SPRIND 2.0 Fully Autonomous Flight

A team combining researchers from the Multirobot Systems Group at CTU’s Faculty of Electrical Engineering and the Czech technology company Fly4Future achieved outstanding results in the German SPRIND Fully Autonomous Flight 2.0 competition. The team demonstrated an autonomous drone system that can interpret instructions in natural language, plan its own mission, and operate without GPS or manual control.

The F4F & CTU SkyGraph team competed against six other European technology teams in the 14-month competition organised by SPRIND, the German Federal Agency for Breakthrough Innovation. During the final tests, held from 24 to 27 August 2026 at Fliegerhorst Erding airbase in Bavaria, the teams were challenged to demonstrate autonomous flight systems capable of performing tasks inspired by real-world applications, including package delivery, search and rescue, and security monitoring.

F4F & CTU SkyGraph achieved the best result in the challenge focused on locating and assisting an injured person. During the final demonstration for investors and representatives of European security and defence organisations, the Czech team also demonstrated the most consistent performance in this task, completing it repeatedly and without errors.

From Natural-Language Instructions to Autonomous Missions

The key innovation of the system lies in its ability to translate an ordinary human instruction into an executable flight plan. Instead of following a predefined route or requiring a specialist to program each mission, the drone interprets the task, builds an understanding of its surroundings, and independently determines how to achieve the requested objective.

The system was tested in four scenarios representing potential applications in delivery services, emergency response, and security operations. Tasks included delivering a package to a specific location, identifying a customer waving from a window, locating an injured person and bringing them a medical kit, and tracking a person moving through a monitored area.

Throughout the tests, the drone had to operate without GPS and complete its missions autonomously after take-off. These requirements made the competition a demanding demonstration of the system’s ability to perceive and interpret unfamiliar environments, make decisions, and adapt its behaviour to changing circumstances.

Understanding the Environment Without GPS

The system combines four cameras, a 3D LiDAR sensor, and an onboard computer to localise the drone without GPS, map its surroundings, and identify objects relevant to the mission.

Detected buildings, people, vehicles, and other objects are incorporated into a continuously updated spatial map. A graph-based representation captures not only individual objects but also their relationships—for example, whether a bicycle is in front of a particular building or where a person is located in relation to a window.

This representation allows the system to interpret more complex instructions and plan its next actions based on the relationships between objects in the environment. By connecting natural-language instructions, environmental perception, mapping, and autonomous planning, the technology aims to make aerial robots more accessible to users without specialised robotics or programming expertise.

From Research to Real-World Applications

The results highlight the potential of combining advanced robotics research with practical autonomous systems. The technology could support search-and-rescue operations, help firefighters locate people and monitor fire hotspots, assist security teams in monitoring large facilities, and enable industrial inspections of hard-to-reach areas. Other potential applications include package delivery, precision agriculture, environmental protection, and mapping areas affected by emergencies.

The approach could also reduce the operational complexity of deploying autonomous drones. Rather than requiring a separate platform and control software for every task, a more general-purpose system could adapt to different missions based on instructions from the end user.

“The most important outcome of the competition is not just the result itself, but the confirmation that a research concept is becoming a generally applicable technology,” says Martin Saska, head of the F4F & CTU SkyGraph team, Professor at CTU’s Faculty of Electrical Engineering, and founder of Fly4Future. “Our goal is to allow users to simply describe the outcome they need and let the system determine how to achieve it.”

The competition’s results are already being integrated into Fly4Future’s robotic platforms as the company prepares the technology for commercial applications.

Links:

Autonomní dron vyvinutý Fly4Future a ČVUT může pomáhat hasičům i záchranářům