Robust Grasp Planning from Depth Image for Cluttered Scenes
Översikt
- Datum:Startar 10 juni 2024, 15:30Slutar 10 juni 2024, 17:00
- Plats:Online participation only via zoom
- Språk:Svenska och Engelska
Examiner: Yasemin Bekiroglu
Opponents:
Weilong Chen and Fangdelong Su
Abstract:
Grasping is a fundamental skill in robotic applications in household, warehouse, healthcare, and industrial scenarios, which involves constructing stable configurations between the robot's end-effector and the objects of interest. Antipodal grasping, characterized by its simplicity and efficiency in handling a wide array of object geometries, proves indispensable for applications demanding precise, firm grips. Suction grasping emerges as a versatile solution for objects with challenging surfaces and irregular shapes, providing a non-invasive and adaptable approach to manipulation. In this work, we delve into hybrid-modality robotic grasp planning, investigating antipodal grasping and suction grasp planning on the pixel level from solely depth image as visual input. By predicting the grasp pose on every pixel from visual input, we can produce numerous dense and applicable grasp poses for downstream tasks. Real-world experiment results are presented to demonstrate the robustness and generalization ability of our hybrid grasp planning framework.
Welcome!
Yiting and Yasemin
Online participation only via Zoom:
Join from PC, Mac, Linux, iOS or Android: https://chalmers.zoom.us/j/8362581794
Meeting ID: 836 258 1794
International numbers available: https://chalmers.zoom.us/u/ctXzaGuKj
Meeting ID: 836 258 1794
