Algorithm automatically recognizes sensors and their mathematical modeling

Modern robotic systems—in drones or autonomous vehicles, for example—use a variety of sensors, ranging from cameras and accelerometers to GPS modules. To date, their correct integration has required expert knowledge and time-consuming calibration.

Christian Brommer, Alessandro Fornasier, Jan Steinbrener and Stephan Weiss, all members of the Control of Networked Systems research group at the time their research was conducted, have developed a new integration method and published it in IEEE Transactions on Robotics. It allows robots to automatically identify a newly added sensor by type, estimate its position and orientation, and correctly integrate it into the existing navigation system.

According to Brommer, it is no longer necessary to know which sensor is being used. Whether GPS, magnetometer/compass or speedometer, the data can simply be passed on to the algorithm and the sensor model is automatically recognized.

However, the researchers still need some movement for recognition. “This can be managed, for example, by holding the device in your hand in a laboratory or, as we demonstrate in the paper, during flight with a quadcopter or while driving a car,” says Brommer.

The need for this method is undeniable: GitHub, a platform for open-source projects, has registered more than 14,000 requests from developers using the keywords “sensor model integration.”

“Our work aims to make the integration of sensors into localization solutions such as filters easier, faster and more robust,” says Brommer.

Provided by
Universität Klagenfurt