Title:
CLEANING ROBOT OPTIMAL TARGET PATH PLANNING METHOD BASED ON MODEL LEARNING
Document Type and Number:
WIPO Patent Application WO/2017/161632
Kind Code:
A1
Abstract:
A cleaning robot optimal target path planning method based on model learning. In the method, a robot can be driven to preferentially treat places where there is possibly the most rubbish; with a reinforcement learning architecture and a Dyna-H algorithm as a basis, an exploration mechanism in an R-MAX algorithm is used; in an inter-state distance metric method, a Euclidean distance metric method in the Dyna-H is improved using a self-simulation metric, so that the learning efficiency of a model is improved.
Inventors:
LIU QUAN (CN)
ZHOU YICHENG (CN)
ZHU FEI (CN)
ZHOU YICHENG (CN)
ZHU FEI (CN)
Application Number:
PCT/CN2016/080332
Publication Date:
September 28, 2017
Filing Date:
April 27, 2016
Export Citation:
Assignee:
ZHANGJIAGANG INST IND TECHNOLOGIES SOOCHOW UNIV (CN)
International Classes:
G05D1/02
Foreign References:
| CN101714000A | 2010-05-26 | |||
| CN105320133A | 2016-02-10 | |||
| CN102866706A | 2013-01-09 | |||
| JP2006285547A | 2006-10-19 | |||
| JP2003241836A | 2003-08-29 |
Attorney, Agent or Firm:
NEW SUZHOU PATENT AGENT CO., LTD (CN)
Download PDF: