The first of the projects presented by MRS Director Prof. Martin Saska and his colleagues is the development of a secure drone for monitoring and protecting critical infrastructure—such as nuclear power plants, military facilities, and key industries.
“The goal is to develop an advanced flying robot with onboard artificial intelligence, whose components and software come from domestic or European companies and research institutions. It is thus completely independent of non-European components, which may pose a potential threat of industrial and security espionage,” described Prof. Saska, noting that research into secure robotic systems is based on Czech know-how. More than a hundred researchers from the MRS group and the CTU spin-off Fly4Future are participating in the project.
“The biggest challenge is ensuring the absolute security of the robotic platform. Any component in a drone that collects data from any critical infrastructure can be a potential threat and exploited for espionage. There have been cases where seemingly harmless chips in drones connected to the network and sent collected data even several weeks after deployment. “For example, in the U.S., NDAA-compliant drones are not allowed to use even a single component originating from countries interested in espionage. Similarly, we are developing a unique solution that we will have full control over at the European level,” explained Prof. Saska.
A Safe and Unique Solution in Practice
The project is now in its second year, and the team has already achieved a number of notable successes. “In collaboration with Fly4Future, we’ve managed to develop a reliable platform that’s weather-resistant, modular, and open-source. This means it can be equipped with various sensors. The client can choose the sensor, computer, or camera they need. We can integrate everything very quickly,” said the researcher. And that’s not all! The team has also integrated a secure flight controller, a computer, and a communication module—mesh communication—into the platform under development. “This is critical in these applications, and practically no other solution of this scale exists on the market. Furthermore, the product we are developing has been designed as a multi-robot system from the very beginning. It is precisely the collaboration of teams of flying robots that is key for complex missions in vast areas, where it is important to perform tasks quickly and where competing solutions practically do not support this,” emphasized Prof. Saska.
The team also regularly tests the drones in the field—above open-pit mines in northern Czechia. “The advantage of testing in these areas is that they offer a truly vast scale. And that’s where the benefits of multi-robot systems really come into play. The mines span tens of kilometers and feature a diverse environment. There are tunnels, pits, dust, and often poor GPS reception because the drones fly close to the rocks. So, for us, it’s an interesting and highly challenging environment in which to fly drones,” added Prof. Martin Saska.
Drones in swarms as the perfect solution for challenging terrain
The second research project presented, which focuses on developing drones capable of flying in dynamic swarms while responding very quickly and flexibly to changes, began this past January. Its objectives also overlap with critical infrastructure monitoring, surveys of complex terrain, and deployment in crisis situations. Fast swarm flight is important in applications where drones must fly close to one another and react to each other. In addition to numerous civilian applications and basic research inspired by the natural world, the FEL CTU team is integrating requirements related to national security and defense—including the aforementioned resistance to GPS outages and communication failures—into swarm research as part of the TAČR PRODEF projects.
To achieve this globally unique goal in their research, the team draws on information from nature. “Specifically, we are currently focusing on the study of birds. We worked on this with Dr. Martin Šálek from the Institute of Vertebrate Biology at the Czech Academy of Sciences, who helped us capture unique and highly dynamic data on the movement of birds flying in flocks. We then analyzed the data in collaboration with Prof. Iain Couzin’s team, who leads one of Europe’s largest centers dedicated to the study of collective behavior in nature at the University of Konstanz. Specifically, we are working with them to determine exactly how high dynamics function within flocks. How do birds manage to swerve very quickly at high speeds, avoid obstacles, and never collide? What information do they use? We directly translate the findings into artificial intelligence that controls robotic swarms. We’re striving to ensure that drones fly in a group just as reliably, quickly, and dynamically as birds,” described Prof. Saska.
How does the process of transferring data from bird behavior analysis into the “body” of a machine work? “We’re trying to figure out how birds perceive the other individuals in their flock and how they use that information to perform coordinated maneuvers. We use the data we collect to train neural networks and other components of the onboard intelligence used to control flying robots. In addition, we are developing sensors that are similarly fast to those birds use to react to rapid events around them. This technology of very fast cameras is also directly inspired by nature,” the scientist explained.
First Major Success
Despite the fact that the project has been underway for barely three months, the team has already achieved its first significant research breakthrough, according to Prof. Saska. “We hypothesized that birds flying in flocks use more information than just the relative positions of their neighbors. In collective robotics, however, only position was commonly used. Thanks to the aforementioned study of birds, we have now discovered that birds try to predict the future movements of other individuals in the flock. And thanks to this, they can fly close together and very dynamically,” emphasized Prof. Saska. “We’ve already managed to conduct an experiment where a drone also tries to predict what its neighbor will do and, based on that, takes faster action, enabling the drones to fly in formation dynamically and with great precision. The improvement in maneuverability reached up to 60 percent, which is incredible progress—and achieved practically within a single month. I believe that thanks to this close collaboration between biologists, robotics experts, and drone specialists, we will soon achieve truly groundbreaking results in the field of dynamic swarms and can also help us understand the behavior of flocks in nature,” summarized Prof. Saska.
