Real-time, Robust 6D Pose Estimation and Tracking for Dexterous Manipulation
Friday, May 08, 2020, 03:00pm - 04:30pm
Speaker: Bowen Wen
Location : Remote via Webex
Advisor: Kostas Berkis
Committee Members: Abdeslam Boularias, Ahmed Elgammal, Desheng Zhang
Event Type: Qualifying Exam
Abstract: Many manipulation tasks, such as picking, placement or within-hand manipulation, require the object’s pose relative to the manipulating agent, robot or human. At the same time, these tasks frequently involve significant occlusions, which complicate the estimation and tracking process. This work presents a framework based on RGB-D data. It aims towards robust pose estimation under severe occlusions, such as those arising due to a hand occluding a manipulated object. It also aims for short response times so as to allow for responsive decision making in highly dynamic setups. The proposed framework leverages the complementary attributes of deep learning and 3D geometric reasoning, so as to achieve high accuracy as well as generalization to different robotic manipulation scenarios. Additionally, only synthetic data are required for training, making the proposed approach applicable to new tasks, circumventing the overhead of collecting labeled real world data. Extensive evaluation on multiple real world benchmarks demonstrates superior performance when compared with state-of-the-art approaches.