It can detect radiation as well as enemy units. Demonstration of AI-controlled swarm of drones and robots impressed both army and firefighters

(Press Release, Brno, 9 October 2025) Scientists from Brno University of Technology (BUT) and the University of Defence (UD) have been carrying out a project focused on the development of a swarm of drones and robots controlled by artificial intelligence algorithms for four years now. In August and September, they demonstrated the system to representatives of defence and rescue forces. Two experiments carried out at the Military District of Březina near Vyškov confirmed its potential for intelligence and reconnaissance use – in detecting radiation sources and in mapping areas containing enemy military equipment.

A team of researchers and developers of drone swarms and robots from FEEC BUT. | Author: Jakub Rozboud

The predecessor of the project, officially titled An Artificial Intelligence-Controlled Robotic System for Intelligence and Reconnaissance Operations, can be considered the Orpheus series of ground robots, developed at BUT since 2003 for both military and civilian applications. The upgraded Orpheus is part of the current system, which also includes two other types of ground robots, three types of drones, and an operator station with communication infrastructure. The result is a comprehensive system that, thanks to advanced automation, requires minimal operator intervention.

Unmanned ground vehicle Orfeus. | Author: Jakub Rozboud

“Conventional drones are usually operated manually, where one drone is controlled in real time by one operator. Our project is much more complex. A large swarm of drones accompanied by several ground robots can be operated by a single operator, who assigns a mission, and all machines autonomously execute it as one coordinated system. Moreover, since we use multiple machines equipped with various types of sensors providing different but relevant information, we can also reduce the need for additional specialists,” explained project leader Petr Marcoň from the Faculty of Electrical Engineering and Communication, BUT (FEEC BUT).

Delta – the most commonly used drone in the swarm, flight endurance 40 minutes with a full payload of 2 kg, equipped with sensors, e.g., for generating a digital terrain model;
Falcon – the largest drone in the swarm, flight endurance 24 minutes with a full payload of 9 kg, used to transport the ground robot Hermes;
Ares – drone with 1 kg payload capacity, 40-minute flight endurance, used for delivery missions.
The swarm is complemented by unmanned ground vehicles (robots) Orpheus, Scout, Hermes and the operator station.

All software and most of the hardware were developed by the project team or are open source, which contributes to the system’s cybersecurity.

The Falcon drone performing its mission. | Author: Jakub Rozboud

Specialists from the Czech Armed Forces and the Czech Fire Rescue Service (HZS) were able to observe a demonstration mission in the field during August and September 2025. This represented the long-planned culmination of the project before its conclusion in December 2025. “We simulated two possible scenarios. In August, it was the case of a dirty bomb used in a populated area. We identified and measured radiation to assess whether the environment was safe for soldiers and civilians. The second scenario in mid-September involved reconnaissance of a large military zone – to determine whether enemy troops or armed vehicles were present within the defined area,” said Petr Marcoň.

First Experiment – Detection of a Radioactive Source

The UD experts were responsible for preparing the scenarios and for tactical assessment of the mission. The first scenario simulated the movement of units from Vyškov northwest toward the Jeseník locality. During their advance, the drone swarm detected a weapons cache containing weapons of mass destruction emitting gamma radiation. “The aerial reconnaissance identified the threat, which was subsequently confirmed by ground robots. Thanks to the sensors on both drones and robots, the unit commander could visualize the detected source and then decide how to proceed with securing it,” explained Jan Nohel from the Faculty of Military Leadership at UD. He praised the fact that the project’s communication system is compatible with the army’s BVIS information system and that the acquired data can be shared with other military units.

Preparations of the swarm before the experiment – Delta drones. | Author: Jakub Rozboud

“The experiment was particularly valuable because we were able to work with real radioactive sources and test the swarm’s radiation detectors (part of so-called CBRN detectors, editor’s note),” added project co-investigator Luděk Žalud from FEKT BUT.

Second Experiment – Reconnaissance of Enemy Territory

The goal of the approximately 15-minute experiment was the identification and mapping of enemy positions and military equipment using eight drones cooperating with three ground robots over an area of about 40 hectares. Experts from FEEC BUT demonstrated all phases of the mission – from planning (mission setup, area definition, and task allocation) to preparation, execution, and post-mission data processing. “In the demonstration, we combined all available drone route planners – a metaheuristic planner using AI algorithms for route optimization, a systematic planner that scans the area in detail, and a manual planner. In real operations, one type chosen by the operator would suffice, usually the one ensuring the fastest acquisition of information, but we wanted to show how the options can complement each other,” explained Petr Marcoň.

A swarm of drones and robots begins a reconnaissance mission at the operator's command with the aim of locating enemy military equipment. | Author: Jakub Rozboud

Each drone recorded the entire mission using different types of sensors. They are equipped with RGB cameras, multispectral imaging, thermal cameras, laser scanners, and the aforementioned radiation detectors. The collected data can be transmitted in real time as a video stream, although this is not always advantageous in a combat environment. “Due to high data loads, it is better to limit transmissions to short text messages and download full data after the mission. Then, using a powerful neural network, the data can be processed into high-resolution 3D maps with real-time positions of military equipment,” said Luděk Žalud.

What the Experiments Showed and the Follow-up Project

The speed of data processing and the resulting highly accurate georeferenced maps impressed the representatives of the Czech Armed Forces. “The outputs enhance the operational awareness of the unit commander in areas that are otherwise inaccessible. In doing so, they help protect the health and lives of soldiers and other specialists,” said David Anthony Procházka from the General Staff of the Czech Armed Forces.

Participants in the event were able to follow the drones' routes and the progress of the experiment. | Author: Jakub Rozboud

“The goal of the experiments was to prepare scenarios that would demonstrate the current state of the project, including its strengths and weaknesses. We already know that in the follow-up project, we will focus on more resilient communication tools and localization issues, since GPS data – on which the current swarm still relies – can be jammed or spoofed,” noted Luděk Žalud. He added that interest in the system from various public and civilian entities has exceeded the researchers’ initial expectations.

The follow-up project, titled Multirobot System for Autonomous Field Operations with Multisource Navigation, will address, among other things, communication and localization of drones. Since 1 July 2025, it has been supported by the TAČR PRODEF program. The main investigator is BUT, with partners including the University of Defence, companies LTR and Fly4Future, CTU in Prague, and the Military Research Institute.

“I have seen many projects focused on the use of unmanned systems, but compared to them, this joint project of BUT and UD can truly address a real problem and translate it into a user-applicable solution. There are increasingly more situations where we could deploy such a system,” emphasized Martin Nekula, specialist of the Fire Rescue Service of the Moravian-Silesian Region.

Representatives of the Fire Rescue Service also participated in the subsequent discussions. | Author: Jakub Rozboud

“We have experience deploying drones during emergency events such as fires or chemical accidents. We can offer the project a broader ‘playground’ and invite its representatives to real-world operations,” added Petr Ošlejšek, Deputy General Director of the Czech Fire Rescue Service, noting that he sees the project’s potential in creating a unified platform and synergy for different users.

“We designed the project even before the war in Ukraine. When it began, a number of technical and military questions emerged that revealed new areas for our future work. We are pleased to be able to continue this research and are convinced that through it, we are helping to address real problems within the security community,” concluded Alexandr Štefek, Vice-Rector for Digitalization and Informatization at the University of Defence.