CS Events

Qualifying Exam

Failure Explanation Planning for Robotic Manipulation: Planning with Discrete Uncertainty and Many Objects

 

Download as iCal file

Friday, September 18, 2020, 02:00pm - 04:00pm

 

Speaker: 

Rui Wang

Location : Remote via Webex

Committee

Prof. Kostas Bekris (Advisor)

Prof. Jingjin Yu

Prof. Fred Roberts

Prof. Sepehr Assadi

Event Type: Qualifying Exam

Abstract: Traditional task and motion planning problems in robot manipulation focus on finding valid plans for robot arms to fulfill different tasks. Collision-free plans, however, are not often immediately available in highly-challenging manipulation tasks. The first part of the talk will focus on picking an item in the presence of obstacles, while taking into account the uncertainty of the perception process due to occlusions and partial views. The focus is on the case that no obvious collision-free solution can be found. The objective is to minimize the probability of collision with uncertain obstacles and to maximize the probability of picking the target object given a discrete set of pose hypotheses obtained from perception. It follows the concept of conformant probabilistic planning (CPP) and is modelled as a stochastic variant of the Minimum Constraint Removal (MCR) problem so as to achieve the desired objective. This approach is shown experimentally to be effective and is a promising direction for closing the gap between sensing uncertainty and motion planning in manipulation. The second part of the talk will cover the related topic of multi-object rearrangement. Rearrangement can be used either to move blocking obstacles before manipulating the target object when picking in clutter or it can be the task itself in applications, such as logistics, when multiple objects need to be presented to a consumer in a specific manner. Failure explanation planning is also suitable for solving such multi-object task planning problems, especially for non-monotone problem instances, where an object has to be first moved to an intermediate location before being moved to its target pose so that the problem is solvable. The talk covers the existing exploration of this problem in collaboration with colleagues for the case where the rearrangement process has to respect object-to-object interactions but no manipulator-to-object interactions.

 

Meeting Link: https://rutgers.webex.com/rutgers/j.php?MTID=m83a8049185f5cc3344dcd929867b9571

Meeting Number: 120 115 9525
Password: 1234

Other Notes:
Join by video system
Dial This email address is being protected from spambots. You need JavaScript enabled to view it.
You can also dial 173.243.2.68 and enter your meeting number.
Join by phone
+1-650-429-3300 USA Toll
Access code: 120 115 9525