Andreas Michaels, Sebastian Haug, Amos Albert, and Slawomir Grzonka
Vision-Based Manipulation for Weed Control with an Autonomous Field Robot
Abstract:
Organic farming includes many labor-intensive manual tasks, in particular the need for treatment of individual plants. Some examples for such tasks are selective harvesting and mechanical weed control. The publicly funded project Remote Farming investigates how such tasks may be accomplished by an autonomous robot system, particularly for the use case of mechanical weed regulation in carrot row cultivations. The main topics of the project are plant classification, high precision and high speed manipulation as well as shared autonomy concepts. This paper deals with the vision-based regulation of weed, which is growing close-to-crop. For that purpose a manipulator with parallel kinematic structure is mounted on the mobile robot "BoniRob" as shown in Figure 1. The major challenges for the manipulation are the harsh environment, as well as the requirements arising from speed and accuracy perspective. The parallel kinematic structure manipulator is equipped with two cameras and a specifically developed actuator. One of the cameras is mounted on the robot platform whereas the other is attached to the manipulator. The latter forms an eye-in-hand system for visual servoing. Precision and speed demands must be ensured, even when the field robot moves continuously. Therefore we have to compensate the flexible mounting and vibrations, which are caused by the ego motion of the system on uneven ground and the forces exerted by the manipulator to the mobile platform.
Bibtex:
@article{michaels2013vision,
title={Vision-Based Manipulation for Weed Control with an Autonomous Field Robot},
author={Michaels, A and Haug, S and Albert, A and Grzonka, S},
journal={Proc. Agricultural Engieering, Landtechnik AgEng},
pages={289--294},
year={2013}
}
Paper (pdf file):
full paper