573 0

Research on Hybrid Localization Algorithm based Multi-Group Robot Service System with Application to Shopping Mall Environment

Title
Research on Hybrid Localization Algorithm based Multi-Group Robot Service System with Application to Shopping Mall Environment
Author
계성남
Advisor(s)
Byung-Ju Yi
Issue Date
2016-02
Publisher
한양대학교
Degree
Doctor
Abstract
Intellectualization of life is a general tendency due to the proliferation of technology and science. Based on this concept, this dissertation presents multi-group localization algorithms for multi-group robot service system (MGRS). Shopping cart problem is considered as an exemplary multi-group robot service system. The MGRS is designed to provide customers with co-service by multiple carts and allows multiple customers operation, simultaneously. In MGRS, a cart that carries personal belongings of the customer follows the customer automatically and provides real-time position information to the customer. To better help customer during the shopping trip, two types of shopping cart named by active shopping cart (ASC) are developed. Differential-driven wheels are mounted on the chassis of the shopping cart, which can switch the operating modes of the shopping cart from a passive mode to an active mode or vice versa. A single cart localization algorithms based on hybrid sensor system with application to an ASC are proposed firstly. Collecting a large number of ZigBee measurements, a regular distribution of ZigBee in a given environment is summarized by a probabilistic map. Based on the probabilistic map, a probability localization method is proposed and applied to locate the global coordinate of the ASC. And a simple hybrid sensor system combining the Zigbee and odometry is used to improve the localization performance of the active shopping cart. Through experiments, we corroborate the feasibility of the proposed single cart localization algorithm. Secondly, the multi-group robotic service system is presented. The MGRS is defined as an extension of multi-robotic system whose basic component is a robot group. In the defined MGRS, the device/robot number in a robot group will be set according to the task requirement. Therefore, even a single device/robot can be considered as one robot group. For providing customers with being co-serviced by multiple carts, a cart recognition algorithm using support vector domain description (SVDD) by LRF is proposed. To detect and track a cart by another cart with LRF, we define cart features in LRF data and employ a support vector data description method. The effect of the cart recognition algorithm is verified through experiment in a given indoor environment. Lastly, multi-cart localization algorithms for MGRS is proposed. To fulfill estimating the location of MGRS, a hybrid external localization algorithm based on combination of QR location information and ZigBee location estimate is proposed. The probability localization method based on ZigBee proposed for single cart localization is improved by using a Bayesian Classifier. The feasibility of using a camera to get the position information of QR code is discussed and Wi-Fi based QR localization algorithms are proposed. The proposed hybrid external localization algorithms reduces the fluctuation caused by overlap of Gaussian distribution and influence of the environment noise. And recognition of customer-cart groups in MGRS is realized by ZigBee blind nodes on the cart. By employing the proposed HELA based multi-cart localization algorithms, the MGRS is verified to realize accurate and stable localization of multiple customer-cart groups through three experiment trials. It is remarked that the MGRS is capable of achieving the goal of providing customers with co-service by multiple carts and allowing multiple customers operation, simultaneously.
URI
https://repository.hanyang.ac.kr/handle/20.500.11754/126934http://hanyang.dcollection.net/common/orgView/200000428150
Appears in Collections:
GRADUATE SCHOOL[S](대학원) > ELECTRONIC SYSTEMS ENGINEERING(전자시스템공학과) > Theses (Ph.D.)
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

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

BROWSE