Reseach topics

We actively seek for a cooperation with students at all levels of their studies. Come and join us to work on challenging, real-life oriented problems. We offer to students various research topics for their thesis and projects. Some of them are more industry orientated (i.e. working on a task required by some of our industry partners, solving their real-life problem), others are more theoretical (i.e. focusing on developing new fundamental algorithms and theory).

In the list below you can see a short description and motivation for solving the problem/project. After the student chooses the area of his/her work, the topic is more deeply specified. If you still hesitate, just make an appointment with sojkam1atfel [dot] cvut [dot] cz (Michal Sojka) (@wentasah) or Přemysl Šůcha (suchapatfel [dot] cvut [dot] cz) and come to see what we are working on.


Spolupráce s Fakultní Nemocnicí Hradec Králové

Cílem spolupráce je návrh algoritmů pro plánování péče o pacienty. Cílem je vyvinout algoritmus pro tuto oblast, schopné pracovat s neurčitostí parametrů. V rámci projetu je možná stáž v rámci programu Erasmus v Belgii.


Optimalizace provozu klinické laboratoře

Chceme navázat na úspěšný projekt s laboratoří PREVEDIG, kde se nám podařilo pomocí optimalizačních algoritmů a datové analýzy zvýšit výkon laboratorního systému o cca 30% ( Práce zahrnuje analýzu dat a návrh algoritmů

EATON Lab projects

EATON offers a large number of industrial projects related to embeded devices that will be applied to modern manufacturing facilities. Find out more here.


Využití strojového učení pro řešení kombinatorických problémů

Strojové učení je jedna z možnosti, jak zefektivnit prohledávání prostoru řešení kombinatorických problémů. Ukazuje se, že velmi zajímavých výsledků lze dosáhnout, pokud se spojí znalosti z oblasti kombinatorické optimalizace s možnostmi hlubokých neuronových sítí. Práce zahrnuje návrh neuronových sítí a vývoj algoritmů.

Machine learning for production data analysis

The data available in the information system of the production company are often incomplete. Namely absence of the data like processing time and changeover time makes it nearly impossible to construct production plans and schedules. On the other hand, there is often a correlation between a production order in hand and a similar one in the historical data comprising real production time measured at the factory floor. In order to capture this correlation, we aim to use Machine Learning techniques. Due to our collaboration with industry, the data from the real production will be used to derive processing times and changeover times in order to optimize production.

Distributionally robust stochastic optimization

The field of stochastic optimization deals with the decision making under uncertainty of parameters (e.g., duration of activities, electricity consumption, profits). The problems are typically stated as mathematical optimization problems with the aim of minimizing the expected value of suitably defined loss function concerning the selected probability distribution over the parameter space. Although the stochastic optimization provides more realistic and robust decision making for real-life problems than deterministic optimization, its applicability is still relatively limited to the problems where the form and parameters of uncertainty are precisely known. Hence, in recent years, distributionally robust stochastic optimization has started to attract more attention. In contrast to regular stochastic optimization, its distributionally robust counterpart enables to optimize with respect to a broad set of different distributions, which improves the robustness, resource consumption, and profits. The goal of this project is to investigate the computational tractability of different ambiguity sets, decision rules, their performances, and practical applications to scheduling problems (we recommend this topic to more mathematically-inclined students).

Detailed list

Circular economy

Waste is becoming one of the mankind's most serious problems. The population keeps growing and the world is already drowning in garbage. A system of circular economy aims at the continual use of resources and eliminating waste. Circular systems employ reuse, sharing, repair, refurbishment, remanufacturing, and recycling to create a closed-loop system, minimizing the use of resource inputs and the creation of waste, pollution, and carbon emissions. The circular economy aims to keep products, equipment, and infrastructure in use for longer, thereby improving the productivity of these resources. Waste materials and energy should become input for other processes. The circular economy presents numerous challenges. We aim to address these challenges using combinatorial optimization methodology.

F1/10 -- Model of an autonomous car

Self-driving cars are future of public transportation and a lot of skilled engineers will be needed to make this technology safe and reliable. This topic could be your entry to this future world. The goal is to improve the algorithms in our current car and participate in F1/10 Autonomous Racing Competition ( with that car. In addition to that, we collaborate with industrial partners on the development of real autonomous cars and the techniques developed here may be ported into real cars.

Thermal-aware scheduling for embedded systems

Many devices need to satisfy strict thermal requirements without using external fans or massive heat sinks. As an example, we consider embedded systems used in avionics, for which the additional weight and power constraints need to be taken into account. Since the avionics systems need to enforce critical timing and safety guarantees, it is crucial to design and implement thermal/power-aware management techniques for them. The student will observe the behavior of a real embedded system, measure and analyze data, and propose novel thermal/power-aware management policies. The proposed policies will then be tested on real hardware, and their benefits will be evaluated.

Detailed list


Implement State of the Art Path Planning Algorithm for Parking Car

The current boom in the vehicular industry is self-driving capabilities. Semi-automated vehicles are gaining more and more advanced features like adaptive cruise control, line keeping, or parking assistance. The task of this topic is to review the current state of the art algorithms for path planning of automated parking cars, choose one, and implement it.

Detailed list

Data-driven approach for scheduling problems

Deep learning is a rapidly growing discipline in the last ten years for many scientific disciplines. However, the way to utilize the neural network for scheduling is unclear and rarely studied in the literature [2]. Nevertheless, data-driven approaches have clear advantages for solving combinatorial problems, as has been recently shown in [1]. The target of the project is to propose an effective data-driven algorithm for the selected scheduling problem.

Detailed list


[1] Bouska, Novák, Šůcha, Módos, Hanzálek. "Data-driven Algorithm for Scheduling with Total Tardiness" ICORES. 2020

[2] Vinyals, Oriol, Meire Fortunato, and Navdeep Jaitly. "Pointer networks." Advances in neural information processing systems. 2015.


image credits:

Scheduling of surgical cases

Operating theaters are typically one of the most expensive facilities for every bigger hospital. Therefore, their utilization needs to be carefully planned. The goal of this topic is to design an algorithm able to schedule individual surgeries to maximize the utilization of available operating theaters while taking into account the uncertain duration of individual surgical cases. The experiments will be carried on real data.


[1] Shuwan Zhu, Wenjuan Fan, Shanlin Yang, Jun Pei, Panos M. Pardalos, Operating room planning and surgical case scheduling: a review of literature, Journal of Combinatorial Optimization, April 2019, Volume 37, Issue 3, pp 757–805.

[2] Wang Yu, Zhang Yu, Tang Jiafu, A distributionally robust optimization approach for surgery block allocation, European Journal of Operational Research, Volume 273, Issue 2, 1 March 2019, Pages 740-753.


image credits:

Take on the intersections (smart infrastructure and autonomous cars)

Nowadays, autonomy of driving is tightly connected to highways and navigating in well-known closed areas. In the future, when the autonomous cars reach the edge of the cities, new scenario comes up -- intersections. Come and join us on creating a model of an intersection of the future for F1/10 car, which can be later expanded to real cars.


image credits:

Optimization of laboratory workflow

Medical laboratories play an irreplaceable role in the diagnosis of patients. Analysis of medical samples is very important for the right diagnosis of patients. Moreover, the results need to be available on time. Therefore, every laboratory is trying to minimize the response time, known as Turn Around Time (TAT). TAT is a difference between the time when the results of the analysis are reported and thetime when the sample arrives at the laboratory. The goal of this thesis is to analyze several different scenarios of laboratory automation. Then, base on the analysis, propose an algorithm able to optimize the laboratory workflow leading to a reduction of overall TAT. The experiments will be carried on real data.

Detailed list


[1] Wenhua Li and Xing Chai. The medical laboratory scheduling for weighted flow-time. Journal of Combinatorial Optimization, Nov 2017.

[2] Ali Azadeh, Milad Baghersad, Mehdi Hosseinabadi Farahani, and Mansour Zarrin. Semi-online patient scheduling in pathology laboratories. Artificial Intelligence in Medicine, 64(3):217 – 226, 2015.

Fundamental research in scheduling & open problems

Sometimes, even the fastest computers are not able to solve the given combinatorial problems in a reasonable time. This does not necessarily mean that these problems are impossible to solve, but a deeper theoretical understanding of the problem is required to implement efficient algorithms. The task of this topic is to study fundamental properties of the given scheduling problem and implement efficient algorithms that exploit these properties (we recommend this topic to more mathematically-inclined students).

Reliable hardware for autonomous cars

Future self-driving cars will require not only tremendous computational power for processing data in real-time but also increased reliability and fault-tolerance. While sufficient performance is easily provided by modern hardware, the reliability of such hardware is not considered in "automotive-grade". This topic addresses the challenge of increasing the reliability and determinism of Xilinx UltraSCALE platform, used by our industrial partners in their autonomous cars, by implementing so-called "Predictable Execution Model" on this platform and improving it by using new hardware features of the platform.

Detailed list

Scheduling of the Time-Triggered communication on Deterministic Ethernet

The Ethernet technology is one of the most widespread technology in the world. Nowadays, the network has the usage even in the areas where it was not intended. It becomes more and more common even in highly critical applications, where any failure can cause jeopardy of life, because of the increasing bandwidth demands of e.g. autonomous driving systems. The car or airplane manufacturers endeavor to modify the Ethernet protocol to be also capable of the deterministic and hard real-time communication - TTEthernet, 802.1Qbv, 802.1Qbu and 802.3br. In such a time-triggered Ethernet standards, the communication follows a schedule known in advance. Design and development of an algorithm, which is able to create efficient schedules, is the main objective of this diploma thesis topic.

Detailed list



Industrial Informatics Research Center leads the Computer Engineering study branch of popular study program Open Informatics at the Faculty of Electrical Engineering of CTU in Prague. For more information check out the site for Computer Engineering [in Czech]

IIRC secures a number of undergraduate and graduate courses. See the links below.