Route Planning for a Tractor in an Agriculture Field with Obstacles

Abstract

In this article, we propose a trajectory planning method for a tractor in a rural environment. Our method is based on the simultaneous use of offline and real-time acquired data. The offline data is the geographical map of the agricultural field, which is assumed to be known, as well as vehicle driving data prerecorded by several experienced drivers. The depth video and the location of the vehicle are acquired in real time and used with existing data for the detection and avoidance of obstacles. The planning is seen as a constrained optimization problem. The constraints are related to the presence of obstacles which are considered as spatiotemporal events recognized through their geometric structure represented in the depth frames. We show that the proposed approach validated by using real data is effective and fulfills the real-time requirement.