Drones and AI Are Changing the Rules in Crisis Situations

Unmanned systems, in this case drones (UAS), have enabled us to establish situational awareness in the field within minutes, which ten years ago was not possible without helicopters or physically sending teams to the site, says Željko Riha, Vice President of the Association Unmanned Systems in Crisis Situations. He adds that the application of artificial intelligence in their work is steadily growing, as AI is increasingly used for automated analysis of video and photo material — detecting people, fire hotspots, assessing damage to infrastructure, and classifying types of terrain.

According to him, they train AI models on their own field footage, including data from Croatia’s geographic specificities — dense forests, rocky karst terrain, urban areas — which increases the accuracy of the system. However, he emphasizes that this is still in the early phase, namely the testing stage.

How have modern technologies such as unmanned systems changed the way we respond to major accidents and natural disasters compared to methods from, say, ten years ago? Can you cite a concrete example of a recent intervention (search, flood, fire) where the use of such technology was decisive for the outcome?

Unmanned systems, specifically drones (UAS), have enabled us to establish situational awareness on the ground within minutes, something that a decade ago required helicopters or the physical arrival of response teams. Today, within the first hour, we already have aerial footage, 2D maps, thermal imaging, and hazard data providing critical information to incident commanders.

One example I would highlight is the use of unmanned systems during floods in Croatia. UAS were deployed to monitor critical dike points, identify newly formed water flows, and deliver visual data to commanders within the Civil Protection system. This allowed evacuation and dike defense decisions to be made faster and more accurately.

The key advantages in the initial phase are definitely the speed of acquiring visual data, faster crisis assessment, access to hard-to-reach areas, and minimal risk to human lives in hazardous zones. Real-time data transmission to command and operational centers is also essential.

In many emergency services, unmanned systems are already standard tactical equipment, especially within Civil Protection. However, full integration into SOPs still depends on training and the availability of equipment at local levels. I am confident that soon we will see full integration of unmanned systems such as drones into SOPs.

When we talk about artificial intelligence (AI), in what ways is it applied in analyzing data collected from the field, for example, via drones? Are AI systems used for automatic detection of missing persons, identification of fire hotspots, or assessment of infrastructure damage from above?

The use of artificial intelligence in our case is growing steadily. AI is applied for automated analysis of video and photo material — detecting people, identifying fire hotspots, analyzing damage to infrastructure, and classifying terrain types.

There are cases where AI systems automatically recognize missing persons (e.g., using thermal cameras), detect fire hotspots, and even generate assessments of building stability after earthquakes.

Processing time depends on the complexity of the analysis, but generally ranges from minutes to around ten minutes for image and video material. Final decisions are always made by intervention commanders and expert teams — AI supports, but does not replace human judgment.

We train AI models on our own field footage, incorporating Croatia’s geographic specificities — dense forests, rocky terrain, urban environments — which improves system accuracy.

I would emphasize that all this is still in an early stage, in testing. During real interventions, pilots and observers remain crucial as part of the drone team on the ground. But I am convinced that soon we will have reliable and effective autonomous drones trained with AI models.

How complex are the systems you operate, and what kind of training and certification do your operators undergo to act effectively in stressful and unpredictable conditions? What are the main operational challenges in the field — weather conditions, inaccessible terrain, battery life, or something else? How does cooperation and communication look between the team operating unmanned systems and traditional ground teams such as firefighters, mountain rescue (HGSS), or first aid units?

BSKS was established to bring together professionals and civilians who already fly unmanned systems, whether professionally or as a hobby. It is a group of enthusiasts who want to contribute their knowledge and experience to strengthen society and raise Civil Protection to a higher level.

Our operators within BSKS, in addition to standard training required by the Civil Aviation Agency, participate in joint exercises with the State Intervention Unit of Civil Protection and firefighters. This improves not only their flying skills but also their coordination with other services in the field.

The biggest challenges in the field are weather conditions (wind, rain), inaccessible terrain without suitable take-off/landing points, limited battery life, and radio spectrum saturation (interference).

Cooperation is crucial. On the ground, we act as an extended arm of the incident commander. Data is shared directly with firefighters, Civil Protection, and first aid teams, most often through integrated communication systems or mobile command centers.

What about situations when technology fails?

These situations are not planned, but given the nature of crisis response, we often push drones to their limits, sometimes knowingly, to acquire critical information as quickly as possible. Although we promote safe drone use, we accept the risk of signal loss and train for such scenarios to minimize potential hazards.

Beyond standard visual cameras, what other types of sensors and equipment do you use on unmanned systems to increase efficiency in various scenarios, such as night searches or dense vegetation? To what extent do thermal cameras help locate missing persons, and how reliable are they? Do you use technologies such as LiDAR for creating 3D models of collapsed structures or mapping landslides?

Thermal imaging has proven extremely important during interventions, so we use thermal cameras as well as multispectral sensors. Night missions are often carried out using a combination of thermal imaging and lighting modules.

Thermal cameras are highly reliable for searches, but require experienced operators to interpret data correctly, as environmental factors (e.g., heated rocks) can cause false alarms.

So far, we have not directly used LiDAR technology, but we are testing it for creating precise 3D models of collapsed structures, especially useful after earthquakes or in landslides, where visual detection is not sufficient.

BSKS and our activities are still primarily focused on education, training flights, and testing new capabilities. We are not yet ready to fully engage in advanced drone operations such as package delivery or cargo transport. But we monitor global trends, attend conferences, and study such applications. When we decide to pursue advanced drone operations more seriously, it will mean our members are already responsible participants in air traffic and conduct their missions safely.

Looking into the future, what new technologies and solutions can we expect in crisis management over the next 5 to 10 years? Is there research into advanced concepts such as drone swarms for rapid search of large areas, or fully autonomous systems operating without direct human control? Is there a vision of a unified digital platform that would aggregate data in real time from all sources — drones, satellites, citizens, and ground teams — to create a complete situational picture?

Personally, and this view is shared by colleagues, we expect further development of autonomous systems, advanced AI models for real-time situation analysis, and drones with higher payload capacity for logistics tasks. This is critical to improve efficiency in the field. Autonomy in flight operations is already well developed today. Concepts like drone swarms are still in research and development. In the near future, we see autonomous systems handling initial assessments, with humans stepping in for critical decision-making.

The Republic of Croatia already actively uses the NICS system, which can integrate in real time data from drones, satellites, ground cameras, and even citizen reports. I believe NICS will become even more integrated into our operations.

Key prerequisites for development are clear and flexible legal frameworks, as well as financial investment in equipment and training. Systematic education and training of local services are crucial, along with strengthening IT and AI capacities, particularly through cooperation with academia. We are already working closely with the Faculty of Electrical Engineering and Computing in Zagreb, and that collaboration is excellent.