Abstract-We present a novel on-line approach for extrinsic robot-camera calibration, a process often referred to as handeye calibration, that uses object pose estimates from a real-time model-based tracking approach. While off-line calibration has seen much progress recently due to the incorporation of bundle adjustment techniques, on-line calibration still remains a largely open problem. Since we update the calibration in each frame, the improvements can be incorporated immediately in the pose estimation itself to facilitate object tracking. Our method does not require the camera to observe the robot or to have markers at certain fixed locations on the robot. To comply with a limited computational budget, it maintains a fixed size configuration set of samples. This set is updated each frame in order to maximize an observability criterion. We show that a set of size 20 is sufficient in real-world scenarios with static and actuated cameras. With this set size, only 100 microseconds are required to update the calibration in each frame, and we typically achieve accurate robot-camera calibration in 10 to 20 seconds. Together, these characteristics enable the incorporation of calibration in normal task execution.
I. INTRODUCTIONTo enable a robot to interact with objects observed by a camera, and to learn from these interactions, it is invaluable to know the pose of the camera in the robot's reference frame. This allows for the robot to situate perceived objects with respect to itself, as illustrated in Fig. 1. The camera pose can be obtained using a robot-camera calibration procedure, often referred to as hand-eye calibration. Typically such calibration requires a time-intensive procedure in which first a fixed sequence of robot configurations are determined and executed while a video is recorded. This video is then processed off-line with a vision-based pose estimation method in order to detect the pose of a calibration object or markers attached to the robot. Finally a batch optimization is performed to determine the calibration parameters. The hand-eye calibration problem is not simply a pose estimation problem since both the position or pose of the calibration object and the camera pose with respect to the robot are unknown. So as a side product of the calibration, the object's pose or position in world coordinates is obtained as well.In many real-world situations, the camera pose relative to the robot may change over time due to changing temperature, humidity, or due to a contact or collision with the environment. It may also be necessary to quickly add