Next-level drone tech: Hungarian lab masters autonomous precision delivery

When we hear about delivery drones, many of us imagine futuristic scenes: hovering devices dropping off orders directly on balconies or in front of garden gates. However, reality is far more complex. The greatest challenge lies in enabling these devices to navigate often unpredictable urban environments autonomously, safely, and reliably. Researchers at the HUN-REN SZTAKI Systems and Control Lab (SCL) have achieved globally unique results in this area by developing intelligent control solutions that allow drones to react in real time to environmental changes – while consuming minimal energy.

A drone that sees, thinks, and decides

For a drone to fly safely, it must know where it is, what’s around it, and which direction it should take. Pre-programmed flight paths aren’t enough – rapid, improvised adaptation is required. A passing bird, a crane erected for construction, or even a sudden gust of wind can force a change in the planned route. Autonomous drones can rely on a range of sensors for orientation: GPS, accelerometers and gyroscopes, cameras, and LIDAR – which operates similarly to radar, but uses lasers instead of radio waves to map out objects and obstacles in the environment.

This brings us to one of the fundamental dilemmas of the technology: we can equip drones with the best possible sensors, but these are heavy, consume a lot of energy, and complicate the system. Yet a drone can only carry limited weight, and its battery capacity is finite. That’s why smart, efficient control solutions are needed – the goal isn’t to detect and record every bit of data, but to perceive and process just enough to make fast, safe decisions.

Researchers at HUN-REN SZTAKI SCL are developing software systems that can make quick and reliable decisions based on the available data. “These systems not only plan ahead, but constantly re-plan the drone’s movements in response to even the slightest unexpected change,” says Tamás Péni, researcher at SCL. “It’s not enough to know where the obstacle is at a given moment – the drone must also predict how that object might move and adjust its own motion accordingly, maintaining appropriate safety distances.”

To achieve this, flight path planning occurs on two levels. The first is the global level, which functions much like using a GPS to drive toward a destination. Based on available maps and data, the drone preplans its route. However, when an unexpected situation arises – say, an obstacle appears in its path – the drone switches to local-level planning, adjusting its flight path or even rethinking its entire route. If multiple drones are in the air at once, they can also communicate with one another. For example, if one detects a new obstacle – like a crane that wasn’t marked on the map – it can immediately share this information with the others. This way, all drones work from a single, real-time “map” and can safely reach their destinations.