Vision-based minimum-time planning of mobile robots with kinematic and visibility constraints

Abstract

Abstract This paper proposes a vision-based minimum-time trajectory planning method for mobile robots, which takes into account kinematic constraints for linear/angular velocities and accelerations, as well as the visibility constraint. Different from existing methods, by means of homography-based pose estimation, the vision-based trajectory planning is formulated as a constrained optimal control problem in the scaled Euclidean space, which is solved by using the Gauss Pseudospectral Method (GPM). Specifically, the homography matrix is estimated and then decomposed to obtain the relative rotation angle and the scaled translation between the current pose and the desired one, which are expressed in the scaled Euclidean space. Then, kinematic constraints are taken into account in this space, while the visibility constraint is formulated by mapping the Euclidean homography matrix to the image space. To our best of knowledge, it is the first reported approach to solve the vision-based minimum-time trajectory planning problem for wheeled mobile robots, which can help improve the working efficiency in realistic visual servoing systems. Extensive simulation and experimental results with comparison to other related methods are presented to demonstrate the effectiveness of the proposed approach.