Mobile robotic applications need precise information about the geometric position of the individual sensors on the platform. This information is given by the extrinsic calibration parameters which define how the sensor is rotated and translated with respect to a fixed reference coordinate system. Erroneous calibration parameters have a negative impact on typical robotic estimation tasks, e.g. SLAM. In this work we propose a new method for a continuous estimation of the calibration parameters during operation of the robot. The parameter estimation is based on the matching of point clouds which are acquired by the sensors from multiple static viewpoints. Consequently, our method does not need any special calibration targets and is applicable to any sensor whose measurements can be converted to point clouds. We demonstrate the suitability of our method by calibrating a multi-sensor system composed by 2 lidar sensors, 3 cameras, and an imaging radar sensor.