Czech experts develop robots to help search for missing persons

Researchers at the Czech Technical University in Prague are using artificial intelligence to train robots, teaching them to search unfamiliar or hard-to-reach areas. In the future, these intelligent robots could help search for missing persons or victims of accidents.

Scientists at Czech Technical University’s Faculty of Electrical Engineering are working with a number of robots, some of which resemble tiny vehicles while others look like dogs or giant spiders.

Miloš Prágr|Photo: Martin Pařízek, Czech Radio

Doctoral student Miloš Prágr uses a game console controller to drive one of the robots, a black and yellow vehicle on four rubber wheels, out of the lab and into the corridor.

“It's a four-wheeled robot of small to medium size that we use mainly to inspect sites or facilities.  It is equipped with several cameras. The two located at the top are both colour and depth cameras, which allows them to sense the geometry of the surrounding environment. The other set of cameras are simple colour cameras. We use them to detect various objects or people.”

The robot is designed to be able to immediately orient itself in any unfamiliar environment, such as a hallway, a cave or a shaft. To do that, it uses spatial sensors to create a model of the environment’s geometry, explains Mr. Prágr:

Photo: Martin Pařízek, Czech Radio

“Based on this model, we are able to create a map of the environment. We can see, for instance, walls or flat surfaces that we can use to move the robot around. We can see obstacles that we should avoid. In such a situation, the robot is already planning its steps using the algorithms that we develop.”

Jan Faigl, head of the Computational Robotics Laboratory at the university’s Centre for Artificial Intelligence, explains in what way such algorithms can be useful:

“The sensors that measure the robot’s surroundings send electrical signals and we sample them in order to transform them into digital form. In that way we get various data, for instance about distance, such as you know from the internet.”

Jan Faigl|Photo: Martin Pařízek, Czech Radio

Using this data, the scientists can learn that a robot has identified an obstacle in the distance and have information about the specific distance.

However, the measurements are never completely accurate. That's where artificial intelligence comes in, using prior knowledge to assess what the object or obstacle is.

“When it comes to mapping an unknown terrain, when we see something that looks like a rock, we only see its front part. However, based on previous experience, the algorithm calculates that there really is a stone behind what we see.”

Jan Faigl illustrates the technology with an image that a robot takes when scanning its surroundings. The human eye can only see a few pixels in detail. However, based on the large amount of data it has learned, the algorithm is able to determine whether it could be a missing person.

Photo: Martin Pařízek, Czech Radio