A Dynamic Motion Planning System of Fabric

Fabric dynamic motion planning is challenging. Fabric motion and its are affected by many hard-to-estimate factors in the environment, resulting in fabric dynamic motion that cannot be modeled accurately. In the proposed system, we roughly approximate the fabric state based on the segmented fabric point cloud. The fabric motion is then represented by a time series of the approximated state. Then a neural new type of neural network system is proposed to model the fabric motion using the approximated state. A simple trajectory planning method is designed to plan a robot endpoint trajectory that connects the start point and the goal point without collision between the fabric attached to the robot end-effector and the obstacle. The effectiveness of the proposed system is demonstrated in collision avoidance experiments under different obstacle settings.