Visual path following with obstacle avoidance for quadcopters in indoor environments

Abstract

In this paper, we propose a hierarchical visual servo control scheme for following visual paths with quadcopters in indoor environments. The control scheme is able to deal with three tasks simultaneously, considering as order of hierarchy: first, collision avoidance task, second visibility task (keep point features in the camera’s field of view), and finally, visual servoing task. The strategy allows a commitment of active tasks (also the smooth deactivation of tasks) for avoiding conflicts between them. The proposed scheme uses a monocular RGB front-facing camera as single sensor, i.e., only 2D information is required for the control law computation. That information is used for online reactive visual path following, i.e., no 3D reconstruction of the environment is required, and no path or motion planning for obstacle avoidance is performed while the robot follows the visual path. The effectiveness of the proposed control scheme is supported by a stability analysis and experiments with a low-cost commercial quadcopter in scenarios with static obstacles.