Alessandro Tarchini (MathWorks)
This year marks a significant milestone in development of Stateflow – MATLAB add-on for modeling and simulation of combinatorial and sequential decision logic using state machines and flow charts. Stateflow now allows creating Stateflow charts for execution as MATLAB objects, without the necessity to use Simulink environment. This enables a novel approach to developing MATLAB applications – instead of classical programming, we can design application control logic using Stateflow Charts. This technique will be demonstrated on creating a simple image recognition application.
Michal Blaho (Humusoft)
Zajímavé změny v základních modulech MATLABu a Simulinku, další rozšíření a nové produkty.
Jaroslav Jirkovský (Humusoft)
Deep Learning enables you to solve computer vision tasks such as image classification, object detection and semantic image segmentation, or tasks in signal recognition and advanced control system design. Applications include automotive applications – ADAS and autonomous driving, medical applications – image diagnostics and MRI, satellite imaging, speech recognition and system monitoring. The latest tools in the MATLAB environment provide graphical design and editing of deep learning models – Deep Network Designer tool, model support for 3-D image data, efficient object detection using YOLO detector, graphical applications for labeling images, video, signals and audio, support for model exchange via ONNX format and automated deployment of the resulting models to target devices via C or CUDA code generation.
Jaroslav Jirkovský (Humusoft)
A complete novelty is the Reinforcement Learning toolkit, a technique that enables the application of deep learning to solve complex tasks in the field of automatic and autonomous control of systems and robotics.
Jan Studnička (Humusoft)
An overview of MATLAB tools for different types of optimization. We will present the possibilities of solving a wide range of optimization problems using Optimization and Global Optimization Toolbox, automatic calculation of gradients for the optimization solver using symbolic calculations, Bayesian optimization of hyperparameters of machine learning algorithms, optimization of parameters in Simulink models, optimization of portfolios and others.
Jaroslav Jirkovský (Humusoft)
Using MATLAB and Simulink to develop software for condition monitoring and predictive maintenance. Predictive models are the basis for estimating the remaining useful life (RUL) of equipment. Different types of models can be used, depending on the available information from the operation of the monitored equipment. The whole process involves several stages, from data collection and selection of suitable indicators of equipment condition, through the design and testing of the predictive model, to the deployment of the resulting solution within enterprise systems.
Michal Blaho (Humusoft)
Inaccurate measurements from sensors make it difficult for several systems to succeed. To be able to estimate system states such as position or rotation more accurately, it is necessary to merge the available sensor information into a single model. Sensor fusion, which is included in the Sensor Fusion and Tracking Toolbox, is used for this task. During this paper, we will demonstrate the creation of virtual scenarios, position and orientation estimation, estimation filters, and object tracking using a multi-object tracker.
Jaroslav Jirkovský (Humusoft)
In the Simulink environment you can create algorithms (= graphically model them), verify them by simulation and then automatically generate a program in C/C++ or HDL from them. However, programming a functional algorithm is not the only step in the development of software and hardware systems. It is complemented by defining the requirements and planning the architecture of the resulting solution, designing and managing tests, or validating and integrating the solution within a broader whole. MATLAB has recently introduced several tools that allow these tasks to be addressed in a coherent way with the actual design of functionality. These include Simulink Requirements, Simulink Test, Simulink Check, and Simulink Coverage. The most recent of these tools is System Composer, which is used to design and analyze the architecture of SW systems and which provides a logical bridge between the definition of requirements and the actual implementation of algorithms.
Martin Kožíšek (Humusoft)
Presentation of COMSOL Multiphysics, COMSOL Server and COMSOL Compiler simulation tools for mathematical simulations of physical processes. The lecture is followed by a COMSOL Multiphysics workshop in the afternoon part of the program.
Martina Mudrová (Humusoft)
A brief overview of current MATLAB licensing options.
Jana Sárená (Humusoft)
Introduction of the dSPACE real-time platform. The use of hardware-in-the-loop testing in the automotive industry. Testing of autonomous driving scenarios.
Jan Studnička (Humusoft)
Tips, tricks and handy tools in MATLAB and Simulink in the form of an evening quiz over a delicious drink.
Matouš Lorenc (Humusoft)
Design and application of antennas for microwave tissue heating. The workshop is designed to introduce the COMSOL Multiphysics environment and physical interfaces for simulating microwave fields and heat propagation in biological tissues. Both complete beginners interested in FEM simulations and advanced users will find it useful.
Igor Podlubný, Jana Pócsová, Andrea Mojžišová, Tomáš Škovránek (Technická univerzita v Košiciach)
We will present our experience with using MATLAB for our research and applications, with an emphasis on its intrinsic ability to support creative thinking. When it comes to publishing our results, we follow the idea of reproducible research and accompany our publications with our MATLAB toolboxes. In recent years we even used MATLAB's Live Editor (in the past we were using export to LaTeX or HTML) for writing not only documentation to our toolboxes, but also research papers, as well as materials for teaching, directly in MATLAB. Using MATLAB for teaching mathematics and other courses allowed us to change the way how students perceive and understand mathematical tools, and therefore to help students in getting better results and succeeding in various student competitions.
Robert Grepl, Martin Appel (MECHSOFT s.r.o.)
Petr Kolář (Geofyzikální Ústav AV ČR), Matěj Petružálek (Geologický Ústav AV ČR)
Convolutional neural network (CNN) is used to identify the acoustic emission phenomena arising from the loading of rock samples. While the actual architecture of the CNN is virtually identical to the manual recommendations, its input and output have been modified considerably. The records – originally time series – enter the identification algorithm in the form of a spectrogram. The final identification is then not given directly by the output of the CNN, but is a combination of the probabilities of all classes considered.
The presented research was published in 2022 in the paper „A two-step algorithm for acoustic emission event discrimination based on recurrent neural networks“ (full paper on request from the author)
Martin Šiler (Ústav přístrojové techniky AV ČR)
High-resolution optical imaging deep inside the tissues (brain) is only possible with endoscopes. Multimode optical fibers can be used as such an endoscope with a minimal cross-section. However, these are not able to transmit the image directly, as they pseudo-randomly reshape it. This process is characterized by a large transmission matrix whose size can reach tens of GB.