Implementation of a grasp planning pipeline for the humanoid robot neoDavid

Kurzfassung

Robotic grasping is a fundamental challenge in autonomous manipulation, requiring precise coordination between perception, motion planning, and control. This thesis presents a comprehensive grasp planning pipeline that integrates a variable stiffness analysis to enhance grasp stability and adaptability. The proposed framework addresses key limitations in traditional grasp planning by incorporating stiffness modulation, optimizing grasp execution for a wide range of objects and grasping scenarios. The grasp planning pipeline consists of several core components: object detection, grasp pose generation, collision checking, inverse kinematics computation, and base positioning for mobile manipulators. To optimize the grasping strategy, a novel quality metric is introduced, leveraging the stiffness matrix in task space to evaluate grasp feasibility while maintaining a balance between robustness and compliance. Additionally, the pipeline is fine-tuned using a statistical optimization approach based on the Taguchi method, significantly reducing the number of required experimental trials while systematically identifying optimal grasp parameters. This research contributes to the advancement of robotic grasping by bridging the gap between theoretical modeling and practical implementation, providing a flexible and robust grasp planning framework that can be extended to complex manipulation tasks in dynamic environments. The effectiveness of the proposed framework is validated through both simulation and real-world experiments, using a robotic system equipped with variable stiffness actuation.