Exciting Trajectories for Extrinsic Calibration of Mobile Robots with Cameras

The paper presents a method to precompute the optimal (so-called exciting) trajectories for the extrinsic calibration of a mobile robot equipped with the wheel odometry and one or more cameras. Considering the fact that the calibration is formulated as a non-linear least-square problem, the method is based on the analysis of the cost function properties in the neighborhood of the solution. By maximizing the determinant of the Hessian matrix, one makes the problem better defined and improves its robustness with respect to the measurement noise. Another convenience of the method is the possibility to reinforce additional constraints, like the visibility of a calibration object and the trajectory feasibility. The source code of the application is publicly available as a part of visgeom project.