Study and Hardware Design using Received Signal Strength Indicator Positioning Technology with Genetic Algorithms

Autor: Rui-Ming Yang, 楊叡明
Rok vydání: 2019
Druh dokumentu: 學位論文 ; thesis
Popis: 107
In recent years, emergence of Internet of Things (IoT) make people rely on electronic products with analyses of big data, such as wearable devices, smart medical products, and environmental monitoring apparatus. We hope the technology can help people to obtain the location of everything, such as electronic tracking labels for merchandise, old people, and animals, etc. The labels are usually cheap, low power consumption, and long-lived battery. Thus, development of different positioning algorithms to improve positioning accuracy is the trend in many localization studies. The principle of wireless localization is to obtain the relative position of the target by extracting the received signal characteristics, such as azimuth of arrival (AoA), time difference of arrival (TDoA), received signal strength indicator (RSSI), and so on. The RSSI is a low-power method without extra sensors but it is vulnerable to the environment and variation of devices resulting in poor localization. In this thesis, we propose a progressive global based symbiosis-genetic algorithm (PGBS-GA) for RSSI localization. It combines the excellent ability of complete searching algorithm and the information of multiple anchors with the physical and mathematical characteristics to find the improved noise compensation value of each RSSI, and improve the localization accuracy. In this thesis, Python was used to develop and verify the algorithms. The simulation results show that the algorithm performs well in different noise environments. The average positioning error is only 47 m in the area of 120000 m2, which is reduced by 100 m if the conventional method is used. The minimum error can be as low as 8 m and the maximum error is 76 m. The Arduino microprocessor and LoRa wireless communication module were used to set up the experimental apparatus for the outdoor RSSI measurement. Three sets of positioning data were obtained. The average error is about 34.85 m. In comparison with the other algorithms, the error is reduced by 20~30 m. The proposed algorithm is quite stable and uses little amount of data. However, it needs complicated searching and computation. The computation time by software is about 0.74 s/gen. If it is desired to complete 40 generations within 0.5 ms, that indicates the hardware accelerator with 12.5 us/gen. is required. Especially, the Sub Evaluator and the Global Evaluator occupy 17% and 17% computation times, respectively, so their accelerating hardware was designed and implemented using Xilinx Sparten-6 XC6SLX150T and ISE Design Suite 14.5 for synthesis and APR. The post-APR frequency is 47.689 MHz with the throughput of 11.922 M frames/sec, which is 87,874 times faster than the software computation. If the other computation is performed by hardware, the equivalent comuting time of 10.9 us/gen. may be obtained.
Databáze: Networked Digital Library of Theses & Dissertations
Externí odkaz: