better control results of NMPC at the cost of higher computational complexity. In this work, we present the NMPC method that collects many interesting findings from other authors who dealt with predictive controllers and adds our knowledge, which leads to a standardised predictive control method suitable for controlling linear or nonlinear systems with state-space representation.

Predictive control is a very general method of designing an optimal controller, allowing the use of different criteria, but also a large number of different models of the controlled system. In this paper, the derivation and real application of two types of predictive controller are shown - for internal description (expressed by a discrete-time state-space model) and for external description (expressed by discrete-time transfer function). In an ideal simulation both controllers give identical results, but in real conditions different behavior will be displayed. The aim of this paper is to demonstrate the differences between the two solutions from an application point of view, to discuss the strengths and weaknesses of both types of predictive controllers and to compare their behavior in real conditions. The result is not the “best” controller but understanding their differences and facilitating the selection of the right form for a specific application.

Since its first demonstration more than a half century ago, magnetic levitation (MagLev) has gained eminent scientific attention from both the fundamental and applied points of view. In essence, MagLev shows highly nonlinear dynamics, described with nonlinear differential equations. Thus, in order to exploit the MagLev phenomenon, both mathematical models and control algorithms must be constructed. Frequently authors use simplifications of the model, and in doing so, limit the application of the MagLev model around a nominal operating point. In these simplified cases, the MagLev models may contain parameters that are not represented by proper physical quantities. Thus, in this work, we revised the issue of MagLev modelling from the first-principle approach. More specifically, we theoretically derived expressions for the interaction between the magnetic fields of the solenoid and a small magnetic object. The behaviour of the inductance on a distance from the solenoid was then described. The suggested MagLev modelling concept was verified experimentally, confirming the validity and correctness of the proposed MagLev mathematical model. The results presented here could thus be regarded as highly beneficial for formulating more complex MagLev designs exploitable in the field of model predictive control of the position of a levitating object.

The authors are presenting Process Control Laboratory at the Faculty of Electrical Engineering and Informatics, University of Pardubice. The laboratory is equipped with six GUNT training systems covering common technological variables such as level, flow, pressure, temperature, speed, and position. The authors created SW support for MATLAB and Simulink environment. An internal part of the GUNT training systems is the LabJack U12 data acquisition card. LabJack Dynamic Link Library (DLL) is used to operate training systems from MATLAB. Authors created the M-S function to allow experimenting from Simulink. The possibilities of the proposed solution are demonstrated in several control applications.

The Model Predictive Control (MPC) concept and its realization is explained together with some real-world laboratory application examples. Usually, future control errors and control action changes are penalized in the cost function (objective) by predictive controllers. The cost function can be seen as a function of future control actions utilizing the process model in a form of the predictor. A derivation of the predictors for the transfer function (external) and state-space (internal) model is indicated in the paper. An analytical solution to the given optimization problem is possible in an unconstrained case. A quadratic programming strategy must be used in case of the occurrence of the process constraints that should be respected by the controller. The authors apply both algorithms to two types of dynamical systems—to proportional (stable) system and to integrating (unstable) system and they demonstrate the influence of the penalization parameters (weights) in the cost function on the control quality. The authors aim to highlight the MPC strategy and its potential, and on the other hand, mention some bottlenecks and risks associated with model-optimisation-based methods.

The paper is devoted to the education and teaching of process control and automation. Various laboratory equipment is used to explain and better understand the theory and to gain practical experience. The authors have designed and developed a simple electrical dynamical system RCDue (dynamic model with passive RC components and Arduino Due as measurement and communication unit) that allows students to perform various laboratory experiments – e.g. static and dynamic characteristics measurements, modeling, experimental identification, control design and application of from the simplest strategies to advanced methods. Specifically, in this paper, the authors focus on experimental identification.

Model Predictive Control (MPC) is an advanced process control method that is widely used for controlling both linear and under some modifications for non-linear systems. The aim of this work is to show a way how to apply MPC to a non-linear Magnetic Levitation System (MLS) and its capability of stabilization and closed-loop performance. This work is a continuation of the previous article where the laboratory plant CE 152 MLS was identified, and a non-linear model was designed. This paper proposes a control circuit consisting of linearized discretized non-linear MLS model, Extended Kalman Filter (EKF) algorithm for state estimation and linear MPC. The results are verified in simulation and real-world experiment.

Microplastics, i.e. synthetic polymers that have particle size smaller than 5 mm, are emerging pollutants that are widespread in the environment. In order to monitor environmental pollution by microplastics, it is necessary to have available rapid screening techniques, which provide the accurate information about the quality (type of polymer) and quantity (amount). Spectroscopy is an indispensable method, if precise classification of individual polymers in microplastics is required. In order to contribute to the topic of autonomous spectra matching when using spectroscopy, we decided to demonstrate the quality and efficiency of neural networks. We adopted three neural network architectures, and we tested them for application to spectra matching. In order to keep our study transparent, we use publicly available dataset of FTIR spectra. Furthermore, we performed a deep statistical analysis of all the architectures performance and efficiency to show the suitability of neural networks for spectra matching. The results presented at the end of this article indicated the overall suitability of the selected neural network architectures for spectra matching in microplastics classification.

The ability to grasp objects is one of the basic functions of modern industrial robots. In this article, the focus is placed on a system for processing the image provided by a robot visual perception system leading to the detection of objects grasping points. The proposed processing system is based on a multi-step method using convolutional neural networks (CNN). The first step is to use the first CNN to transform the input image into a schematic image with labeled objects centers of gravity, which then serves as a supporting input to the second CNN. In this second CNN, original input and supporting input images are used to obtain a schematic image containing the grasping points of the objects. This solution is further compared with a network providing grasping points directly from the input image. As a result, the proposed method provided a 0.7% improvement in the average intersection over union for all of the models.

Some analyses state that buildings contribute to overall energy consumption by 20–40%, which, in the context of the recent geopolitical energy crisis, makes them a critical issue to study. Finding solutions for better energy management in buildings can have a significant impact on the energy sector, thus reducing EU energy dependencies and contributing to the fulfillment of the REPowerEU goals. This paper focuses on proposing a simplified model of a residential house considering the main appliances, heating and cooling, a photovoltaic system, and electric vehicle recharging. Weather and solar irradiance forecasts are taken into account. The model predicts the energy demands of a house based on online weather forecasts and the desired indoor temperature. The article also focuses on the analysis of how weather forecast uncertainty affects energy demand prediction. This model can be used to better understand and predict the energy demand of either a single house or a set of houses. A multi-objective optimization approach that takes into account the preferences of users/inhabitants is developed to provide a compromise between the price paid for the electricity and temperature comfort. The authors plan to apply the proposed model to a residential house’s real-time control system. The model will be tuned, its predictions will be tested, and it will be used for energy demand optimization.

Programming industrial robots in a real-life environment is a significant task necessary to be dealt with in modern facilities. The "pick up and place" task is undeniably one of the regular robot programming problems which needs to be solved. At the beginning of the “pick and place” task, the position determination and exact detection of the objects for picking must be performed. In this paper, an advanced approach to the detection and positioning of various objects is introduced. The approach is based on two consecutive steps. Firstly, the captured scene, containing attentive objects, is transformed using a segmentation neural network. The output of the segmentation process is a schematic image in which the types and positions of objects are represented by gradient circles of various colors. Secondly, these particular circle positions are determined by finding the local maxima in the schematic image. The proposed approach is tested on a complex detection and positioning problem by evaluation of total accuracy.

Precise object localization in an industrial environment is a significant task affecting follow-up processes for a pick and place application. One of the solutions to effectively ensure the success of this task is to use modern methods of machine vision. Machine vision is still a highly evolving topic, in which the use of approaches based on convolutional neural networks is rising. And so in this contribution, an innovative engineering approach based on convolutional neural networks is proposed for an object localization task. The approach is based on an atypical image segmentation, where the individual objects are represented by two colored gradient circles. These circles represent significant parts of the object like its center or ending. Each object type (class) is determined by a specific color. By use of a local maxima finder, all circles in an image are transformed to points. With knowledge of these points the coordinates and rotations are calculated. The proposed approach was tested on a legitimate localization problem with 100% precision, more than 99.52% recall on the positioning task and with an average of 6 minutes angle variance per object.

By multivariable decentralized control, changing one set-point in result acts as a disturbance to other control loops. This can be solved by using multivariable controller or compensator. In this paper, a novelty approach to control non-square eighth-order system with four inputs and three outputs is demonstrated using a static compensator that guarantees autonomy in the steady state (changing one input, affects one output) and also optimal solution for non-square overdetermined systems (systems with more manipulated variables than controlled variables). To evaluate the control quality of this method, the system is also controlled without static compensator for comparison.

Grasp point detection is a necessary ability to handle for industrial robots. In recent years, various deep learning-based techniques for robotic grasping have been introduced. To follow this trend, we introduce a convolutional neural network-based approach for model-free one step method for grasp point detection. This method provides all feasible grasp points suitable for parallel grippers, based on a single RGB image of the scene. A case study, which shows the outstanding accuracy of the presented approach as well as its acceptable response time, is presented at the end of this contribution.

Paper describes an implementation of a low cost but nontrivial laboratory system for automation and control theory teaching purposes. The system called RCDue consists of two parts - a dynamic system module and a control unit providing measurement, control and communication with MATLAB through USB serial port. Wiring of the dynamic system module can be modified to get a system of the desired behavior. Application example named as R5C4 with four RC circuits is described together with its mathematical model and calculated step responses. The system can be used in laboratory for practical education of C or MATLAB programming, modelling and identification or control theory subjects.

Magnetic Levitation Systems (MLS) are usually highly non-linear systems with a great sensitivity to the size of the control input. Therefore, special emphasis should be placed on the correct identification of all unknown MLS parameters. This paper describes the principle and procedure for identifying laboratory plant CE 152 MLS with an emphasis on automatically processing identification data. Moving-Average Filter (MAF) and Fast Fourier Transform Filter (FFTF) methods are compared to filtering input data noise. Key parameters are then estimated using the Least Squares Method (LSM). The results are verified in simulation and real-world experiment using a simple PID controller.