Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
MISSING TARGET SEARCH METHOD BASED ON REINFORCEMENT LEARNING ALGORITHM
Document Type and Number:
WIPO Patent Application WO/2021/103248
Kind Code:
A1
Abstract:
Provided is a missing target search method based on a reinforcement learning algorithm. The method comprises the following steps: step S1, data pre-processing: comprising the discretization of time and space, the discretization of a target movement trajectory, and the standardization of the search difficulty in different times and spaces; step S2, construction of a reinforcement learning training environment: constructing a reinforcement learning training environment, wherein training environment information includes expected search costs, at different search moments, of objects starting from different positions at different times and the probability of same transferring to different positions at different search moments; step S3, offline training of a time-space search model: defining states and behaviors, and performing adaptive optimization on the model; and step S4, online time-space search decision making: iteratively determining, on the basis of the trained time-space search model in step S3, a time-space search sequence by using a greedy policy, and executing a time-space search. According to the present invention, the search cost for finding the position of a target at a target moment is effectively reduced, and a target search task is completed under the constraints of the search cost.

Inventors:
YU ZHIYONG (CN)
HAN LEI (CN)
HUANG FANGWAN (CN)
GUO WENZHONG (CN)
Application Number:
PCT/CN2019/128552
Publication Date:
June 03, 2021
Filing Date:
December 26, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
UNIV FU ZHOU (CN)
International Classes:
G01S19/19
Foreign References:
CN109521447A2019-03-26
CN101339240A2009-01-07
US20190108555A12019-04-11
Other References:
LEI HAN, YU ZHIYONG ,ZHU WEIPING, YU ZHIWEN: "Spatiotemporal Search Optimization for Vehicle Destination Inference", COMPUTER ENGINEERING, vol. 46, no. 2, 15 February 2020 (2020-02-15), pages 28 - 34, XP055816144, ISSN: 1000-3428, DOI: 10.19678/j.issn.1000-3428.0053543
Attorney, Agent or Firm:
FUZHOU YUANCHUANG PATENT AND TRADEMARK AGENT LTD. (CN)
Download PDF: