拡張カルマンフィルタを用いた水田用小型ロボットの座礁検出に関する研究

抄録

In this paper, the authors propose a method to detect collision of the robot in paddy fields to avoid the stranding of small robot in shallow water. The proposed method is to model the acceleration of the robot by the differential equation, including the unknown collision, and estimate the time and amplitude of the collision. The extended Kalman filter is used for estimation. Two types of models in which the unknown collision is modeled by the acceleration and velocity are discussed. As the result, we understand that the time of the collision can be estimated from the time of the estimated value of the velocity-dependent collision term becomes maximum values. Furthermore, the result shows that we can decide the robot becomes stranding based on the amplitude of the estimate of the velocity-dependent collision term because the estimated value becomes large when the height of obstacle becomes large.