15943

14239

كارشناسي ارشد

مخابرات - سيستم

اصفهان : دانشگاه صنعتي اصفهان

1399

پانزده،94ص، مصور، جدول، نمودار

محمد صادق فاضل

سمانه حسيني، محمد قوامي

مهران صفاياني، نغمه سادات مؤيديان

1399/08/10

كتابنامه

مهندسي برق

مهندسي برق و كامپيوتر

1399/08/10

2645021

Localization of person inside the building using the UWB drone and deep learning approach Leyla Nosrati l nosrati@ec iut ac ir July 5 2020 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisor Assoc Prof Mohammad Sadegh Fazel fazel@iut ac ir First Adviser Assoc Prof Samaneh Hosseini Semnani hosseini@iut ac ir Second Adviser Prof Mohammad Ghavami ghavamim@lsbu ac uk Abstract Elderly or sick people who have just been discharged from the hospital and live alone in their homes need a varietyof help As they may forget to do something albeit partially that could lead to a serious safety threat In this regard ahome surveillance system based on wireless sensor networks can take care of these people at their homes and if there areany emergencies the system can send a message to caregivers or surrounding hospitals using a radio modem to take pre cautionary measures Accurate localization in indoor environments with ultra wideband UWB technology has long beenattracted much attention However due to the presence of multipath components or non line of sight NLOS propagationof radio signals indoor UWB localization has been converted to a critical challenge Using several anchors in the indoorenvironment is one of the existing solutions But large indoor areas require a large number of anchor nodes On the otherhand in the case of unexpected events that lead to the destruction of existing infrastructures the fixed anchors cannot beused In this thesis a novel localization framework based on the transmitting signal from a moving UWB capable droneoutside of the building and its received signal regarding the modified SV channel model is presented After preprocessing ofthe received signals three new methods for reducing the indoor localization error are proposed To improve the performanceof the indoor localization system in the first method two machine learning algorithms including support vector machine SVM and multi layer perceptron MLP using extracted features of the received signals are implemented Also in thesecond method two deep learning algorithms including MLP and convolutional neural networks CNNs using raw receivedsignals are implemented The simulation results show that the architecture designed for the convolutional neural networkbased on the hybrid dataset the combination of the database related to the time and power of the received signal providesa mean absolute error MAE of about 3 cm Therefore this type of architecture offers better performance compared to theprevious methods Also in the third proposed method by using the reinforcement learning the drone is learned to fly to

محمد صادق فاضل

سمانه حسيني، محمد قوامي

مهران صفاياني، نغمه سادات مؤيديان