The intensive advances in robotics have deeply facilitated the accomplishment of tedious and repetitive tasks in our daily lives. If robots are now well established in the manufacturing industry, thanks to the knowledge of the environment, this is still not fully the case for outdoor applications such as in agriculture, as many parameters are varying (kind of vegetation, perception conditions, wheel–soil interaction, etc.) The use of robots in such a context is nevertheless important since the reduction of environmental impacts requires the use of alternative practices (such as agroecological production or organic production), which require highly accurate work and frequent operations. As a result, the design of robots for agroecology implies notably the availability of highly accurate autonomous navigation processes related to crop and adapting to their variability. This paper proposes several contributions to the problem of crop row tracking using a four-wheel-steering mobile robot, which straddles the crops. It uses a 2D LiDAR allowing the detection of crop rows in 3D thanks to the robot motion. This permits the definition of a reference trajectory that is followed using two different control approaches. The main targeted application is navigation in vineyard fields, to achieve several kinds of operation, such as monitoring, cropping, or accurate spraying. In the first part, a row detection strategy based on a 2D LiDAR inclined in front of the robot to match a predefined shape of the vineyard row in the robot framework is described. The successive detected regions of interest are aggregated along the local robot motion, through the system odometry. This permits the computation of a local trajectory to be followed by a robot. In a second part, a control architecture that allows the control of a four-wheel-steering mobile robot is proposed. Two different strategies are investigated, one is based on a backstepping approach, while the second considers independently the regulation of front and rear steering axle position. The results of these control laws are then compared in an extended simulation framework, using a 3D reconstruction of actual vineyards in different seasons.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.