Michal Blaho (Humusoft)
Interesting changes in MATLAB and Simulink core modules, other extensions and new products. During this post we will talk about new features in:
And news in tools for:
Jaroslav Jirkovský (Humusoft)
Deep Learning allows 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 bring innovations for different levels of work with this technology, from beginners to advanced users.
New users will appreciate the ready-to-use Deep Network Designer graphical application. The application is used from the design of the deep learning network, to the use and modification of the supplied networks, to data preparation and actual learning. The graphical tool is a wizard where the individual steps are arranged in a logical order and guides the user through the entire deep learning process.
For advanced users, a new extended development framework is provided in which the build and learning process of deep learning networks can be completely customized and personalized. This will allow the creation of advanced deep learning architectures such as generative adversarial networks (GANs) and Siamese networks, the use of custom training loops, user input of loss functions, and the use of automatic differentiation or weight sharing. It is also possible to train networks with multiple inputs and multiple outputs.
Michal Blaho (Humusoft)
The development of robotic and autonomous systems is one of the modern areas of research. Researchers and engineers of such systems strive to design and fine-tune algorithms that meet the most stringent requirements in areas such as motion planning or perception of the environment for mobile robots, UAVs or manipulators. A frequently used solution is Robot Operating System (ROS), which helps in the acquisition and analysis of sensor data. In this talk, we will discuss interesting MATLAB tools for the development of robotic and autonomous systems.
Jaroslav Jirkovský (Humusoft)
Machine learning uses data and creates a program to perform a given task. The core of the resulting program is a mathematical model that evaluates outputs based on input data. The goal of machine learning is to set the parameters of the model so that the evaluation of outputs is done with maximum accuracy and minimum error. The basic tasks of machine learning are classification, regression and cluster analysis. MATLAB provides features for the complete development of machine learning-based applications, from data preparation to model building and learning to the implementation and deployment of the resulting algorithms as server applications or intelligent embedded systems.
Jaroslav Jirkovský (Humusoft)
Simulink is a graphical environment of block diagrams, in which it is possible to quickly and easily model the behaviour of dynamic systems or model various algorithms. Many approaches can be used and combined to describe the modelled systems from different perspectives. Models of systems can be created based on a mathematical description by differential or difference equations, or use the so-called physical modelling approach, where models of systems are created by composing elements representing real objects. In addition to the dynamic description of systems, Simulink can use state-based modelling in the form of state automata and control logic or model systems based on entity generation and processing. Of course, it is also possible to incorporate MATLAB functions and algorithms directly into Simulink graphical models. Simulink can thus be described as a „multilingual“ environment for modeling and simulation-based system design – Model-Based Design.
Michal Blaho (Humusoft)
An overview of current options for creating, deploying and sharing applications created in MATLAB and Simulink, with a focus on two new tools – Simulink Compiler and MATLAB Web App Server. Simulink Compiler allows you to share Simulink simulations as standalone executable applications. Standalone executable applications can use MATLAB graphics and user interfaces designed in MATLAB App Designer. In order to cosimulate Simulink models with external simulation tools, it is possible to create a separate Functional Mockup Unit (FMU) from the model. MATLAB Web App Server is a new tool that allows you to host applications created in the MATLAB Simulink environment in the format of interactive web applications. End users can run and use web applications using a web browser without installing additional software.
Martin Kožíšek (Humusoft)
Introduction of simulation tools COMSOL Multiphysics, COMSOL Server, COMSOL Compiler. Interfacing with MATLAB and modern trends: applications, digital twins and optimization.
Martina Mudrová (Humusoft), Jiří Nárožný (Humusoft)
A brief overview of the current possibilities that MATLAB and its complementary services offer for teaching and education, not only in the field of higher education.
Martin Kožíšek (Humusoft)
Introduction of two new modules in the new COMSOL Multiphysics 5.5.
Matouš Lorenc (Humusoft)
In the form of a mini-course, where the lecturer will introduce and mainly set up the task, we will touch upon the topics „Power spectral density“, „Shock response“ and model lightening as the first step in the process of creating a „Digital Twin“.
Matouš Lorenc (Humusoft)
We compare the experimental data with a simulation set up by the speaker in front of the audience.
Jana Sárená (Humusoft)
The dSPACE platform has its place in the automotive world. But it's not the only area where dSPACE makes engineers' work more efficient. We will show you the possibilities and examples from the field where dSPACE has been deployed.
Martin Kožíšek (Humusoft)
Workshop for all technicians who want to set up a numerical simulation of a physical process under the guidance of a lecturer (and with the support of a printed manual). Suitable also for complete beginners.
Jan Studnička (Humusoft)
Tips, tricks and nifty new features in MATLAB that will make your work easier when creating graphs, in the form of an evening „drink & learn“ session.
Blažej Kubizna (Schaeffler), Samuel Kecík (Schaeffler)
As part of the development of this product, Schaeffler Kysuce is involved in the development of the temperature system. Therefore, it is very important to develop a cooling system that is responsible for controlling the temperatures in the system. A thermal model of the gearbox has been developed which includes both an external and an internal cooling circuit. The external circuit is a water circuit (aqueous liquid cooling) which is developed directly at the customer's site and only data is provided. In the following figure, a diagram of the cooling circuit can be seen, as well as the Schaeffler horizon.
Martin Šiler (Ústav přístrojové techniky AVČR, v.v.i.)
When developing algorithms, it is still possible to run into physical limits of computing, whether it is the power of the processor or the amount of memory available. In such cases, one possible route is to move from high-level languages such as MATLAB, LabVIEW or Python to lower-level languages such as C/C++ or Fortran. This gives us, for example, more control over code execution, the ability to combine local processing of large data sets, control over parallelization, or moving complex computational operations to the graphics card. The price for this is a significantly higher memory footprint and the difficulty of debugging code and visually checking the results of mathematical operations.
In a project that deals with optical imaging deep inside tissues (brain) by promoting optical fiber with minimal cross-section, we encountered this type of hardware limitation. We need to process gigabytes of measured data in the shortest possible time using about 100,000 2D Fourier transforms and then use these results for further measurements. By rewriting the algorithm from LabVIEW to C, we achieved more than 100 times faster computation time.
During the talk, practical experience will be demonstrated how to prepare a functional interface to an external library (DLL) in MATLAB for its testing and for verifying the correctness of the obtained results.
Filip Rak (Zuri.com SE), Ondřej Procházka (Zuri.com SE)
Identification of the parameters of a model of a perpendicular take-off and landing aeroplane based on experimental flight test data. The model is then used to simulate the flight characteristics of a real aircraft.
Robert Grepl (MECHSOFT s.r.o.), Martin Appel (MECHSOFT s.r.o.), Martin Formánek (MECHSOFT s.r.o.)
For our customers we needed a measurement/control HW that would be more powerful than Arduino, cheaper than NI/dSPACE and faster than conventional UART/USB communication. So we put some smart heads together and …
In this talk, we'll present some secret SW/HW tricks to implement a control loop at up to 2kHz over conventional USB 2.0.
We will demonstrate the solution with the following examples: