News

On the morning of Friday, July 10, things were really hopping at the airport in Moravská Třebová. However, it wasn’t the members of the local flying club who were responsible—it was the student teams with their rockets. The finals of the nationwide Czech Rocket Challenge 2026 were taking place there. This is a team competition for high school and college students, in which they design, test, build, and launch their own rockets. The evaluation is not based solely on the altitude reached (which is usually around 500 to 600 meters), but also on whether the teams are able to land the rocket; other criteria include the sophistication and complexity of the design, the team’s report, the research conducted, and so on. The event is organized by the Czech Rocket Society, an association of college students who are passionate about aerospace and rockets. The Faculty of Information Technology was represented at the event by a new student team called Testing in Production, consisting of Petra Továrková, Jan Kostka, David Daniel Šlapal, David Alois Bureš, and Štěpán Oram. The team was one of 18 (out of a total of 70 registered teams) that made it all the way to Friday’s competition finals.

Our freshmen performed very well: Even the successful rocket launch on Friday alone is no small feat. “The rocket launched flawlessly from the launch pad, successfully deployed its parachute, and landed safely — which is also a critical moment in the entire process. Using a GPS module and drone tracking, we were able to locate it quickly,” says Petra Továrková, describing the highlights of the two-day event in Moravská Třebová, adding: “A remarkable achievement for our team was winning the category for the most accurate calculation and estimation of the apogee, that is, the highest point of the flight.” You can view a web visualization of the rocket’s flight here.

Connecting Different Worlds

The rockets themselves, weighing about one kilogram, catch everyone’s eye at first glance. The one from the Testing in Production team is designed as a modular platform, roughly 90 cm long, that can be easily disassembled and serviced right in the field. Its main structural element is a steel threaded rod running through the center of the rocket, which connects the individual parts and transfers the load during launch and parachute deployment. The outer PVC fuselage serves primarily as an aerodynamic fairing. The rocket features a 3D-printed nose and tail section, a mechanical rescue system with a spring and parachute, and onboard electronics that use sensors to monitor flight progress and transmit data to the ground station.

Behind the physical object, of course, lie dozens of hours of less visible work in the fields of information technology, embedded systems, and software engineering. The subsequent step from the predictable world of software into the harsh environment of physical reality may represent the greatest challenge for the Testing in Production team. Jan Kostka confirms this: “In IT, we’re used to the fact that when code doesn’t work, it throws an error; we fix it and try again. But physics doesn’t give you a second chance. Suddenly, we had to figure out from scratch the aerodynamic shape and design of the rocket with the correct center of gravity so that it would fly smoothly and stably. We had to select materials that were light enough but could also withstand the enormous forces, vibrations, and G-forces during engine ignition. Integrating our software with the hardware so that everything would function reliably in this uncompromising real world was a huge and incredibly valuable learning experience for us.” Precise integration of software and mechanical design is the key to success—this included implementing reliable real-time telemetry data transmission to the ground station and the ability to remotely control the rocket during flight, including, for example, a mechanism for emergency parachute deployment.

The competition projects also include conducting specific research. The team from FIT focused on software analysis of the triboelectric effect (or static electricity) generated during flight. Static electricity poses a real risk in the aerospace industry: It can damage onboard avionics, produce electromagnetic interference that can lead to a loss of communication with the ground station, and even cause degradation of surface materials. The experiment aims to measure the triboelectric voltage during individual phases of flight, verify the resilience of an electrical circuit exposed to this voltage, and assess the extent of the risk involved.

The Beginnings and Future Visions

Building and testing rockets isn’t exactly your average hobby. How did the student team from FIT get into this? “Actually, it was a complete coincidence. The team’s founder is Petra Továrková, who first heard about the possibility of building student rockets at a professional conference. She was immediately excited by the idea. When we were sitting around drinking punch at Christmas, she brought it up as a topic of discussion and right away suggested that we join her and start a team,” recalls David Daniel Šlapal, reflecting on the project’s beginnings. Other members with their own expertise subsequently joined. “Even though I’m studying IT, I’m very interested in 3D printing, engineering, and engine design. When I found out the team needed someone to help with the rocket design and models, I jumped at the chance,” says Štěpán Oram, describing how he got involved. All members are motivated by the opportunity to gain experience far beyond the boundaries of everyday computer science — the fact that they can not only write functional code but also directly integrate it with a complete hardware design and mechanical construction that will then literally fly into the sky.

Working on a project like this while studying computer science — especially as a freshman — isn’t for everyone. Faculty projects have strict deadlines of their own, and working on the rocket adds dozens of hours of development and testing on top of that—often at the expense of free weekends and peaceful nights. “Seeing the results of our own work come to life in reality is our greatest motivation. In IT fields, you often only see the output as text or graphics on a computer screen. The moment when you plug in a board you designed yourself, program it, and watch with your own eyes as the rocket successfully launches, communicates, and you receive telemetry from it in real time—that’s indescribable,” says David Alois Bureš, identifying the main source of energy for everyone.

The members of the Testing in Production team view this year’s Czech Rocket Challenge as a springboard and a source of valuable know-how. “Our initial goal was to find out whether it was even possible to combine two such different worlds and learn something new in the process. Through the competition, we mastered the basics of rocket science, tested our systems in practice, and put our teamwork to the test. In the long term, however, our ultimate goal is to shift our focus toward much larger and more complex projects. And as for rockets, we’d like to qualify for one of the major international competitions next year,” says Petra Továrková, outlining the team’s next steps, which will help popularize the fields of computer science and software engineering in the space industry. Space is no longer just about mechanical engineering; software plays a critical role in it. “We want to spread this idea among our classmates in the FIT community and among the general public,” Petra concludes our conversation about the Testing in Production project.

We wish our students perseverance and much success. The work they’ve taken on is certainly not easy.

We invite you to the public defense of the dissertation by Amrutha Prasad from the Department of Computer Graphics and Multimedia, FIT, BUT, which will take place on Thursday, July 9, 2026, at 10:00 a.m. in conference room L220. The thesis advisor for the dissertation titled “Parameter-efficient models for domain-specific speech and language recognition” is Assoc. Prof. Ing. Petr Motlíček, Ph.D.

Amrutha Prasad collaborates with Assoc. Prof. Motlíček at the IDIAP research institute in Martigny, Switzerland, which is an internationally renowned center focused on artificial intelligence, machine learning, and speech processing. She is a member of the faculty research group Speech@FIT. In her research, she specializes in automatic speech recognition, language and speaker identification, and the use of speech technologies in air traffic control communication. Her research findings have been published at leading academic conferences, such as Interspeech, IEEE ICASSP, IEEE ASRU, Odyssey: The Speaker and Language Recognition Workshop, and others. She has also participated in international research activities related to the European ATCO2 project, focused on air traffic communication, and was a member of the broader Swiss national team participating in the NIST SRE24 speaker recognition evaluation.

Amruth’s dissertation addresses the question of how to reduce the size of large-scale speech recognition models without significantly compromising their accuracy. Current models, such as XLS-R and Whisper, are very effective at transcribing speech to text and identifying the language spoken in a recording. However, they contain a huge number of parameters and therefore require significant memory capacity, high computational power, and consume a lot of energy. This poses a particular problem in applications where the system must operate quickly and reliably, especially on less powerful hardware or with limited energy resources. This dissertation focuses primarily on communication in the field of air traffic control (ATC). This is an exceptionally challenging environment: Recordings are often affected by background noise, speakers have different accents, the quality of the radio signal fluctuates, and communication is concise and highly technical. At the same time, air traffic control is a safety-critical field in which errors can lead to serious consequences.

The dissertation focuses on two main areas: ASR—automatic speech recognition, i.e., the conversion of spoken language to text; and LID—language identification, which determines the language spoken in a recording. The author’s first research approach involves reducing the final components of the model, which, for example, determine which language was detected. A standard time-delayed neural network, referred to as TDNN, is replaced by a variant called TDNN-F, which uses parameters more efficiently. This model achieves similar capabilities in distinguishing between languages while requiring approximately 30 to 50% fewer parameters. The second approach utilizes a method known as normalization flows. This method allows for more flexible modeling of complex speech data distributions than the traditional PLDA method, which is based on the relatively simplifying assumption that the data approximately follow a Gaussian distribution. On two of the three evaluated datasets, this approach achieved higher accuracy than the PLDA method. The third and most significant part of the dissertation deals with direct reduction of the core of the XLS-R and Whisper models. The weight matrices in the forward layers of these models are replaced by a low-rank approximation created using singular value decomposition (SVD). Only the first r principal components are retained. Simply put, the model attempts to preserve the most important information and eliminate less essential parts of the computation. This method is also combined with the LoRA technique, which allows the model to be adapted during training using a significantly smaller number of parameters. The advantage of the proposed solution is that the model operates more efficiently not only during training but also during actual use, i.e., during inference.

What are the main results of the dissertation? For the XLS-R model, the number of parameters was successfully reduced by approximately 23–50%, with only a minimal decrease in accuracy. For the Whisper model, reducing the number of parameters by 50% did not cause a deterioration in performance during automatic speech recognition on a dataset from the field of air traffic control. However, according to the results, a certain decline in performance was observed for other datasets and tasks. When the number of parameters was reduced by 28%, the error rate actually decreased by 1.8%, meaning that transcription accuracy improved slightly. One possible explanation is that compression removed some redundant parameters and acted as a form of regularization—that is, a way to prevent the model from overfitting to the training data. The main conclusion of the study is that large speech processing models are often unnecessarily large. Their size and operational costs can be significantly reduced without substantially limiting their performance. This can facilitate their deployment, for example, directly on local or edge devices, in systems with limited memory and computing power, in real-time applications, and also in safety-critical fields such as aviation, as Amrutha Prasad states in her dissertation.

We wish Amrutha a successful defense and the best of luck in her future career!

The graduation ceremony for graduates of the MITAI Master’s program will take place on July 2, 2026, in Auditorium D105 (Božetěchova 2, Brno).

The schedule is as follows:

July 2, 2026

• 1. Group 1 at 9:00 a.m.: from the letter A through De, inclusive
• 2. Group 2 at 10:30 a.m.: from the letter Do through J
• 3. Group 3 at 12:00 p.m.: from the letter K through Mas
• 4. Group 4 at 1:30 p.m.: from the letter Mat through Raj
• 5. Group at 3:00 p.m.: from the letter Ruc through Z

Rehearsals for the graduation ceremony will take place half an hour before the ceremony, at 8:30, 10:00, 11:30, 13:00, and finally at 14:30.

Please arrive on time!

Flowers will be available for purchase.