Autonomous valve turning with an optimization-based whole-body kinematic control algorithm

Abstract

This thesis presents an optimization-based whole-body kinematic control algorithm that is val idated on an autonomous valve-turning application with the dual-arm Girona1000 Intervention Autonomous Underwater Vehicle (I-AUV). The proposed approach leverages Control Lyapunov Functions (CLFs) for equality tasks and Control Barrier Functions (CBFs) for set-based tasks to formulate hierarchical quadratic programming (QP) problems that ensure system stability and safety. Key contributions include the development of a unified framework for simultaneous handling of various kinematic tasks, integration of tasks such as Cartesian position control, orientation control, joint limits, and self-collision avoidance, and the extension of a prioritized multi-task control strategy with a new ”connecting slack constraints” method. The thesis also proposes extensions to the kinematic formulation to allow integration with inherently dynamic systems and presents a modular optimization-based kinematic control software stack developed in ROS/Python that integrates all the aforementioned developments. Extensive simulations give a comparative overview of the performance of the proposed con trol framework relative to traditional task-priority methods and other soft-priority optimization based approaches. Furthermore, an in-depth analysis of safety certification is covered, detailing the shortcomings of kinematic-based safety constraints in guaranteeing the safety of dynamic systems. Finally, the autonomous valve-turning experiment showcases the I-AUV successfully using the optimization-based controller to complete the mission while satisfying task priorities and safety constraints. The results demonstrate the potential of CLF-CBF-based QP control for enhancing the autonomy, safety, and operational capabilities of redundant intervention robots, such as underwater robotic systems, tasked with complex manipulation.

9