Navigation and Control for an Autonomous Robotic Fruit Harvesting System

The use of robotics in industrial environments has increased dramatically over the past decade. Advances in automated systems have given rise to a multitude of applications, from automated robotic security to drone delivery services. With research and develop- ment motivated by both economic incentives and accentuated by national and global issues, such as labour shortage, climate change and continuously increasing demand, the agricul- tural industry has recently found interest in automated robotic technologies. Robotics are an ideal solution to crop harvesting. This often repetitive, time consuming, labour intensive and costly task is perfectly suited to automation with a capable system. Automated robotics would enable 24 hour operation, increasing efficiency and improving resource management, whilst significantly reducing many associated costs. This task from a human’s perspective is straightforward, yet when thought of in a robotic system’s context requires many inter- linked components including navigation and control methods and computer vision systems. The research of this thesis concentrates on the practical implementation of a navigation and control software system for a strawberry picking task. It includes a novel software archi- tecture, a navigation and control method and initial work into a low cost implementation of an automated robotic fruit harvesting system. The software architecture is a configurable, built on top of the Robot Operating System (ROS). It contains perception and action systems to coordinate dual arms with a mobile base platform to pick strawberries. This architecture is able to host laser SLAM for building farm maps for autonomous navigation. The design and implementation of the software architecture, and the evaluation of the laser SLAM map- ping building are included. The autonomous navigation and control software for the system incorporates the provision for both robotic arms and mobile base. The motion strategy of the mobile base and dual arms for picking strawberries is proposed, and the analysis of its performance in both a laboratory setting and an agricultural environment is included. Finally the thesis contributes the implementation of a single board computer for the software architecture. The focus of this portable implementation is the performance analysis of instance segmentation and SLAM systems based on an RGBD camera. The experimentation to deter- mine the viability of implementing the system overall on to a single board computer is also introduced.