Collision Detection and Stranding Prediction of Small Robots for Paddy Fields against Soft Obstacles Using Wavelet Transformation

Abstract

This paper proposes a method for detecting collisions with deformable soft obstacles and strandings of small robots before they become inoperable. The proposed method represents the robot's behavior as a mathematical model of acceleration with unknown exogenous inputs which represent collisions and detects the features of the unknown inputs estimated by an extended Kalman filter using wavelet transformation. The effectiveness of the proposed method is verified by water tank experiments, and problems are discussed.