Teaching a robot dog to navigate the wild
Doctoral researcher Yelyzaveta Pervysheva from SpacEconomy partner Tampere University is working on a fascinating space-related topic that is, at the same time, very much down to earth: a robotic dog.
How could it walk through a snowy forest, climb over rocks, cross muddy terrain—and still know exactly where it is?
Of course, it has its heavenly guardian angels: satellites.
This question of determining location lies at the heart of ROBOSAT, a European research project in which we combine satellite navigation, 3D maps, artificial intelligence, and legged robots to enable autonomous navigation in the most challenging natural environments.
"I focuse in this project on how Global Navigation Satellite Systems (GNSS) can be made more accurate and resilient using real-world data collected by robots", Yelyzaveta writes.
Today, most navigation technologies are designed for cities, cars, and environments with clean satellite signals. But nature is messy. Forests block signals, mountains reflect them, and interference is increasing globally. Yet these are precisely the environments where robots could one day save lives – during wildfires, floods, avalanches, or search-and-rescue missions.
At the center of the project is a four-legged robot equipped with satellite receivers, cameras, 3D laser scanners, and motion sensors.
The robot can traverse forests, rocky landscapes, snow, and uneven terrain while recording synchronized environmental and satellite data. This creates something truly unique: a large-scale dataset that links real natural environments with satellite navigation performance.
Most people assume that satellite positioning always works. In reality, trees, terrain, signal reflections, and interference can significantly degrade accuracy. In urban areas, 3D building models help correct these errors. In natural environments, such detailed models rarely exist.
ROBOSAT fills this gap by combining satellite data with precise 3D environmental information and machine-learning methods that estimate how reliable each position truly is.
Yelyzaveta's research focuses on analyzing raw satellite signals, detecting disturbances, and training algorithms to recognize when navigation becomes unreliable.
The goal is not only greater accuracy, but also safer decision-making for robots, vehicles, and emergency systems.
ROBOSAT is designed as an open research platform. The project will provide open datasets, open software tools, and digital services that allow researchers and companies to use real robot data from natural environments. This supports innovation far beyond the project itself.
"For me, ROBOSAT is about more than technology", says Yelyzaveta.
"It represents a step toward safer rescue missions, improved disaster response, and a future in which robots can support humans in the most challenging places on Earth."