Visual-RRT*: Generating Asymptotically Optimal Trajectories With Vision-Based Controllers

Abstract

In this article, we present a planning method for robotic systems that, by construction, guarantees the planned trajectories to be executable in closed-loop with a visual servoing controller, in addition to assuring their asymptotic optimality in the state space. The design of this method requires one to obtain a locally optimal steering method, locally stable closed-loop controllers used to track the optimal trajectories, and a mapping between the motion of the robot in the state space and the corresponding motion of the visual features in the image space. As a result, the proposed approach allows the system to navigate in a closed-loop manner among nonconvex obstacles, following asymptotically optimal trajectories. The method is validated using two systems: a nonholonomic robot and a robotic arm. Real-world experiments are carried out for the first system.