Adaptive Simulation-Trained Cloth Manipulation Control with Human Guidance for Real-World Robotics Tasks

This study presents a comprehensive system integrating expert-built simulations with human-guided real-world control for robotic tasks. Due to potential spatial constraints in actual factory environments, errors during workspace assembly, and the challenges in measuring anisotropic parameters of deformable fabrics, we aim to adapt robotic arms to real-world conditions through direct human guidance, enabling the transfer from simulation control to real-world operation. Key contributions include (1) a normalizing method to bridge the gap between simulation and reality, (2) the employment of force-position control, and (3) the use of human guidance for model environment updates. Experiments demonstrate the system's effectiveness in bridging the gap between simulation and reality and enhancing the applicability of simulation-trained models to complex, real-world scenarios.