Uncertainty-Aware Manipulation Planning Using Gravity and Environment Geometry

Factory automation robot systems often depend on specially-made jigs that precisely position each part, which increases the system's cost and limits flexibility. We propose a method to determine 3D pose of an object with high precision confidence, using only parallel robotic grippers no parts-specific jigs. Our automatically generates sequence actions ensures real-world physical matches assumed sub-mm precision. Furthermore, we use “extrinsic” actions, gravity, environment gripper geometry significantly reduce or even eliminate uncertainty about object's pose. show in simulated real-robot experiments our outperforms previous work, at success rates over 95%.