Modeling, identification and control of an omnidirectional wheeled manipulator for intralogistics applications in shared workspaces

Efficient material flow, also known as intralogistics, plays a crucial role in agile production. Despite the potential of robotic solutions, their complete integration into automating intralogistics processes, particularly in collaboration with humans, remains a significant challenge. This is primarily due to the limitations in flexibility, cost-effectiveness, and safety certification of current robots. To address these challenges, this thesis focuses on the development of a model-based, whole-body controller that can handle navigation and manipulation tasks simultaneously while ensuring human safety. To achieve efficient execution of manipulation tasks, it is essential to employ rigid or compliant actions. Therefore, the whole-body controller has been specifically designed with a strong emphasis on dynamics, leveraging the theory of impedance control. By adopting a model-based approach instead of alternatives like optimization-based control, we can ensure safety and reliability, mitigating the risk of failure and improving our understanding of the overall system's functioning. The chosen model-based approach requires deriving an accurate model of the robot. In this thesis, we provide a detailed description of the dynamics and kinematics derivation for an omnidirectional wheeled manipulator. These derivations serve as the foundation for implementing the proposed whole-body controller and integrating a validated safety algorithm that incorporates impact dynamics and injury biomechanics. Additionally, the ability to perform coordinated base-arm motions with dexterity is a desirable characteristic for a mobile manipulator. To demonstrate this capability, we present an example that utilizes artificial potential fields. Ultimately, all these components are seamlessly integrated into a unique framework, which undergoes rigorous testing in MATLAB/Simulink and subsequent simulation in ROS/Gazebo. While the theoretical aspects are applicable to any omnidirectional-wheeled manipulator, the tests are conducted on a digital twin of a real robot currently being utilized in the EU project DARKO.