Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/158088
Title: LEARNING ACTIONS FROM DEMONSTRATIONS FOR MANIPULATION TASK PLANNING
Authors: YUAN JINQIANG
Keywords: Learning from Demonstrations, Combined Task and Motion Planning, Manipulationg Planning, Robotic Learning, Dynamic Movement Primitives, Task Planning
Issue Date: 24-Jan-2019
Citation: YUAN JINQIANG (2019-01-24). LEARNING ACTIONS FROM DEMONSTRATIONS FOR MANIPULATION TASK PLANNING. ScholarBank@NUS Repository.
Abstract: This thesis presented the development of framework for a robotic system that allows non-expert users to operate and perform complex manipulation tasks. The study includes a new method to represent and learn the geometric constraints in manipulation actions, and a new planner based on Combined Task and Motion Planning (CTAMP) architecture. The main contributions of this thesis are: (1) a two-stage learning from demonstration technique for robots: a. learning actions from demonstrations, and b. using the learned actions for planning, (2) a general model to represent the geometric constraints of actions and a method to learn the geometric constraints from demonstrations, and (3) a general model of actions that can be used by the CTAMP framework to plan for manipulation tasks. The algorithms are tested and evaluated using the KUKA iiwa robot with items in daily living. The experimental results have shown that the robot can efficiently learn a wide variety of manipulation actions. Additionally, the robot can generate a plan that satisfies robot kinematics and geometric constraints using the learned actions.
URI: https://scholarbank.nus.edu.sg/handle/10635/158088
Appears in Collections:Ph.D Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
YuanJQ.pdf12.05 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

18
checked on Jan 16, 2020

Download(s)

9
checked on Jan 16, 2020

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.