Deploying robots for unstructured industrial quality control remains a bottleneck due to unexpected environmental variations. This particularly occurs when manipulating flexible materials like textiles, where soft, deformable structures and surface reflectance make physical interaction challenging. This doctoral project focuses on robot learning for vision-guided manipulation within industrial settings. The research aims to explore robot learning models to bridge the gap between high-level perception and low-level physical control. Because textiles are soft and change shape, robot learning models can help translate visual feedback and multi-modal instructions directly into adaptive motor skills. Rather than relying on rigid, pre-programmed trajectories, the project will investigate how one or more robotic arms can learn to autonomously adapt and perform accurate, localized measurements on garments guided dynamically by a vision system. Key areas of investigation include representation learning for deformable objects, reinforcement learning for path planning, and cross-domain alignment. The ultimate goal is to create an imitation-based robot learning framework that enables physical platforms to transition from visual detection to physical inspection in real-world manufacturing.

This PhD opportunity is sponsored by Alsco.