Path Planning for Autonomous Work for Agricultural Mobile Robots with Wheels

Abstract

In this paper, we attempted generating a moveable path from the start to goal positions with mobile robots with wheels. Because agricultural mobile robots with wheels cannot move onto a step due to mechanical factors, we have devised a path planning method which considers the gradient on the ground. Q-learning is one of the methods which is often used for path planning. Using Q-learning, we made a robot learn the optimum path with the smallest gradient on the ground. We performed an experiment using a real robot in an actual environment with steps and slopes, and investigated the effectiveness of this method. The effectiveness of this method was verified from the experiment results.