Christoph Stockhammer, MathWorks Germany
Artificial Intelligence is transforming the products we build and the way we do business. It also presents new challenges for those who need to build AI into their systems. Creating an “AI-driven” system requires more than developing intelligent algorithms. It also requires:
Christoph Stockhammer, MathWorks Germany
Reinforcement learning allows you to solve control problems using deep learning without using labeled data. Instead, it uses a model of your system that captures the appropriate dynamics of the environment and learns through performing multiple simulations. This simulation data is used to train a policy which is represented by a deep neural network that would then replace your traditional controller or decision-making system.
Zdeněk Hanzálek, CIIRC ČVUT
Automated car control is a challenge for any engineer who has studied technical cybernetics or a related field. In cooperation with Porsche Engineering in Prague, we have taken up this challenge and today we will show you how to autonomously zigzag the Porsche Panamera between the cones. This complex application includes a number of interesting tasks: development of efficient SW, optimization, image processing, control theory, data processing from camera and lidar, machine learning and real-time programming of parallel HW. Everything has to happen reliably and quickly because we want cars that are reliable and fast.
Zdeněk Kubín, Doosan Škoda Power
Predictive blade diagnostics are increasingly in demand due to the increasing demands on blades due to the incorporation of renewable sources. However, the use of tensiometric measurements or blade tipping is very expensive, so less accurate measurements are made using standard turbine instrumentation. The instrumentation inaccuracy is replaced by robust identification methods and subsequent classification using neural networks.
Jaroslav Jirkovský, Humusoft s.r.o.
Deep Learning allows you to solve computer vision tasks such as image classification, object detection on images and semantic image segmentation, or tasks from signal recognition and advanced control system design. Applications can be found in automotive applications – ADAS and autonomous driving, medicine – image and MRI diagnostics, satellite imaging, speech recognition or system monitoring. The latest tools in MATLAB environment bring graphic design and editing of deep learning models – Deep Network Designer tool, model support for 3-D image data, effective object detection with YOLO detector, graphic application for tagging images, video, signals and sound, support for model exchange via ONNX format and automated deployment of resulting models on target devices through C or CUDA code generation. A complete novelty is a set of tools for Reinforcement Learning, a technique that allows deep learning application to solve complex tasks in the field of automatic and autonomous system control and robotics.
Jaroslav Jirkovský, Humusoft s.r.o.
The use of MATLAB and Simulink tools to develop software for device status monitoring and predictive maintenance. The basis is predictive models allowing estimation of the remaining lifetime of the device (RUL). Models of different types can be used depending on the available information from the operation of the monitored device. The whole process involves several stages from data collection and selection of suitable indicators of the condition of the device through the design and testing of the predictive model to the deployment of the resulting solution within the enterprise systems.
Jiří Sehnal, Humusoft s.r.o.
Autonomous driving of a car is perhaps the biggest current challenge for the automotive industry. For the development of algorithms and their testing, it is necessary to have a number of real-life scenarios from operation on which artificial intelligence learns. To do this, it is necessary to obtain data from operation, perform selection, anonymization, annotation and convert them into parameterizable test scenarios. These are then used to simulate driving of a vehicle in operation.
Michal Blaho, Humusoft s.r.o.
The development of robotic and autonómnych sys belongs to modern areas and research. Researchers and engineers takýchto sys sa trying to design and debug algorithms that splnia najprísnejšie požiadavky in areas such as the planning of movement or perception okolia for mobile robots, UAV alebo manipulators. A frequent part of the solution is also the Robot Operating System (ROS), which helps in the acquisition and analysis of data zo snímačov. In this lecture we will talk about the interesting tools of MATLAB for the development of robotic and autonómnych sys