Extrinsic Multi Sensor Calibration under Uncertainties

Highly accurate extrinsic sensor calibration is crucial for environment perception of robots as it allows to fuse information from different sensors. On todays robotic platforms, e.g. autonomous cars, a variety of different sensors with different measurement characteristics is used. Current calibration approaches ignore the individual sensor characteristics. In this work, we derive an approach for optimal calibration in a probabilistic sense under consideration of these characteristics. Our method can be used with any type of sensors e.g. camera, 3D LiDAR, line scanner and radar. We show in simulation and on real data that our approach significantly outperforms state-of-the-art approaches.