Hybrid Navigation System for Lely Mixing and Feeding Robot

Abstract

To date, autonomous robots have been widely used and appear huge advantages in dairy farming that is a labor-intensive industry. Vector is an automatic feeding system developed by Lely to feed cows accurately and flexibly with minimum labor requirements in dairy farms. As the autonomous mobile robot of the Vector, the Mixing and Feeding Robot (MFR) is able to automatically distribute feed for cows based on a reactive navigation system. However, since the reactive approach only responds to stimuli at the moment, MFR is susceptible to errors in the barn environment like the temporary loss of valid ultrasonic or inductive sensing; without following a fence or a metal stripe, MFR is quite limited in its ability to navigate. In this project, a hybrid navigation system is proposed combining the reactive navigation currently used on MFR and the map-based navigation consisting of mapping, localization and path planning. Due to MFR's incapability to build a map of the environment with current sensors, a laser scanner is added for map building. Based on Adaptive Monte Carlo Localization (AMCL), the Striped-based Adaptive Monte Carlo Localization algorithm (SAMCL) is proposed for MFR to localize itself based on the map. Moreover, a control strategy is developed to switch between the reactive navigation and the map-based navigation accordingly. To evaluate the performance, the hybrid navigation system is tested in the simulation model built by the robotics simulator V-REP and it is concluded that the proposed navigation system improves MFR's error-tolerant capacity and navigation performance.