Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
LARGE-RANGE MONITORING METHOD BASED ON DEEP WEIGHTED DOUBLE-Q LEARNING AND MONITORING ROBOT
Document Type and Number:
WIPO Patent Application WO/2018/205778
Kind Code:
A1
Abstract:
A large-range monitoring method based on deep weighted double-Q learning. The method involves: firstly, providing a robot (10) having a Q value table comprising a QA table and a QB table; secondly, an unidentified object (30) entering a large-range space (20) so as to trigger the robot (10); then, the robot (10) sensing a current state s, and determining whether the current state s is a target state; if so, the robot (10) reaching the next state and monitoring the unidentified object (30); if not, the robot (10) reaching the next state, and the robot (10) obtaining a reward value according to the next state; and the robot (10) choosing to update a QA value or a QB value equiprobably, and then updating a Q value, until the optimal monitoring policy is obtained through convergence. Not only does the method solve the problems of a limited monitoring range and a limited camera capacity, the problem of the synchronisation of a plurality of cameras also does not need to be considered, thereby reducing the cost.

Inventors:
ZHANG ZONGZHANG (CN)
PAN ZHIYUAN (CN)
WANG HUI (CN)
Application Number:
PCT/CN2018/082105
Publication Date:
November 15, 2018
Filing Date:
April 08, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
ZHANGJIAGANG INSITUTE OF INDUSTRIAL TECH (CN)
UNIV SOOCHOW (CN)
International Classes:
G05B13/00; G05D1/12
Foreign References:
CN107292392A2017-10-24
CN105867427A2016-08-17
CN106598058A2017-04-26
US20170032245A12017-02-02
US20100094786A12010-04-15
Attorney, Agent or Firm:
SUZHOU FLY PATENT AGENCY (GENERAL PARTERNERSHIP) (CN)
Download PDF: