Self-calibration of a mobile manipulator using structured light

Rising automation requirements in manufacturing lead to an increasing demand for robot self-calibration. Self-calibration becomes a challenging task for mobile robots since the environment is dynamic, at least from the perspective of moving robots. This paper proposes a new self-calibration method by tracking the end-effector with the help of a head-mounted projector (see Fig. 1). Pseudorandom coded light is projected into the environment and single images are decoded. The pattern consists of checkerboard-like corner-primitives, which are detected by a camera-pair, mounted on the robot base. Triangulation yields the primitives' positions. The projector can be described as an inverse camera model, so its pose is determinable by taking the primitive points as reference. The extrinsic hand-projector and camera-robot transformations are given by the commonly known formula AX = ZB. Further optimization and the incorporation of the joint parameters allow the calibration of the manipulator. Self-calibration means that no sort of calibration object like a checkerboard is needed. Since the decoding is applied to single images, the environment is permitted to change, and the robot is allowed to move around freely during calibration. Experimental results of this method show a submillimeter accuracy of the proposed projector tracking as well as improvements of the robot's accuracy in its typical workspace.