期刊名称：Facta universitatis - series: Electronics and Energetics
印刷版ISSN：0353-3670
电子版ISSN：2217-5997
出版年度：2017
卷号：30
期号：3
页码：403-416
出版社：University of Niš
摘要：This paper presents two simple and cost effective indoor localisation methods. The first method uses ceiling-mounted wide-view angle webcam, computer vision and coloured circular markers, placed on the top of a robot. Main drawbacks of this method are lens distortion and sensitivity to lighting conditions. After solving these problems, a high localisation accuracy of ±1cm is achieved at about 5 Hz sampling rate. The second method is a version of trilateration, based on ultrasound time of flight distance measurement. An ultrasonic beacon is placed on a robot while wall detectors are strategically placed to avoid an excessive occlusion. The ZigBee network is used for inter-device synchronisation and for broadcasting measured data. Robot location is determined as a solution to the minimisation of measurement errors. Using Nelder-Mead algorithm and low-cost distance measuring devices, a solid sub 5 cm localisation accuracy is achieved at 10Hz.
关键词：Robot localization; Nelder-Mead; Gnu Scientific Library; USB camera; OpenCV
