Automation, and electronic and electrical engineering

1. Flexible production lines based on artificial intelligence.

Supervisor: dr hab. inż. Paweł Rotter

Auxiliary supervisor: dr inż. Marcin Węgrzynowski

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: One of the biggest challenge of the modern industry is to create flexible production line which can quickly adapt to a new product, i.e. in electronic industry where it happens that production lines are re-equipped many times during the day. The existing assembly process solutions use impedance controlled industrial robots that perform programmed trajectories (i.e. spiral). Their effectiveness depends on the precision of the grip of the object to be assembled and place in which it is to be placed. Such applications require the use of dedicated mechanism (i.e. a shaped gripper) which limits their ability of the adaptation to new products. Flexible production systems can be achieved though the use of artificial intelligence. The approach we propose uses methods from reinforcement learning and deep learning. We assume that the robot used for experiments will gather information about surrounding environment from sensors such as: industrial cameras, depth camera and force-torque sensor. As part of the research work, the following will be developed:

self-learning algorithm, thanks to which robot will be able to quickly adapt to new tasks,

vision algorithm for object inspection, thank to which it will be possible to use universal gripper.

Proposed approach has the potential to significant improve current production lines. In comparison to existing solution, it allows to quickly adapt to new product, as well as the use of universal mechanisms.

Research facilities: Resource base of Fitech/Fideltronik company:

  • computational resources (servers and workstations),

  • industrial equipment needed for research (industrial robots, actuators, sensors etc.),

  • engineering resources needed to create prototypes,

  • factories where research’s results can be tested,

  • funds for conferences (the company got NCBiR grant to create innovative solutions for the industry).

Number of places: 1

 

 

2.  Nanodielectrics for the development of electrical insulating systems with increased reliability, the use of advanced measurement and diagnostic methods for the evalu­ation of modified composite insulating materials.

Supervisor: dr hab. inż. Paweł Zydro, prof. AGH

Auxiliary supervisor: dr inż. Maciej Kuniewski
Department of Electrical and Power Engineering, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The reliability of insulation systems of electrical power equipment determines the uninterrupted supply of electricity. Research works related to the improvement of dielectric properties of electro-insulating materials used in insulation systems of medium and high voltage devices currently include the development of new nanocomposite materials. The research issue concerns the application and development of advanced measurement and diagnostic methods for the assessment of the properties of polymeric nanocomposite materials and technological processes for their production. The tests will be aimed at determining the impact of structural modifications of nanocomposites on their dielectric parameters. The element of research will be long-term aging tests of materials carried out under conditions of synergic action of electrical and thermal stresses.

Research facilities: The Department of Electrical and Power Engineering has research laboratories equipped with research stands and measuring instruments that enable realization of the planned research program both in the long-term aging procedures of dielectric materials under conditions of electrical and thermal stresses (e.g. high quality computer control thermal vacuum chambers with high- and low-voltage bushings) as well as measurements using advanced measurement techniques (e.g. broadband dielectric spectroscopy system, detection and measurements of partial discharges using PRPDA type systems), polarization/depolarization current measurement systems, system for determining of space charge distributions using the PEA (Pulsed Electroacoustic) method.

Number of places: 2

 

3. New approach to time continuous system faults detection in large scale desalination plants by utilizing exact state observers and signal modulation.

Supervisor: prof. dr hab. inż. Witold Byrski

Auxiliary supervisor: dr inż. Jędrzej Byrski

Department of Applied Informatics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: Development of new methods for fault diagnosis in desalination process. Design of dynamic mathematical model of desalination process in state space. Based on the real experiments the identification of the model parameters. In the case of technology faults the analysis of parity equation for the residual vector, which depends on the output/input derivatives and unmeasurable faults. Two different methodologies may be proposed for the solution of fault diagnosis problem – the transformation of parity equation by the use of modulating functions to algebraic equation or by the use of the exact state observers and reconstruction of the unknown input/output derivatives acting in the matrix parity equation. Proving the superiority of accurate state observers over asymptotic estimators such as the Kalman filter. Efficient algorithms for fault diagnosis in large scale desalination plants. New methodologies being the basis for a doctoral dissertation.

Number of places: 2

 

4. Development of integrated microprocessor systems.

Supervisor: dr hab. inż Robert Szczygieł, prof. AGH

Department of Measurement and Electronics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The proposed research topic includes the development of a programming and test environment for the development of modern microprocessor systems. The currently available open architecture of the Risc-V processor is gaining increasing popularity among both commercial companies and leading research centers. It allows creating very flexible user implementations and introducing dedicated extensions. Potential exemplary applications include both IoT applications, security applications, and many more. The detailed direction of architecture development will be determined after preliminary research.

Research facilities: The Department of Metrology and Electronics, AGH UST, in which the research will be carried out, has many years of experience in designing and testing of integrated circuits. The Department has all the software necessary for design and verification of integrated circuits, as well as equipment for testing, both silicon wafers and individual dices.

Number of places: 1

 

5. Signal processing techniques based on local extrema sampling

Supervisor: dr hab. inż. Marek Miśkowicz, prof. AGH

Auxiliary supervisor: dr inż. Dominik Rzepka

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Summary of research problem: The objective of the proposed research topic is the development of a technical framework for a new signal processing mode based on sampling and processing local extrema of an input signal for the purpose of its reconstruction, or for the capture of specific signal properties. The technical scope of the research will include theoretical analysis of local extrema rate, distributions, and envelope in stochastic signals, as well as perfect and approximate signal recovery from its local extrema, and estimation of signal parameters based on local extrema. Furthermore, the proposed research is going to show that the signal processing techniques based on extrema sampling can find many applications in biomedical engineering.

Research facilities: Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering of AGH offers research development facilities including both the scientific equipment and effective academic supervising. The supervisor has been an author of multiple research publications within the scope of the proposed research topic and related topics. The research is going to be carried out in the cooperation with foreign co-investigators.

The proposed topic is covered by the scope of the research project founded by National Science Centre in the OPUS programme.

Number of places: 2

 

 

6. Fast integrated circuits of pixel architecture operating in a single photon counting mode

Supervisor: prof. dr hab. inż Paweł Gryboś

Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of this research topic is to search for and implement a new concept of a hybrid detector for X-ray radiation providing a measurement of the spatial distribution of X-ray intensity with high resolution as well as spectrometric detection of incident photons. In addition to the combination of positional and energy detection of incident photons, this detector will be able to work at high intensity of X-rays and record photons from a wide range of energy. The hybrid detector operating in the single photon counting mode consists of two components: a pixel-based detection element (ensuring fast conversion of incoming X-ray photons into short nanosecond current pulses) and a multi-channel specialized integrated circuit also with pixel architecture, in which each of the reading pixels will be connected to the pixel of the detection element and will process successively the incoming current pulses (amplified, filtered and digitized). The work will focus on the design and testing of specialized integrated circuits.

Research facilities:

  • lab for design and testing integrated circuits

  • along with the progress of the PhD student's work, there is a possibility of applying for a research project (eg NCN)

Number of places: 1

 

7. Multichannel integrated circuits for precise time measurement

Supervisor: dr hab. inż Piotr Kmon

Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of this research topic is to search for and implement a new concept multichannel integrated circuit for precise time measurements – Time to Digital Converter. These type of converters are applied in different areas of science and industry (i.e. high energy physics, distance measurements for automotive or 3D imaging, or PET biomedical imaging). There are well known problems that limit practical implementation of these methods (i.e. low resolution, lack of uniformity, large area occupation or large power consumption) and these limits their further development. The work will focus on the design and testing of specialized integrated circuits.

Research facilities: The proposed in this project prototypes of ASIC in CMOS nanometer technology will be send to production via EUROPRACTICE service. The EUROPRACTICE IC Service, offered by IMEC and Fraunhofer, offers low-cost ASIC prototyping and ASIC small volume production through Multi Project Wafer (MPW). For the European universities and research institutes, EUROPRACTICE offers special educational conditions through a contract with the European Commission in the frame FP6 and FP7 including both access to professional software and special discounted prototype fabrication prices. We are the EUROPRACTICE member and the planned in the project cost of expenditures are based on up-to-date offers for ASIC production obtained from EUROPRACTICE.

The project author is a member of group having constant access to both the software for ASIC design (Cadence and Mentor packages) and laboratory for integrated circuit testing including:

  • probe station Alessi with sets of micromanipulators and active probe for testing naked integrated
  • circuits,
  • wire bonding facility (Kulicke & Soffa Digital Wire Bonder 4500),
  • X-ray generator with step motors,
  • femtosecond laser with pulse picker,
  • Network Spectrum Analyzer HP4195, Semiconductor Parameter Analyzer, RLC meter Agilent 20 Hz - 2 MHz,
  • Keithley for measurements of extremely low currents,
  • National Instruments based different types of complete test stations based on PXI, etc.

 

Number of places: 1

 

8. New solutions low area, low-noise and fast ADC

Supervisor: dr hab. inż. Piotr Maj, prof. AGH

Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of this project is to develop and implement a new concept of a readout integrated circuit (IC) for hybrid radiation detectors of pixelated architecture, which processes electrical pulses generated by photons / particles and simultaneously provides statistical analysis of the radiation energy spectrum, by precise amplitude measurement of each pulse using an in-pixel analog-to-digital converter (ADC) and digital in-pixel signal processing.

Number of places: 1

 

9. New solutions of X-ray detectors for synchrotron applications and electron microscopy

Supervisor: dr hab. inż. Piotr Maj, prof. AGH

Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of this research topic is the implementation of a new concept of data readout from pixel X-ray detectors providing desired parameters for synchrotron experiments and electron microscopy. The most important of which is reading a very large number of frames per second, detection of simultaneous registration of two photons, very low sensitivity on crosstalk from the digital part. An additional goal of the research problem is the construction of an autonomous system capable of independent work on a synchrotron or electron microscope.

Number of places: 1

 

10. Gallium nitride (GaN) based multilevel DC-DC converfters.

Supervisor: dr hab. inż. Robert Stala, prof. AGH

Department of Power Electronics and Energy Control Systems , Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: Application of gallium nitride transistors (GaN) makes it possible to achieve high power density power electronic converter. GaN-based converter can operate with high switching frequency with low switching losses which makes it possible to optimize the converter towards minimization of volume of passive components.

The drain-source voltage of GaN transistors is limited to 650V. To achieve converter for higher voltage a multilevel topology can be applied. In a multilevel converter voltage stresses on semiconductor switches are decreased in relation to converted voltages. It is achieved in more complex topologies with higher number of switches in relation to classic converters.

Research works of GaN -based converters will lead to optimization the converter towards high power density, high efficiency, low volume of passive components and high voltage gain.

Research facilities: The research facility contains suitable laboratory equipment such as oscilloscopes, probes, design and simulation software, power supplies and loads as well as competence for carrying out analytical, simulation and experimental research in the area of power electronics.

Number of places: 1

 

11. Hybrid magnetic gear box for industrial applications.

Supervisor: dr hab. inż. Adam Piłat, prof. AGH

Department of Automatic Control and Robotics , Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of the scientific-research work is to develop a mathematical model and carry out simulation research on the proposed hybrid magnetic transmission, in particular intended for applications in the robotics field. Studies will require analysis of mechanical, electrical engineering, electronic and control dependencies in order to synergize knowledge and technologies used in the prototype structure. Conducting simulation tests is aimed at optimizing the configuration and selection of control strategies depending on the variable static conditions and dynamic work of the transmission. Experimental research will be carried out using the manufactured prototype. These tests will use the infrastructure of the Laboratory of Robotics and Magnetic Levitation of the Department of Automatics and Robotics. The results of experimental research will be used to verify the simulation studies of the discussion on the optimization of the configuration.

Research facilities: Research facilities include software for implementation of multi-domain modeling tasks, desktop computers for calculations and implementation of real-time control tasks using the automatic code generation methodology, programmable systems: industrial controller (PAC), microcontrollers, FPGAs, signal processors, devices with active magnetic levitation technology: single and two actuators suspension, magnetic bearings of various configurations, rigid rotor in magnetic bearings, flexible rotor in magnetic bearings, industrial robots, force sensors, magnetic field sensors, signal generators, oscilloscopes, programmable power supplies, distance sensors.

Number of places: 1

 

12. Transport system with magnetic levitation technology

Supervisor: dr hab. inż. Adam Piłat, prof. AGH

Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of the scientific and research work is to extend the current configuration of the transport system with the technology of magnetic levitation in order to obtain a fully autonomous solution. Scientific and research work will be devoted to the aspects of modeling and nonlinear control taking into account the specificity of the configuration. The implementation of goals will require the development of a complex non-linear mathematical model, conducting analyzes, development of several control strategies with a focus on nonlinear control, including the use of state observation. Conducting simulation tests is to lead to the selection of appropriate control methods for the given functional scenarios of the device. Conducting experimental research in the Laboratory of Magnetic Levitation of the Department of Automatic Control and Robotics will be the basis for verification of the study work. The developed controllers will be verified during real time regime tests.

Research facilities: Research facilities include software for implementation of multi-domain modeling tasks, desktop computers for calculations and implementation of real-time control tasks using the automatic code generation methodology, programmable systems: industrial controller (PAC), microcontrollers, FPGAs, signal processors, devices with active magnetic levitation technology: single and two actuators suspension, magnetic bearings of various configurations, rigid rotor in magnetic bearings, flexible rotor in magnetic bearings, passive-active transport system, kinetic energy storage, force sensors, magnetic field sensors, signal generators, oscilloscopes, programmable power supplies, distance sensors, video detector (300 fps).

Number of places: 1

 

13. Optimal control and management of energy flow in the energy storage and accumulation system

Supervisor: dr hab. inż. Adam Piłat, prof. AGH

Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of scientific and research works is to develop optimal configuration and control strategy in a hybrid energy generation and storage system. This system in its basic configuration consists of a wind farm, photovoltaic panels, a kinetic energy store, and accumulators. It is required to develop a comprehensive mathematical model that takes into account the characteristics of individual devices, energy flow, and working conditions. Such a model will be subjected to study studies in order to develop an optimal control strategy in particular in changing climatic conditions. Next, experimental studies will be carried out using the infrastructure of the Laboratory of Photovoltaics and Magnetic Levitation of the Department of Automatic Control and Robotics. The results of experimental studies will be a discussion for the study work carried out.

Research facilities: Research facilities include software for implementation of multi-domain modeling tasks, desktop computers for calculations and implementation of real-time control tasks using the automatic code generation methodology, programmable systems: industrial controller (PAC), microcontrollers, FPGAs, signal processors, devices with active magnetic levitation technology: single and two actuators suspension, magnetic bearings of various configurations, rigid rotor in magnetic bearings, flexible rotor in magnetic bearings, passive-active transport system, kinetic energy storage, photovoltaic panels, weather measuring station, battery pack.

Number of places: 1

 

14. Self bearing drive

Supervisor: dr hab. inż. Adam Piłat, prof. AGH

Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The aim of the scientific and research work is to develop a mathematical model of the levitating rotor in the magnetic field together with the drive function. The mathematical modeling process will be supported by multiphysical numerical modeling, taking into account material, geometric, magnetic and control dependencies. Finally, the developed configuration will be subjected to simulation tests in the developed simulation environment. Based on the results of the optimization, a prototype structure will be designed and manufactued, which will be subjected to experimental research in the Laboratory of Magnetic Levitation of the Department of Automatic Control and Robotics. Experimental research will concern the development of an automatic control system to implement the stabilization and drive function. The simulation and experimental tests conducted on an ongoing basis will be used to develop the optimal control method for the considered configuration.

Research facilities: Research facilities include software for implementation of multi-domain modeling tasks, desktop computers for calculations and implementation of real-time control tasks using the automatic code generation methodology, programmable systems: industrial controller (PAC), microcontrollers, FPGAs, signal processors, devices with active magnetic levitation technology: single and two actuators suspension, magnetic bearings of various configurations, rigid rotor in magnetic bearings, flexible rotor in magnetic bearings, passive-active transport system, kinetic energy storage, force sensors, magnetic field sensors, signal generators, oscilloscopes, programmable power supplies, distance sensors.

Number of places: 1

 

15. Implementation of real-time control tasks using complex dynamic models solved in real time

Supervisor: dr hab. inż. Adam Piłat, prof. AGH

Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: The purpose of the research work is to develop a mathematical model of a complex dynamic system that will be embedded in the programmable digital system (s), which will be real-time solution model of the real object. It is assumed that the mathematical model will apply to dynamic systems with time constants of the order of milliseconds or shorter. The implementation of this objective requires describing the dynamic system with appropriate equations, adjusting, testing and optimizing numerical methods due to the quality of calculations, implementation in programmed devices, developing a suitable control strategy using the considered model, and then implementing a real-time control task with a selected device available in the Magnetic Levitation Laboratory of the Department of Automation and Robotics, conducting experimental research and developing results.

Research facilities: Research facilities include software for implementation of multi-domain modeling tasks, desktop computers for calculations and implementation of real-time control tasks using the automatic code generation methodology, programmable systems: industrial controller (PAC), microcontrollers, FPGAs, signal processors, devices with active magnetic levitation technology: single and two actuators suspension, magnetic bearings of various configurations, rigid rotor in magnetic bearings, flexible rotor in magnetic bearings, passive-active transport system, kinetic energy storage.

Number of places: 1

 

16. Dynamics of measurements in Smart Grid and assessment of their suitability for control.

Supervisor: dr hab. inż. Andrzej Bień, prof. AGH

Auxiliary supervisor: dr inż. Szymon Barczentewicz

Department of Power Electronics and Energy Control Systems Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: Modern "Smart Grid" electricity for proper operation must be managed and controlled based on the current time state. The possibility of proper control of the Smart Grid is connected with a measurement system that goes beyond the current "Smatr Metering" system. The measurement data collected and processed are related to real time and must be synchronized like the PMU system - phasors. Assessment of the possibility and effectiveness of synchronization for different groups of measured quantities is a proposed research task. Phasors, power and energy measurements as well as electricity quality indicators will be assessed.

Research facilities: The research facilities of the Department have laboratories for experimental work on physical models, construction of installations up to 10kW and equipping with real time measurement systems. Simulation tools are also available. The current scientific cooperation with Polish and foreign partners in an excellent way increases the facilities laboratory.

Number of places: 1

 

17. The particle detector triggering system for the ALICE experiment (CERN) based on FPGA device.

Supervisor: prof. dr hab. inż. Marek Gorgoń

Auxiliary supervisor: dr inż. Mirosław Jabłoński

Computer Vision Laboratory, Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: As part of scientific cooperation between the ALICE experiment (CERN) and AGH, scientific work is being conducted in which AGH specialists build systems and devices for the new particle detector in the ALICE experiment. One of the directions of works is to develop a particle detector triggering system. The PhD student's task will be to involve in the design work consisting in developing the logic of the reconfigured FPGA system and after obtaining a possible CERN doctoral scholarship, also participating in the commissioning and diagnostic work in the first period of the new detector's work.

Research facilities: The AGH Vision Systems Laboratory has access to the necessary software licenses required to design reprogrammable systems. It is also planned to provide FPGA boards, identical to those used in the ALICE Experiment for the Laboratory in the course of the cooperation.

Number of places: 1

 

18. Control systems for autonomous, unmanned aerial platforms using data fusion, based on SoC heterogeneous systems.

Supervisor: prof. dr hab. inż. Marek Gorgoń

Auxiliary supervisor: dr inż. Tomasz Kryjak

Computer Vision Laboratory, Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science, and Biomedical Engineering

Summary of research problem: Autonomous, unmanned aerial vehicle (drones) are currently a very dynamically developing field of scientific research and practical applications: from typically military (intelligence, attacking targets) to commercial (good delivery, area surveillance, search operations - both in the field and inside buildings ). One of the essential components of a drone is the computing platform. It should be characterized by high computational efficiency (real-time processing of data from many sensors: camera or cameras, LiDAR, radar and control algorithm, communication with the ground) and energy efficiency (due to limited battery capacity). In addition, it should enable the introduction of improvements to the used algorithms. The above requirements are met by modern heterogeneous computing platforms in which the processor system is connected to reconfigurable logic (FPGA device), as well as - in the latest generation - with dedicated modules for neural networks implementation (ACAP devices). The topics of the proposed research will focus on demonstrating the advisability of using these platforms for drones. The specific research issues include: methods of simultaneous localization and mapping (SLAM - based on video and IMU sensor data fusion), navigation in a difficult environment (obstacle detection and trajectory planning), cooperation of several drones in the realization of selected missions (e.g. mapping of a given area, transport of large items) and broad understood surveillance (of areas, industrial installations or road traffic).

Research facilities: The Computer Vision Laboratory has a research base that allows to start research in the proposed area. Ph.D. candidates can use: 3 drones (1 large - hexacopter, 2 small - quadcopters), controllers, telemetry modules and cameras (vision and thermal imaging). In addition, they will gain access to a number of heterogeneous computing platforms with Zynq SoC and Zynq UltraScale + MPSoC devices (in future also with next-generation ACAP devices), embedded GPUs (as an alternative computing platform), as well as efficient general purpose computers for algorithm prototyping. Work spaces will be also prepared for them. In addition, the Department has 3D printers. An important support for the conducted research will be the Student Scientific Organization AVADER cooperating with the Laboratory, as well as engineering and master thesis related to the topic of drones. Currently, the Laboratory is involved in one scientific project related to the SLAM issue and another related to the processing in heterogeneous devices a video stream with 4K resolution. After their completion, it is expected to apply for projects more directly related to drones: SONATA BIS or OPUS (for the team) either PRELUDIUM for particular Ph.D candidates. Due to the popularity of the topic and numerous practical applications, it is also possible to cooperate with the industry.

Number of places: 3

 

19. Prediction model of an autonomous vehicle’s behavior based on Artificial Intelligence methods.

Supervisor: dr hab. inż. Krzysztof Oprzędkiewicz, prof. AGH

Auxiliary supervisor: dr inż. Piotr Bania

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Nowadays, automotive industry is absorbed in a race that, by refining more and more advanced driver assistance systems (ADAS), leads to achieve full autonomous cars. Automotive industry is one of the fastest growing branches in the world of technology. The technological stack of such a car includes various sensors, perception algorithms and many functionalities that ensure driving safety. One of the most crucial functionality of autonomous car is selection of appropriate behavior and the planning and maintenance of the driving trajectory. As it was mentioned above, one of the most important functionality which autonomous car has to ensure is to judge the traffic situation and undertaking appropriate actions, so as to guarantee safety and reach the goal in an optimal and comfortable way for the user. To ensure this functionality, deterministic or fuzzy logic is usually used. The number and variety of road scenarios, driving dynamics and difficult to predict the behavior of other road users make this task extremely complicated. It is also difficult to guarantee that the standard deterministic algorithm can handle with all of the traffic scenarios on the road. In connection with the above, the aim of this scientific work is to propose new solutions that, using artificial intelligence, will ensure greater efficiency and versatility of autonomous vehicles in various road situations. The research plan provides using machine learning methods, in particular reinforcement learning, and extend it with methods connected to behavioral cloning. This approach will provide modeling of the agent's behavior in a comprehensive manner, enabling him to generalize all presented road scenarios to plan his behavior. These methods, through their ability to model and detect patterns, and thanks to their optimization properties, can successfully compete with the deterministic approach used so far.

Research facilities: Aptiv Services Poland S.A. is one of the largest and most developed companies in Europe that conduct research and development of advanced driver assistance systems (ADAS). For many years, the company has been conducting research and producing innovative products that change the automotive into a better and safer world. Now, there are projects which are being carried out that aim to achieve further levels of autonomy of cars. Thanks to large-scale activities, Aptiv as a research center provides a professional and experienced team of scientists and engineers who, by supporting each other, push forward the development of technology. There are also many test vehicles available in the company, which, equipped with advanced sensors and perception algorithms, serve the entire staff to test and verify the assumptions of their research and to collect the necessary real data. In addition, Aptiv provides access to many modern simulators that allow you to generate a realistic environment model at various levels of detail in the simulated reality. It is also helpful to have access to many high-performance computing clusters that are necessary for the development of systems based on machine learning.

Number of places: 1

 

20. Real-time trajectory generation for autonomous vehicles in dynamic environments with safety constraints.

Supervisor: prof. dr hab. inż. Wojciech Mitkowski

Auxiliary supervisor: dr inż. Krzysztof Kogut

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Generation of a feasible and efficient trajectory for autonomous vehicle is a central problem of autonomous driving technology. Trajectory is planned based on static and dynamic environment models and must fulfill high-level goals, such as achieving defined position in space, while providing comfortable experience for passengers. Due to high diversity of possible road scenarios, use of machine learning (ML) approaches is considered for high-level decision making process in autonomous driving systems, allowing to generalize knowledge obtained from simulations and real world test drives to a variety of situations. Architecture of such system may consist of a behavior planning module, which makes decisions regarding preferred maneuvers and trajectory generation module, which allows to execute chosen behaviors in a safe and efficient manner. Design of a system that assures safety in a testable and verifiable manner while utilizing ML components is an important research problem. Possible solution is to define a mathematical model that allows to compute a deterministic safety envelope, providing time-dependent constraints for planned trajectory ensuring that autonomous vehicle will not cause any collisions of its fault. While drafts of such mathematical models were recently proposed, definition of trajectory planning algorithm that would fulfill such dynamic constraints while generating efficient trajectories remains an open challenge. Main goal of presented research problem is development of trajectory generation algorithms that would be able to plan trajectory for autonomous vehicle in real-time applications and would enable efficient utilization of ML-based behavior planning modules. A high level architecture should be proposed to assure safety of such system, consisting of deterministic mathematical model for safety constraints computation. Deterministic design of safety-critical components of trajectory generation system can be utilized to define requirements for perception system that provides static and dynamic environment model for the trajectory planning system and to ensure transparency and testability crucial in automotive applications.

Research facilities: Aptiv Technical Center in Cracow is one of the largest technical centers in Europe focused on research related to Advanced Driver Assistance Systems (ADAS), automotive sensors, perception algorithms and Autonomous Driving systems. Work related to the described research problem will be performed in close cooperation with other research centers across the Europe and United States, utilizing Aptiv’s research assets such as:

high performance computing clusters capable of processing large databases of real-world perception data and building large-scale machine learning models,

simulators capable of real-time generation of artificial data such as dynamic and static environment models,

test vehicles equipped with full set of automotive sensors such as automotive radars, lidar sensors and vision cameras,

access to large knowledge base build on many years of experience in automotive perception systems and ADAS algorithms.

Described research problem fits into Aptiv’s long-term strategy, which focuses on increasing of capabilities in areas of driving automation systems and providing reliable driver assistance solutions.

Number of places: 1

 

21. Depth completion for radar-like sensors.

Supervisor: dr hab. inż. Paweł Skruch

Auxiliary supervisor: dr inż. Marcin Piątek

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Automotive depth sensors provide measurements which are relatively sparse when compared with the output of common cameras. While the standard cameras used in automotive output images with more than 10^6 pixels, modern automotive lidars produce less than 10^5 measurements per frame. Because of such relative sparsity of the depth measurements provided by lidars, the field of depth completion was started. It is a field which deals with transforming the sparse depth maps produced by a depth sensor into much denser ones. As of today, the best performing (using KITTI Depth Completion Challenge as a benchmark) solutions for depth completion use artificial intelligence and deep learning. Specifically, they are mostly based on the sparsity invariant convolutional layers. The novelty of this project is application of deep learning-based depth completion for radar-like depth sensors. Automotive radars are depth sensors which are much (orders of magnitude) more affordable than lidars. The output of automotive radars is very different than the output of lidars: radar depth measurements are even sparser than that of lidars (roughly 100 detections per frame), moreover the detections are much more irregularly spaced. Additionally, the measurements provided by present day automotive radars have much lower vertical accuracy than lidar measurements. The research project will entail designing neural network architectures that are based on the state of the art solutions for lidar depth completion, however are significantly altered to accommodate the radar depth measurements as input. The project deals with both unguided and guided (using camera) depth completion for radar-like sensors.

Research facilities: Aptiv is a world leader when it comes to automotive radars. Aptiv’s research and development department in Cracow has vehicles fitted with automotive radars, lidars and cameras which will enable the collection of training and validation datasets for neural networks working on depth completion. Additionally the data collection vehicles will be logging low level radar data that enables simulating particular elements in the radar data processing. Aptiv is providing its employees with suitable computational resources for neural network training in the form of local computers with appropriate GPUs, local computational clusters and cloud computing resources.

Number of places: 1

 

22. Analysis and optimization of the electromagnetic field distribution in wiring harnesses of electric vehicle.

Supervisor: dr hab. inż. Paweł Zydroń, prof. AGH

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The research issue is related to the currently conducted research and development of electric vehicle structures, in particular autonomous cars. The use of electric motors for the drive of a car requires the transmission of more electricity than in the case of cars with internal combustion engines. This increase requires a comprehensive change of the car's electric harness system and individual constructions of wire harnesses containing both high-current and signal cables. Planned structural changes must include an increase in the supply voltage values ​​and the power system used (DC or AC). Optimization of wire harness construction is particularly important in the case of autonomous cars, in which the number of sensors, signal cables, as well as the transmission speed and the volume of data transferred significantly increases. Constructions of wiring harness must therefore take into account all of the factors mentioned, which affect the change of the electromagnetic environment, to ensure an adequate level of internal and external electromagnetic compatibility of the wiring harness in the context of ensuring the proper reliability of the entire car. In addition, an important factor affecting the construction of the wiring harness, which must be taken into account, is the generation of heat in the electrical conductors and its effective transfer outside the harness. The proposed topic concerns the analysis of the electromagnetic field in the complex wiring harness of the electric vehicle (in a wide frequency range) and attempts to optimize its structure to ensure correct operation of the power and signal circuits, including the heating of the harness wires and EMC problems.

Research facilities: The scientific and research institution in which the research program will be run as well as the industrial partner will have laboratories as well as technical and measurement infrastructure that ensure the possibility of implementing the research program of the submitted issue. In particular, this applies to specialized shielded chambers and measuring instruments used to perform EMC tests and to determine the value of electromagnetic disturbances that are conducted or transmitted. The research facilities also enable conducting simulation research in advanced environments of numerical analysis, on computers with high computing power and using specialized scientific software.

Number of places: 1

 

23. Application of Dempster-Schafer theory in automotive perception algorithms.

Supervisor: dr hab. inż. Paweł Rotter

Auxiliary supervisor: dr inż. Piotr Bania

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Processing and interpretation of environmental data, acquired from variety of sensors, is one of the main problems within the development process of perception algorithms for the purpose of autonomous driving and driver assist functionalities. The convergence of the model being built to the actual state of vehicle surroundings is extremely important for the safety of road users, including vehicle occupants. Sensors currently used in the automotive industry operate based on various physical phenomena. For some cases, this involves obtaining seemingly exclusive data from the sensors. An example is a lidar beam directed on a transparent surface, that may indicate an incorrect location of the obstacle on the road. Similarly in the case of a radar, not registering a detection from an area does not necessarily mean that the area is traversable for the vehicle. Consequently, this leads to conflicts in sensor readings and difficulties in interpretation. The key is to use an artificial intelligence algorithm that will provide a systematic approach to the interpretation of various information sources and will allow for a consistent determination of resulting uncertainties of uncertainties. Dempster- Schafer theory will be employed to address such issues. The theory will be adapted to several use cases of particular importance in automotive perception systems. A further objective of the proposed research is to create and apply the practical implementation of the Dempster-Schafer theory in low-level fusion of lidar, radar and vision data. Expected results of the work is to increase the efficiency of perception algorithms and reduce computational complexity. Such solution will lead to an algorithm applicable in embedded systems, thus accelerating the commercialization of the developed solutions.

Research facilities: The Aptiv Technical Center in Krakow is a leading research and development unit in the field of electronics and security. Great emphasis is placed on active safety products, perception algorithms and the development of autonomous driving systems. Experience in creating automotive market products allows building complete and safe systems for cars of the future. The company has the required technical facilities to carry out the research work presented. It provides the possibility of using commercial lidar, radar and vision sensors. Test cars are also available, which allow validation of perception algorithms based on data collected in road traffic. Equally important is the possibility of cooperation with research centers of foreign branches of the company, located in the United States and Europe. An insight into the broad knowledge base regarding the development of perceptual algorithms will allow us to focus on solving the main problem of the proposed research. The subject of work meets a long-term company strategy related to the development of autonomous driving systems.

Number of places: 1

 

24. Analysis and detection of anomalies in time series using deep neural networks.

Supervisor: dr hab. inż. Paweł Rotter

Auxiliary supervisor: dr inż. Piotr Bania

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: Advances in computer science and technology led to introduction of numerous embedded systems in cars. Monitoring their performance is a key element in their development and usage. However, the number of collected data is growing with every new component, and discovering faults in their behavior is a challenging problem. The collected data, for example CPU, memory and network usage can be regarded as discrete time series, and traditional techniques known from time series analysis can be applied. However, such models usually require some prior knowledge about collected data, and do not generalize from one type of data to another easily. In the last years, neural networks proved to be a good solution to that problems. Their ability to generalize, and model data let them became one of the main tools in the time series analysis and forecasting. In applications, one of the most used architectures of neural networks are convolution neural networks for image application, recurrent neural networks for sequential data, and generative adversarial network (GAN) recently developed to increase the ability for network to generate data similar to the input. We propose to take advantage of the mentioned network types in the time series analysis. The pattern classification and detection capabilities of convolutional networks can be applied to time-ordered data, with different metrics being treated similarly as channels in images. We plan to adopt attention mechanism to further increase the ability of analysis of sequential data. Finally the GAN architecture can be used to train encoder-decoder type of network, to compute the difference between original data and the reconstruction from encoder-decoder network. This difference called reconstruction error can be used to detect anomalous behavior in the collected data. The potential benefit of that model is the ability to learn normal behavior without supervision, and then online classify new data as anomalous or not. The unsupervised way of learning is a key component, because the amount of collected data makes the labelling process too time and cost expensive.

Research facilities: Aptiv is a global automotive supplier company, with focus on computing and electrical systems. It has a Technical Center in Kraków (TCK), with research groups working in areas of artificial intelligence, perception and sensing systems and automated driving. Moreover the Technical Center closely collaborates with research groups in Germany, United Stated and Sweden. TCK is hosting industrial PhD candidates from recent years, and has a strong ties with the AGH University realized in research collaboration and usage of Cyfronet computing center. The company experience from producing automotive systems allows to perform research project in the described area.

Number of places: 1

 

25. Calibration and auto-calibration algorithms of radar sensors for active safety systems and autonomous driving.

Supervisor: dr hab. inż. Krzysztof Duda

Second supervisor: dr hab. inż. Dariusz Borkowski

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The technical problem faced by the automotive industry is unknown angular errors in the installation of radar sensors on vehicles. These errors usually fall within ± 5 degrees in three axes. In practice, these small errors have a significant impact on the results of the sensor functions. Therefore, it is important to correct these errors. The research problem that will be undertaken in Ph.D. is the estimation of angular errors in the installation of the radar sensor installed on the vehicle. As a part of this thesis two algorithms will be developed:

static one, used for factory calibration of a stationary vehicle sensor,

dynamic one, used for auto-calibration for a vehicle in motion.

The starting point of the work is the technology used by the company (patent EP 3 279 683 A1, Schiffmann J., Liu Y., Schwartz D., Zhu X.) and available publications (Multipath of flat plate radar cross section measurements, 2003 Proceedings of the International Conference on Radar).

Research facilities: Aptiv is a global automotive supplier company, with focus on computing and electrical systems. It has a Technical Center in Kraków (TCK), with research groups working in areas of artificial intelligence, perception and sensing systems and automated driving. Moreover the Technical Center closely collaborates with research groups in Germany, United Stated and Sweden. TCK is hosting industrial PhD candidates from recent years, and has a strong ties with the AGH University realized in research collaboration and usage of Cyfronet computing center. The company experience from producing automotive systems allows to perform research project in the described area.

Number of places: 1

 

26. Wearable low power systems for registration and monitoring of biomedical signals.

Supervisor: dr hab. inż. Marek Miśkowicz, prof. AGH

Auxiliary supervisor: dr inż. Piotr Otfinowski

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The objective of the proposed research topic is the development of a technique of low power low power wearable systems for registration and monitoring of biomedical signals including electrocardiogram (ECG) and photoplethysmography (PPG) by the use of sampling of local extrema. The project is going to contribute to the trend of extending functionality of consumer electronics devices with the systems for monitoring of biomedical signals. The example is the 4th generation of Apple Watch, a line of smartwatches designed by Apple, and introduced in 2018, capable of generating an ECG of the user. The key innovation of the project is taking advantage of the observation that the informative value of many biomedical signals is included in the amplitude and timing of their local extrema. Therefore, in the project, it is proposed to use the local extrema sampling instead of the periodic sampling in order to reduce the amount of data captured and processed in the biomedical data acquisition systems, and thereby to decrease power consumption of the system.

Research facilities: Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering of AGH offers research development facilities including both the scientific equipment and effective academic supervising. The supervisor has been an author of multiple research publications within the scope of the proposed research topic and related topics. The research is going to be carried out in the cooperation with foreign co-investigators. The proposed topic is partially covered by the scope of the research project founded by National Science Centre.

Number of places: 1

 

27. Improving the reliability of power supply in smart grids.

Supervisor: prof. dr hab. Zbigniew Galias

Auxiliary supervisor: dr inż. Szczepan Moskwa

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The research problem includes the analysis of the possibilities of improving the reliability of end users power supply in modern distribution networks (smart grids) by using the functionalities provided by such networks. During the research, the character of recipients and their changing structure as well as dispersed energy sources should be taken into account. The effect of the research should be the development of guidelines for the placement and use of network automation components and algorithms of automation operations in the event of a failure.

Research facilities: The proposed research project will be carried out within the reserach team in the Department of Electrical and Power Engineering led by prof. Zbigniew Galias. Since 2015 the team carries out research on optimization of electrical distribution networks and has knowledge on topics regarding the proposed research project. In 2016-2019, the team carried out a research project „Optimization of power electrical distribution network structures for improved power supply reliability” sponsored by the National Science Centre (NCN). The team plans to apply for another NCN grant covering the subject of optimization of power discribution grids. In case the application is successful the PhD student will have an opportunity to join the team carrying out the NCN research grant.

Number of places: 1

 

28. Magnetization dynamics in magnetic multilayers.

Supervisor: dr inż. Witold Skowroński

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Spin electronics is a dynamically developing branch of electronics in which, apart from the charge of the electron, its spin is also used for processing and storing information. For further development of spintronics, it is necessary to thoroughly study and understand the dynamics of magnetization in particular based on ferromagnetic resonance and spin waves in magnetic multilayer systems. The spin electronics group at the Departmet of Electronics deals with theoretical and experimental work aimed at design and test mutlilayer structures for potential applications as microwave electronic components (oscillators, detectors, phase shifter).

Research facilities: technological line for the production of thin-film systems (clean room; optical, laser and electron lithography; ion etching, layer deposition by magnetron sputtering), microwave measuring apparatus (vector network analyzer, spectrum analyzer, function generator, pulse generator, sampling oscilloscope) probe stations with integrated electromagnets, cryostat with closed helium circuit, measurement apparatus (source-measurement units, power supplies, mutlimeters, generators, oscilloscopes).

Number of places: 1

 

29.Machine learning algorithms in Industry 4.0

Supervisor: dr hab. inż. Ireneusz Dominik

Auxiliary supervisor: dr inż. Krzysztof Lalik

Faculty of Mechanical Engineering and Robotics

Abstract: The issue of research consists in machine learning algorithms. It occurs in the practical application of artificial intelligence achievements to the creation of an automated system capable of improving itself with the help of the accumulated experience (data) and the acquisition, on this basis, of new knowledge, i.e. machine learning. New machine learning algorithms will be developed and implemented for Industry 4.0 technology. One of the most common tasks of machine learning is classification tasks. The work will collect vast amounts of sensor data from machines and the entire production line, as input data for machine learning. Image recognition and standard deviation detection algorithms will be used in quality control with AR augmented reality googles. The algorithms developed will be used in predictive planning, supply chain forecasting and developing new production strategies.

Machine learning is a technology that requires expertise in the preparation of data for training and testing. The implementation of the work will develop algorithms for intelligent control of the production process. The prepared algorithms will be used to test the operation of the production system, as well as to monitor it and adjust to the changing conditions of its operation.

Research facilities: The research facilities of the unit are primarily the research and didactic laboratory of Industry 4.0. The laboratory equipped with this technological line is unique on a national and international scale. In technological terms, this is the country's first laboratory, which allows integration of the latest trends in industrial automation. It realizes fully the foundation of the Industry 4.0 standard of the fourth industrial revolution using the Industrial Internet of Things (IIoT).

The technology line in which the laboratory is equipped fully replicates the technology lines currently used in smart factories. Multi-layered capabilities, both in the horizontal and vertical control hierarchy, are implemented in research and didactical work on many hardware platforms. The systems in which the laboratory is equipped include:

  • Programmable Logic Controllers PLC,

  • Industrial user interfaces,

  • Reference production control Systems,

  • Industrial machine Communication (M2M) and Internet of Things (IoT)

  • Safety Systems (Fail-Safe),

  • Robotic systems,

  • Vision systems and augmented reality systems (AR),

  • Production systems,

  • Supply chain organization, Lean Management and ERP systems,

  • Database Systems,

  • Surveillance and data acquisition systems,

  • Big Data Analysis,

  • Open Platform Communications standard(OPC).

Number of places: 1

 

30. Problems of control systems in road vehicles

Supervisor: dr hab. inż. Jarosław Konieczny

Auxiliary supervisor: dr hab. inż. Waldemar Rączka

Faculty of Mechanical Engineering and Robotics

Abstract: The research works is oriented around the problems of mechanical objects control systems synthesis. The research will be carried out using mobile SUV research platforms. The aim of the research is to design control systems related to improving the comfort and safety of the car passengers. In modern automotive vehicles there are many subsystems that require the synthesis of the proper control law. The synthesis methods based on the theory of optimization and robust control as well as learning using neural networks will be considered.

Research facilities: The research facility of the unit is a Dynamic and Control of Structures Laboratory. The main purpose of the laboratory is to provide scientific and research activities in the field of controlled dynamic structures.

Research is carried on active and semi-active structures using electrohydraulic, electro-pneumatic and electrodynamic elements and assemblies.

The laboratory is equipped to specialized research rigs with PAC control systems and field-programmable gate arrays FPGA. They are used to test dynamic systems and structures. The operation of the test rigs is based on electrohydraulic vibration generators made based on original designs. In addition, the laboratory is equipped with a number of measurement systems with sensors and transducers for signals acquisition. The laboratory is equipped with SUV wheeled vehicles, which are autonomous platforms designed for experimental research. http://www.suvsusp.agh.edu.pl/

Number of places: 1

 

31. Problems of robots autonomy by using neural networks

Supervisor: dr hab. inż. Marek Sibielak

Auxiliary supervisor: dr hab. inż. Waldemar Rączka

Faculty of Mechanical Engineering and Robotics

Abstract: The research works is oriented around the problems of robotic systems control. The research concerns the use of neural networks in the problems of decision making and robot control. The goal of the research will be to develop and implement manipulator control strategy. One of the research areas will be the selection of a strategy for operation through a neural network based on the analysis of data from the vision system. Systems of this type are widely used in many fields of technology and industry. Currently the researchers all over the world is developing the application of neural networks to automate processes and make decisions.

Research facilities: The research facility of the unit is a Dynamic and Control of Structures Laboratory. The main purpose of the laboratory is to provide scientific and research activities in the field of controlled dynamic structures.

Research is carried on active and semiactive structures using electrohydraulic, electro-pneumatic and electrodynamic elements and assemblies.

The laboratory is equipped to specialized research rigs with PAC control systems and field-programmable gate arrays FPGA. They are used to test dynamic systems and structures. The operation of the test rigs is based on electrohydraulic vibration generators made based on original designs. In addition, the laboratory is equipped with a number of measurement systems with sensors and transducers for signals acquisition. The laboratory is equipped with SUV wheeled vehicles, which are autonomous platforms designed for experimental research.

https://kap.agh.edu.pl/en/dynamic-structures-and-systems-lab/

Number of places: 1

 

32. Investigation on quasi-ideal coupled-line sections' realization in monolithic technologies and their utilization in the design of microwave integrated circuits

Supervisor: dr hab. inż. Krzysztof Wincza, prof. AGH

Department of Electronics, Faculty of Informatics Electronics and Telecommunications

Abstract: The project aims at investigation on designing coupled transmission-line sections taking into account constraints that are inherent to monolithic technologies. The focus will be put on achieving superior performance of the designed sections and to investigate wheatear it is possible to design compact monolithic directional couplers featuring bandwidths exceeding one frequency decade.

The main goal of the project is to research the possibility of coupled-line sections’ design that have good return losses and good isolation properties, and at the same time feature the required coupling between excited and coupled lines. In general coupled-line geometries that will be considered within the project can be complex, consisting of a number of appropriately inter-connected conductors in inhomogeneous dielectric medium. The problem of such coupled-line section realization has been research and experimentally investigated in the techniques of strip-transmission lines for both geometrically symmetrical and asymmetrical conductors. It has been shown that the principal condition of ideal coupled-line section realization to be fulfilled in the case symmetrical conductors is the equalization of phase velocities of the waves propagating in the structure. Alternatively, it can be formulated as the equalization of effective dielectric constants for coupled line geometry under even and odd excitations. The more general condition for ideal coupled-line realization has been derived for the case of geometrically asymmetrical conductors for which it is required that the capacitive and inductive coupling coefficients are equalized. Additionally, in both cases an additional impedance condition needs to be fulfilled to obtain good return losses measured for the assumed reference impedance.

The project aims at developing methods for the design of coupled lines which allow to equalize capacitive and inductive coupling coefficients under the limitations specified by the chosen monolithic technology.

Research facilities: Access to the Laboratory for Microwave Techniques, that allows for full characterization of the microwave circuits in the frequency range up to 43 Ghz.

Number of places: 2

 

33. Development of algorithms for the detection of biomarkers of selected diseases in the exhaled breath using semiconductor gas sensors

Supervisor: dr hab. inż. Artur Rydosz

Department of Electronics, Faculty of Informatics Electronics and Telecommunications

Abstract: The aim of the work is to conduct research on exhaled breath analysis for the detection of biomarkers for selected diseases using semiconductor gas sensors based on the metal oxides. During the research, commercial sensors as well as sensors developed in the research team will be used. Semiconductor gas sensors have a number of advantages such as high sensitivity and ease of mass production. However, they also have the main disadvantage of cross-sensitivity to interference gases, hence the matrix of semi-selective sensors is used, and information on the concentration of a specific compound is generated on the basis of the response from all sensors, in this case it will be a biomarker, e.g. acetone, which is a biomarker for diabetes . The main goal of the research will be the development of front-end readout electronics for sensor array with variable sensor sizes along with detection algorithms based on neural networks and machine learning. The possibilities of using such matrices in modern Internet of Things (IoT) solutions will also be explored.

Research facilities: The works will be carried out in a modern laboratory Integrated Laboratory of Sensor Nanostructures at the Department of Electronics AGH. The laboratory is equipped with the necessary measuring equipment and infrastructure for research on gas sensors. The works will be carried out as part of projects whose supervisor is the promoter: SONATA National Science Center, pt. "Studies on the effect of GLAD technology on 3S properties (sensitivity, selectivity, stability) of gas sensors with an increased response to biomarkers of diabetes in the exhaled breath."

Number of places: 1

 

34. Spatial audio signal processing using machine learning methods

Supervisor: dr hab. inż. Krzysztof Wincza, prof. AGH

Faculty of Informatics Electronics and Telecommunications

Abstract: The main focus of the proposed research is to investigate and develop deep learning algorithms suitable for spatial processing of audio signals recorded using an array of microphones. The Ph.D. student will perform the research work into advanced audio signal processing methods with the main focus on sound event detection and classification, parameter estiamtion of an acoustic sound scene, as well as localization and signal estimation of sound sources. The research work will also concern the reproduction of the recorded spatial sound using headphones or loudspeakers.

Research facilities: The Ph.D. student will be provided with supervision of senior reserach staff and very good work conditions for performing the research work. The Ph.D. student will be provided with an office space in the Signal Processing laboratory, computer with the required professional software, high-class audio equipment (high-quality microphones, audio interfaces, sound cards, etc.). The databases with signals of various acoustic sound sources and an access to the high-power computational resources such as servers at the Department of Electronics and Cyfronet for performing the research work will be provided. The described research work will be performed along the research tasks of the NCN OPUS project entitled „Analiza zastosowania uczenia maszynowego w przestrzennym przetwarzaniu sygnałów dźwiękowych”, in which Dr Konrad Kowalczyk is the principle investigator. The aim of the project is to develop a set of audio signal processing techniques based on deep learning for spatial audio processing purposes.

Number of places: 1

 

35. Speech signal processing in challenging acoustic conditions

Supervisor: dr hab. inż. Krzysztof Wincza, prof. AGH

Auxiliary supervisor: dr inż. Konrad Kowalczyk

Faculty of Informatics Electronics and Telecommunications

Abstract: The rearch work of the Ph.D. student will be into advanced signal processing techniques to speech signals. The main focus of the reserach work will be into applying deep learning methods to speaker recognition/identification, separation of speaker signals, and signal enhancement in challenging acoustic conditions characterized by the presence of noise and room reverberation. The research work combines the elements of classical model-based signal processing (beamforming, Wiener filtering, non-negative matrix factorization) with the elements of machine learning, in particular those based on deep neural networks.

Research facilities: The Ph.D. student will be provided with supervision of senior reserach staff and very good work conditions for performing the research work. The Ph.D. student will be provided with an office space in the Signal Processing laboratory, computer with the required professional software, high-class audio equipment (high-quality microphones, audio interfaces, sound cards, etc.). The databases with audio signals and an access to the computational resources required to perform the research will also be provided including servers at the Department of Electronics and Cyfronet.

The research work will be performed within the research tasks of the First TEAM project entitled „Audio Processing Using Distributed Acoustic Sensors” financed by the Foundation for Polish Science, in which Dr Konrad Kowalczyk is the principle investigator. The aim of the project is to develop audio signal processing techniques which enable voice communication on distance for a person who is freely moving within the acoustic environment, based on signals recorded using distributed microphones.

Number of places: 1

 

36. Research on microwave measurement methods and biofunctionalized microwave circuits' design for microbiological sensory purposes.

Supervisor: dr hab. inż. Sławomir Gruszczyński, prof. AGH

Department of Electronics, Faculty of Informatics Electronics and Telecommunications

Abstract: The goal of the project is to research the novel selective bacteria detection methods with the use of biofunctionalized microelectronic structures in microwave frequency range. The project aims at developing novel biosensors based on microwave measurements of circuits’ properties with specifically bind bacteria species. The microwave sensors have been chosen due to their possible high sensitivity, selectivity and real-time operation. To achieve high sensitivity broadband measurements of circuits’ parameters, i.e. transmission-line impedances, and propagating constants are to be considered. The enhancement of detection results from the analysis of frequency dispersion of the measured parameters, which has been already preliminary verified by the authors of the project, and published in Biosensors and Bioelectronics journal, by measurements of capacitive sensors’ properties with bind adhesin-LPS of E. coli B bacteria. Within this project to ensure the high selectivity to the bacteria, their identification will be assessed by using specific LPS-recognition-molecules: LPS-binding proteins and antibodies which recognize the LPS molecules of bacteria surface. As a result the bacteria to be detected will be bind to sensor’s surface allowing for reliable measurements. The research within the project will be focused on the detection of the Gram-negative bacteria especially E. coli, however, the developed methods can be further utilized for the detection of other bacteria types upon change of the specific LPS-binding proteins. The developed methods can be applied to improve bacteria detection in clinical diagnosis, food analysis, bioprocess and environmental monitoring. Traditional methods are time consuming, laborious and require specialized measurement equipment. Therefore, it is justified to investigate the novel biosensors based on sensitive biological elements and microwave circuits. The approach proposed by the authors of the project will allow for developing of a novel and potentially very useful methods for realization of unique and cheap diagnostic techniques.

Research facilities: Access to the Laboratory for Microwave Techniques, that allows for full characterization of the microwave circuits in the frequency range up to 43 Ghz.

Number of places: 1

 

37. Application of light for activation of the resistive-type response of the sensor array to gases

Supervisor: prof. dr hab. inż. Katarzyna Zakrzewska

Department of Electronics, Faculty of Informatics Electronics and Telecommunications

Abstract: The aim of the PhD thesis is to application of light to activate the response of gas sensors. The research will lead to realization of the array comprising several sensors addressed individually by the light beam. Semiconducting gas sensors that are sensitive to the changing concentration of the detected chemicals usually require an elevated temperature of operation that causes an increase in the power consumption of the device. Illumination with a suitable wavelength and intensity of the light beam will eliminate the need of heating. This work has experimental and technological character.

Research facilities: Integrated Laboratory of Sensor Nanostructures at the Department of Electronics AGH, where the PhD thesis will be carried out is well prepared for this type of research. It is equipped with the measuring units and infrastructure for studies of the gas sensor responses. Research will be performed within the framework of the National Science Center project OPUS 12, entitled “Semiconducting n-n and n-p nano-heterostructures in optically supported systems of resistive type gas sensors”.

Number of places: 1

 

38. Application of artificial intelligence methods to electronic control of chosenprocesses

Supervisor: prof. dr hab. inż. Andrzej Kos

Department of Electronics, Faculty of Informatics Electronics and Telecommunications

Abstract: Artificial intelligence methods are commonly used to design learning process control systems. Theimplementation of the research topie begins with specifying the topie that the PhD student is interested in. ft is possible to implement an electronic intelligent control system, eg in moto-technique, sailing, mountaineering, defense industry, civil industry, etc. The topie will be refined together after getting acquainted with the personal interests of the doctoral student.

Research facilities: The supervisor guarantees the provision of an individual office and full access to all the equipment necessary to carry out the research problem successful/y. There is also a possibility to carry out research within scientific projects.

Number of places: 1

 

 

39. Thin film, silicon free photovoltaic cell CuOZnS.

Supervisor: dr hab. inż. Konstanty Marszałek

Faculty of Computer Science, Electronics and Telecommunications

Abstract: Silicon photovoltaics is well known and technically become the most popular technology with almost maximum available parameters so to lower production costs and find popular chip application there are a lot of investigations of thin film materials (CIGS, Perovskites, CdTe, etc) as a new candidates for nonexpensive photovoltaic power plants which instalations are growing up extremally fast in the world.. One of quite new candidate for silicon free cell are thin film of p-type CuO and n-type ZnS prepared in chip chemical processes. Investigation of these materials as well as photovoltaic diode CuO/ZnS as competitors for silicon one, starting from physical properties through diode characteristic up to application this cellto energy generation will be yhe aim of proposed PhD thesis.

Research facilities: Electronic Dept. is equipped with technological equipment for thin film deposition. We have also photovoltaic lab. Nap outdoor photovoltaic lab. In Miekinia village nearby Krakow. PhD student could make most of necessary experiment in these labs. Additionally we have access to Photovoltaic Lab. In Kozy and Technological lab CEZAMAT in Warsaw where doctorant could provide additional technological processes and mesurements.

Number of places: 1

 

40. Hardware acceleration of artificial intelligence algorithms.

Supervisor: dr hab. inż. Ernest Jamro

Faculty of Computer Science, Electronics and Telecommunications

Abstract: One of the main problems of artificial intelligence is insufficient available computing power. Therefore, special hardware accelerators are used, eg. GP-GPU, programmable devices FPGA, dedicated processors, ASIC circuits that enable acceleration of the most time-consuming calculations. The subject of the proposed work can develop one (or more) of the below issues:

  • examining various accelerators existing on the market, checking their effectiveness and proposing own accelerator architecture,
  • designing an FPGA system, with a properly selected architecture that can include, for example, reducing the bitwidth of calculations, skipping multiplication by zero (sparcity of coefficients, input data),
  • modification of existing AI algorithms so that they require less hardware resources or were simply more adapted to the available hardware resources,
  • compression of input data, coefficients of neural network, so as to reduce data transmission from external memory

Research facilities: The supervisor and the research team have extensive experience with AI algorithms, in particular their implementation in FPGA chips and other hardware accelerators. It is possible to use the computational resources of Academic Computing Center Cyfronet AGH (which plans to buy AI accelerators) or other computing centers.

Number of places: 1

 

41. Use of Software Defined Radiotechniques to improve the operational performance of radio systems.

Supervisor: dr hab. inż. Witold Machowski

Auxiliary supervisor: dr inż. Cezary Worek

Faculty of Computer Science, Electronics and Telecommunications

Abstract: The research will address both design and low-level signal processing techniques to improve the performance of radio systems using Software Defined Radio techniques. The study will focus on methods of the selection of appropriate system solutions and components, verification of their parameters, PCB design, and modeling of high-frequency phenomena in electrical circuits. Additionally, research will address low-level signal processing techniques. Abovementioned activities are to lead to the improvement of achievable parameters of radio receivers and transmitters. One of the objectives of the research will be to verify the possibility of increasing the sensitivity of receivers, reducing the impact of any unwanted signals, improving the spectral quality of transmitters, etc. Software Defined Radio systems are systems in which the operation of basic electronic blocks (mixers, filters, modulators and demodulators, detectors) is performed through a computer program. Thanks to this, the obtained parameters can be significantly improved, e.g., by reducing the influence of variable parameters of electronic components. Additionally, they can be easily reconfigured, prototyped, and verified during various types of tests. For signal processing to be carried out, it is planned to use mixed techniques based on multiprocessor software and implementation of critical algorithms in FPGA systems to increase the efficiency of data processing.

Research facilities: Wireless sensor and control networks, Department of Electronics, IET Faculty, AGH University of Science and Technology in Krakow (http://www.wsn.agh.edu.pl/?q=en) Programm PW-Sat2 is a student satellite project started in 2013 at Warsaw University of Technology by the Students Space Association members. PW-Sat2 is a 2-unit CubeSat satellite, designed for Low Earth Orbit. PW-Sat2 was launched on 3rd of December 2018 onboard SpaceX Falcon 9 rocket. Research project on the development of a base station "Industrial research and development work on the development of a prototype of a new type of battery-powered water counter that will communicate with an autonomous base station for at least 10 years in a wireless, fully automatic manner".

Number of places: 1

 

42. Deep neural networks in computer vision.

Supervisor: dr inż. Joanna Jaworek-Korjakowska

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: In recent years, deep neural networks have frequently been used in the field of computer vision and image processing. They give a chance to solve current problems, in which full automation has not yet been possible. These are, for example, the issues of analyzing medical images or images with a high level of noise. Technological development, associated with the increase of computing power of modern computers, allows the development of new, much more complex architectures of neural networks. Thanks to this, it is also possible to use ensemble methods that combine the advantages of many different algorithms.

Research facilities: This problem requires access to devices with large memory resources and computing power. Research facility is the Academic Computer Center CYFRONET AGH, which provides access to high-performance computers (supercomputers).

Number of places: 1

 

43. Deep neural networks in anomaly detection in diagnostic signals.

Supervisor: dr inż. Joanna Jaworek-Korjakowska

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The main goal of the National Synchrotron Radiation Centre SOLARIS is to provide scientific community with synchrotron light. In order to perform high quality research it is essential to monitor subsystems that are responsible for beam stability. The main goal of this research problem is to propose a deep neural network architecture for anomaly detection in diagnostic signals. Its task is to identify abnormal status of the sensors and classify them.

Research facilities: Research carried out based on the data from the control and diagnostics system of the National Synchrotron Radiation Centre SOLARIS. The Centre also provides the memory, storage and computational resources necessary to carry out the scientific work. If there is a need to make calculations that go beyond the capabilities of the Centre, the additional facility is the Academic Computer Center CYFRONET AGH (high-powered computers).

Number of places: 1

 

44. Methods of machine learning for event-driven data

Supervisor: prof. dr hab. Mirosław Pawlak

Auxiliary supervisor: dr inż. Dominik Rzepka

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The objective of the research in the frame of proposed research problem is development of new methods for classification, clusterization, segmentation, prediction or detection of anomalies for phenomena and systems where significant data is produced at the instant when the predefined event occurs. The proposed scope of studies consists of analysis of time series, including nonuniform series, methods of detection and defining events, as well as verification of possible effective solutions using state-of-the art methods of machine learning and learning sequences. Furthermore, the proposed research is going to show the applicability of the proposed methods for biomedical engineering.

Research facilities: Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering of AGH offers research development facilities including both the scientific equipment and effective academic supervising. The supervisor has been an author of multiple research publications within the scope of the proposed research topic and related topics. The research is going to be carried out in the cooperation with foreign co-investigators. The proposed topic is partially covered by the scope of the research project founded by National Science Centre.

Number of places: 1

45. Autonomous anticipatory systems

Supervisor: prof. dr hab, inż. Andrzej M. Skulimowski

Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering

Abstract: The growing complexity of autonomous systems and intelligent technologies raises new research questions, including a need to investigate artificial intelligent and autonomous decision systems (AADS) – a new class of systems defined in https://link.springer.com/chapter/10.1007/978-3-319-11298-5_14. The related research methods include anticipatory networks (http://www.tandfonline.com/doi/full/10.1080/00207721.2012.670308,  https://doi.org/10.1007/978-3-319-51969-2_6), causal decision systems, the theory of freewill and creativity in artificial systems, multicriteria optimization and optimal control, cognitive architectures. The doctoral research will also encompass real-life applications, such as coordinated control of swarms and formations of inspection robots, research robotics, decision support systems (DSS) and expert systems dedicated to solving real-life problems. The doctoral research will be carried out in cooperation with the team members of the Decision Science Laboratory, benefitting from the synergies with other research areas such as global expert sys­tems (GES), data mining, machine learning, predictive analytics, quantum and population-based algorithms and operations research.

Research facilities: PhD students will get a workplace at the Decision Sciences Laboratory, Department of Automatic Control and Robotics. The laboratory is equipped with rich computer and multimedia infrastructure, advanced software, such as robot simulators, anticipatory network analytics, specialized decision support systems (DSS) and expert systems. The students will be able to take part in preparing and carrying out international research projects related to the area of their doctoral research, specifically projects within the EU Horizon 2020 and Horizon Europe programs.

Number of places: 3