2022 Volume 58 Issue 4 Pages 213-220
Recently, the use of small mobile robots in smart agriculture has been explored. Small robots have the advantage of being less invasive to the environment and crops, but also have the problem of being prone to stranding due to collisions with the geological features. Many studies have been conducted on the collision between robots and obstacles. However, most of them focus on avoiding collision or stopping the robot quickly after collision in order to protect robots and workpieces. On the other hand, since obstacles to field robots are assumed to be soft and deformable obstacles such as soil and grass, there are other options for dealing with obstacles besides the above: push off. In order to take such an action, one has to determine the scale of the obstacle and decide whether to push or pull. However, there are not many studies that use the force received from a deforming obstacle as an indicator of behavioral decisions. In response to the above problem, this study proposes a method for estimating the time and magnitude of a robot's collision with an obstacle and determining the possibility of stranding due to continued action. The target robot is a small weeding robot for paddy fields that has balance floats in front and behind its body and runs on two wheels. The proposed method models the acceleration by a differential equation with an unknown collision, and the collision term is estimated by an extended Kalman filter. The effectiveness of the proposed method is confirmed by numerical examples and experiments.
