Christian Brommer, Alessandro Fornasier, Jan Steinbrener, and Stephan Weiss from the Networked Systems Control Group have developed a new method for robot sensor integration. The related results have been published in the IEEE Transactions on Robotics.

Modern robot systems, such as drones and autonomous vehicles, rely on various sensors including cameras, accelerometers, GPS modules, and others. Previously, the correct integration of these sensors required specialized knowledge and time-consuming calibration. The new method developed by the research team allows robots to automatically identify newly added sensors by type, accurately estimate their position and orientation, and seamlessly integrate them into the existing navigation system.

Brommer explained that there is now no need to know the type of sensor being used. Whether it is a GPS, magnetometer/compass, or speedometer, the data can be directly fed to the algorithm, which automatically identifies the sensor model. However, researchers still need to perform some operations to achieve recognition, such as holding the device in hand in the laboratory, or operating it while flying a quadcopter or driving a car. These operations are also demonstrated in the paper.

Data from the open-source project platform GitHub shows that more than 14,000 developers have used search terms related to "sensor model integration," highlighting the necessity of this new method. Brommer stated that this work aims to make the integration of sensors with positioning solutions such as filters easier, faster, and more robust.