A Pose Estimation Algorithm for Agricultural Mobile Robots using an RGB-D camera

Abstract

This work focuses on the development of a novel algorithm for autonomous navigation for wheeled mobile robots in non-structured environments, such as agricultural fields. The main task of the robot is to self-locate by resorting on a cost-effective multi-sensor system including an IMU, a GNSS receiver and an RGB-D camera. Specifically, an Information Filter is proposed to overcome the issues related to the measurement uncertainties of the single sensor. The strategy is developed under the ROS framework and validated in a simulated environment built in Gazebo. Simulation results show an effective performance improvement, encouraging a field validation of the proposed methodology.