Perception-Based High-Precision Robotic Assembly and Packing in the Presence of Uncertainty
Reliable and efficient tight-clearance robotic assembly and packing are critical for real-world applications in manufacturing, logistics, and related industries. However, tight-clearance robotic assembly remains challenging, as small uncertainties in object modeling, pose estimation, or robot motion can lead to failures, and existing solutions are often limited to specific objects or tasks.
This dissertation proposal addresses these challenges by introducing a suite of novel and general approaches. These include a general-purpose method for autonomous robotic assembly of complex, arbitrary parts under six-dimensional uncertainty, a human-robot collaborative framework that combines human dexterity and problem-solving with robotic accuracy and strength, and a shape-invariant perception algorithm for detecting unknown 3D objects via RGB-D sensing, coupled with a 3D bin-packing strategy that optimally arranges and tightly packs items into containers.
Advisor: Professor Jing Xiao
Committee: Professor Loris Fichera, Professor Jane Li, Professor Xiangnan Kong
