자율무인잠수정의 주행경로 인식을 위한 퍼지칼만필터 설계에 관한 연구

손경민 부경대학교 대학원 2014 국내석사

Recently, the global issues about the depletion of fossil fuels have brought much concerns about the technology for the collection of marin resources. The marin plant is playing an important role in obtaining resources from ocean. Since the human activity is restricted in marine environment, the autonomous underwater vehicles (AUV) have been developed to conduct tasks in the marin environment. AUV are categorized into wireless AUV and wired AUV. Currently, the AUV using tethering cables communicating between AUV and users have been widely used. However, they have limited in their working ranges due to the tethering cables. For this reason, the needs of advanced wireless AUV are increasing. The autonomous navigation system is required for the AUV to carry out the tasks in remote areas. The navigation systems are categorized into dead reckoning navigations, which utilizing environmental data from mounted sensors, and positioning navigations. Although the dead reckoning navigation can correspond well to the environmental variations based on the data from sensors, the error accumulation occurs while the navigation algorithms, which consisting of derivative equations, are running. In contrast, positioning navigation is less affected from the error accumulation problems due to the use of real time data from GPS and acoustic sensors. However, the vehicle needs to rise to the surface periodically to receive data from GPS, and the acoustic sensor system is easily affected by the condition of sea water, such as temperature, salinity, and density. For this reason, the signal processing technique to reduce the error accumulation has been developing along with improving the precision of sensors in dead reckoning navigation systems. The paper investigated the path recognition of AUV using signal processing technique with fuzzy logic and kalman filtering. To develop the kalman filter, a model that simulates output signals from sensors was developed. Also this study investigated an algorithm that detecting moving paths when the movement of vehicle occurs to reduce the error accumulation while tracking paths. When the moving of the AUV is detected, the fuzzy logic and kalman filter are activated to process signals. In order to verify the navigation system, a simulator, enabling the straight line motion and circular motion, was built. Finally, the best signal processing method between the fuzzy logic and kalman filter was selected and applied to the performance verification of the AUV.

인공지능을 활용한 IR-UWB 레이더의 물체인식에 관한 연구

이동헌 부경대학교 2017 국내석사

본 논문에서는 IR-UWB Radar를 이용하여 물체마다 특징을 추출하여 구분하고자 한다. 본 논문에서 언급한 특징이란 물체가 구성된 재질에 따라 IR-UWB Radar에 반사되어 돌아오는 신호의 크기 및 그 신호의 주파수 특성 즉, 이산 푸리에 변환을 수행해보면 각각 물체마다 주파수 특성이 다르게 나타난다는 것을 확인할 수 있을 것으로 예상되었다. 이러한 이론을 실험으로 증명하기 위해 본 논문의 실험에서는 그림5.3과 같이 IR-UWB Radar를 책상 위에 고정시키고, 수직 선상 위치에 측정하고자 하는 대상 물체를 위치시켰다. 또한, IR-UWB Radar를 조작 및 제어하기 위해 MATLAB을 연동하여 실시간으로 측정할 수 있도록 하였다. 본 논문에서는 앞에서 언급한 물체의 특징을 추출하기 위해 IR-UWB Radar의 출력은 다른 주파수의 간섭을 최소화하고 본 실험에 사용된 IR-UWB Radar의 특성에 맞는 주파수 대역만을 얻기 위해 밴드패스 필터를 설계 하였다. 이러한 밴드패스 필터는 약 5∼8GHz 대역만을 통과하게 된다. 본 논문에서는 FIR Bandpass Filter를 통과함으로써 신호의 일그러짐과 플로팅 성분이 제거된 IR-UWB Radar의 출력 신호를 이산 푸리에 변환을 진행 하였으며, 이러한 신호처리를 통해 물체마다의 주파수특성을 확인 할 수 있었다. 이렇게 수집된 주파수 특성으로 데이터베이스를 구축하고, 이 데이터를 이용하여 인공 신경망을 구성하여 학습을 진행하였다. 또한 FIR Bandpass Filter를 통과한 IR-UWB Radar의 출력 신호를 TOA(Time Of Arrival) 방법을 통해 물체와의 거리를 측정 할 수 있었다. 상위에서 언급한 실험 및 연구를 위해 본 논문에서는 2장에서 IR-UWB Radar System에 관해 설명을 하며, 3장에서는 IR-UWB 레이더의 수신된 신호의 처리과정을 언급 및 설명한다. 그리고 이러한 신호처리를 통해 수집된 데이터를 이용하여 인공신경망을 구성하여 물체 구분을 위한 알고리즘을 4장에서 설명한다. 마지막으로 5장에서는 IR-UWB Radar을 이용하여 실제 실험데이터 및 실험환경에 대한 언급과 결과를 나타내었다.

都市鐵道 變電設備用 사이리스터 듀얼 컨버터 竝列運轉을 위한 最適의 모드 轉換 알고리즘 硏究

한성우 부경대학교 2017 국내박사

Most of the systems that supply the electric power required for the operation of the electric motor car of the Urban railway convert the AC into DC and supply it through the trolly wire. The rectifier for converting AC to DC is diodes rectification a method and thyristor rectification a method. The rectification method using a thyristor is a configuration in which a thyristor converter of 12 pulses is connected in a forward direction and a reverse direction, The thyristor converter connected in forward direction operates to supply electric power to the trolley wire. Thyristor converter connected in reverse direction, is operated in reverse mode when regenerative electric power is generated due to regenerative braking at the time of stopping the electric train and the voltage of the trolley wire rises. The thyristor converter can save about 20% of the operating power of the electric car by returning the voltage of the surplus trolley wire to the AC main by reverse mode operation. Therefore, a rectification method using a thyristor is advantageous for improving the operation efficiency of the Urban railway and reducing the greenhouse gas. However, the thyristor dual converter has a voltage fluctuation rate in the hysteresis band at the time of mode conversion and has a different voltage fluctuation rate depending on the load current slope at the time of mode conversion. The purpose of the thyristor dual converter is to keep the trolly wire voltage constant even with load changes. Therefore, it is important to minimize trolley wire voltage changes through stable mode conversion even when the load direction changes. In this paper, we propose a dual converter parallel operation that enables normal operation of electric motor car even when the trolly wire facility is simplified and the dual converter of the adjacent substation fails. The proposed dual converter parallel operation control method is as follows. First, it is a sequential mode conversion algorithm using cyclic current. The dual converter requires a zero-current discontinuity section to prevent short circuit when conversion between forward and reverse modes by load current. This section causes undershoot and overshoot in the DC output voltage. Even when the dual converter operates in parallel, there is instability in the DC output voltage. The sequential mode conversion algorithm using the circulating current can obtain a stable DC output voltage at the time of mode conversion even if the slope of the load current changes or has a value near the zero current. Second, it is an unbalanced voltage compensation algorithm. The reference value of the voltage error compensator is changed according to the error value even if there is voltage sense error of each dual converter that may occur in the actual system. This suppresses the overshoot and undershoot of the DC output voltage caused by the parallel converter operation. The unbalanced voltage compensation algorithm prevents the circulating current divergence between the parallel operation dual converters and improves the phenomenon that the load current is concentrated in one dual converter. The parallel operation control method of the proposed thyristor dual converter is analyzed by PISM simulation and the manufactured of 10 kW small scale dual converter verified by experiments.

센서 기반 넘어짐 동작을 인식하기 위한 딥러닝 모델 아키텍처 설계

조소현 부경대학교 2022 국내박사

With the popularization of Personal Mobility Vehicles (PMV) such as bicycles, electric scooters, etc., user demand is increasing. PMVs are being used as some substitutes for walking or using public transportation, and Sharing System have made it easier for individuals to access the means without having to own them. In particular, as the demand for delivery has increased significantly in the past few years, the number of drivers for motorcycles, bicycles, and electric scooters as a means of delivery has increased significantly. The global electric scooters market size is expected to grow at a Compound Annual Growth Rate (CAGR) of 7.8% until at least 2030. However, as the number of users increases, the occurrence of large and small traffic accidents is increasing. In the case of a two-wheeled vehicle accident, the body of the rider is exposed to the impact as it is, and thus the accident makes injury more serious. Also, after the first collision, a secondary collision occurs due to surrounding structures. In the process, a significant number of riders suffered serious head and neck injuries. Reducing injuries from crashes is important to protect the safety of occupants. For this, a technology to recognize and judge the current movement state through real-time information about the change in the rider's posture is needed. In this paper, I performed performance evaluation for motion detection according to the deep learning algorithm when Inertial Measurement Unit (IMU) sensor data on the rider's movement is given, and through this, I proposed a new improved architecture. A convolution operation was performed on each axis of the acceleration and angular velocity sensors, and Residual block was used to design it. Through this, the characteristics of each axis were analyzed individually, and the accuracy was greatly improved. In order to build the dataset, I conducted accident experiments using a mannequin. I attached a sensor to the back of the mannequin and collected information on acceleration and angular velocity according to the movement of the rider. I extracted and analyzed acceleration and angle information to find out the motion characteristics, and made datasets for Deep Learning. In the case of a deep learning model, the architecture is implemented using algorithms such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Long Short Term Memory (LSTM). In order to select an appropriate deep learning technique for the dataset used in this study, I performed performance evaluations on the output according to the change of the algorithm while maintaining hyperparameters such as layers and nodes. For this dataset, the CNN-based method showed the best performance, and analysis along each axis and time axis of the sensor was also performed. Based on these studies, I designed a new improved architecture TAMS (Time Attention for Multi Sensor) using the CNN algorithm. TAMS showed good performance with fewer layers and only 1/5 epoch than the comparison model. To check the operation of the model trained using the dataset, an experiment was performed through the test bed. The model has been ported to an embedded system based on Raspberry Pi, and when an accident is detected, the operation can be confirmed by deploying the airbag. As a result, the Raspberry Pi module detected the accident in real time and activated the airbag immediately after the first collision, confirming that the accident was determined. From the results of each experiment, real time motion detection of passengers is possible through deep learning using a single IMU sensor, and it was confirmed that the performance was improved compared to the existing model through TAMS design. In addition, it is expected that data for PVM other than the bicycle data used in this study will be available.

영상기반 객체인식을 위한 딥러닝 모델의 활성화 함수 설계

우주 부경대학교 2022 국내박사

Trams are considered as next-generation public transportation driven by electricity as eco-friendly energy. The subway is also operated by electricity, but the tram has the advantage of being easier to install tracks than the subway. Also, it is installed on the ground, so anyone can easily access it. In this paper, the experimental purpose is finding out a more suitable dataset for object recognition in trams, and what performance changes the self-developed Rish activation function makes in the object recognition model. To identify, the experiment was conducted as follows. First, each YOLOv4 model was trained with different datasets with different classes, respectively. The YOLOv4 models were compared to which dataset is easier to learn and compared to which dataset is more suitable for developing an object recognition model for autonomous driving through a camera installed in a tram. The Rish activation function and the Mish activation function used by the original YOLOv4 model were compared in the same way as the dataset experiment.

RAMS 통계분석을 활용한 도시철도 전동차 출입문 부품 고장분석 및 개선에 관한 연구

김상균 부경대학교 2018 국내석사

The entrance and exit system of the city rail system is a device that handles the entry·exit of passengers. However, it is an important device affecting the operation of the city railroads since failure in the event of failure directly affects the safety of the passengers The type of belt-type door system was analyzed using the analysis of the RAMS integrated system statistics that are set up in Busan Metropolitan Railway for improved passenger service and safe operation of the city railroads. As a countermeasure to improve it, First, developing and applying the door belt pulley bearing and upper roller replacement equipment reduced the annual average failure rate of the belt pulley, the bearing, and the upper roller by 56.7 percent. Second, the development and application of the door operated testing machine reduced the annual average failure rate of the door belt pulley bearings and upper rollers by 91.7 percent by recognizing bearing defects and bad assembly conditions before the motor vehicle was shipped. The results are verified by statistical analysis of the RAMS integrated system.

IR-UWB 레이더를 이용한 3차원 위치 측위에 관한 연구

최명훈 釜慶大學校 2017 국내석사

Nowadays, researches on locating systems have been actively conducted in various fields such as security, military and smart devices. A typical system used in smart devices and automobiles is Global Positioning System (GPS). This system finds a destination based on signals from satellites. In an emergency, one can also communicate one’s position so that others can respond. However, GPS has its weak points. With GPS it is not possible to track location indoors, and if a disturbance occurs, it is difficult to detect the object again. Ultrasonic sensors, cameras and infrared sensors are typically used to detect objects. But since these sensors are highly affected by their surrounding environment, there are many errors caused by the environment, and detection distance is very limited. This paper uses IR-UWB to detect objects. The IR-UWB radar system radiates regular impulse signals from TX antennas and checks the times to receive the impulse signals that are reflected from the target to measure the Time of arrival (TOA). It has low power consumption and high energy efficiency, and has a very wide spectrum band. To remove unwanted signals, the average value of the Raw-data is removed and a band-pass filter is used. The coordinates of the target is found by measuring the distance values. Two distance values (X, Y) are required to find two dimensional coordinates. Three distance values (X, Y, Z), using the Pythagorean theorem, are required to find three dimensional coordinates. This is called Trilateration. The experimental results in this paper had ranging errors of 3~10cm. These errors were caused by the process of moving one radar to each position. The results confirm that three dimensional coordinate measurement using IR-UWB is possible.

Lighting Control of LED Luminaries Using LED Sensor

Abu Yusuf Md. Salehuddin 부경대학교 대학원 2014 국내석사

조명제어에서 제어 대상은 밝기, 색온도, 연색지수이다. 조명제어를 위해서는 이들 값을 인식할 조도센서, 값비싼 칼라센서 등이 필요하다. 대형 조명장치에는 수많은 조명기로 구성되어 있으며, 조명장치의 균일한 조명환경을 유지하기 위해서는 각 조명기에 대한 개별적인 조명제어를 해야 한다. 따라서 조명장치의 조명제어는 매우 값비싼 제어가 된다. 조명의 색온도와 연색지수는 광의 Red, Green, Blue 스펙트럼 분포로 결정된다. 그런데 LED는 전기를 가하면 빛이 발생하고, 역으로 LED에 빛을 가하면 전기가 발생하는 반도체이며, 또한 Red LED는 Red 광 스펙트럼 분포에 비례한 전기신호를 발생할 수 있고, Green LED는 Green 광 스펙트럼 분포에 비례한 전기신호를 발생하고, Blue LED는 Blue 광 스펙트럼 분포에 비례한 전기신호를 발생한다. 따라서 특정한 색온도와 연색지수를 갖는 빛에 대한 특성을 R, G, B LED의 전기신호를 해석하여 판단할 수 있다. 그러나 LED의 광에 대한 전기신호는 심한 비선형적인 특성을 가지고 있으므로, 본 연구에서는 LED를 센서로 활용하기 위하여 Fuzzy Logic을 적용하여 보았다. 제작된 조명기에 조도센서, R, G, B LED센서를 장착하여, 조명기가 내는 밝기, 색온도, 연색지수 값을 피드백하여 지정된 밝기, 색온도, 연색지수 값과 비교하여 그 오차에 대하여 PI(비례, 적분) 보상기로 보상하여 균일한 조명을 구현 하였다. 따라서 본 연구는 조명제어를 위하여 고가의 칼라센서 대신에 R, G, B LED를 활용할 수 있음을 보여 준다.

소형풍력발전기를 위한 시뮬레이터 개발에 관한 연구

박원현 釜慶大學校 2017 국내석사

As the Paris Agreement takes into effect, the reduction of greenhouse gas emissions is mandatory. It would be inevitable to develop and apply new and renewable energies which are one of the best ways to replace fossil fuels in order to reduce emissions. Most of all, wind power generator produces electricity by converting wind power to energy, drawing much attention as one of new and renewable energies after the Paris Agreement’s conclusion. Above the rest, small wind power generation system is highly regarded for its handy installation, so various forms of blades and towers are being produced in the light of balance with the surroundings. There had been wind tunnel experiments with similar environment to the actual wind power by embodying wind power generation system. However, it had been costly and limited in time and space. Therefore, this paper means to draw the similar result to the actual wind tunnel experiment by producing a wind power generation simulator. The simulator is composed of induction motor, generator, inverter, battery controller and battery. Since the rotation speed of the wind power generator blade is variable in response to counter electro-motive force of the battery and the generator, Matlab-Simulink is used for the simulation in order to figure out the behavior properties of helical blade for this paper. Motor control algorithm and maximum power point tracking(MPPT) control algorithm are developed for the motor to control behavior properties derived from the simulation. In doing so, this paper has compared energy production from generator according to the rotation speed of each blade, with energy production from actual wind tunnel experiment. The result shows the similar conclusions with a 5 percent margin of error.

LED 가로등의 색도를 이용한 자율주행 차량용 차로 위치측위에 관한 연구

정재훈 부경대학교 대학원 2017 국내석사

Recently, many studies on autonomous vehicles are under way. Location recognition and positioning system for cars are one of the very essential parts in unmanned vehicles. These systems are on the basis of GNSS(Global Navigation Satellite System), using GPS (Global Positioning System)and it calculates location of the receiver using triangulation according to the moment when the satellite signal reaches to the receiver on the surface of the earth by using satellite network moving around the space orbit to measure the position. About seven-meter margin of measurement error occurs, resulting from various error factors. In addition, a margin of error of DGNSS(Differential GNSS) is merely about two meters, compared to GNSS by revising errors on a satellite clock, the ionosphere, the and the track, using Master Station on the ground, but accurate measurement is difficult to make in urban multi-path environment as GNSS is. As part of the improvement, researches are in progress to enhance accuracy of the position, fusing INS, Vision, Radar, Terrestrial Magnetism Sensor, Wi-Fi, etc. INS-GPSis one of the most typical sensor fusion positioning system, having problems with sharp occurrence of measurement errors in accordance with time. Plus, its performance is degraded in GPS interfering spots due to its high dependence on GPS. In short, sensor fusion positioning technique continues to be studied with many unsettled matters.[1-13] To supplement problems with the current positioning system of vehicles, this paper suggests a positioning technique on the road by analyzing chromaticity coordinate, judging from color temperature of LED street lights and tunnels which are one of infrastructures on the roadway. The positioning technique developed in this paper is expected to be applicable when it is difficult to recognize lanes on account of their poor conditions and in GPS interfering spots by examining LED lights with different color temperatures on the respective roads for positioning locations on the road. Many research bodies are studying on the positioning system and it is considered to improve the performance as the technique covered on this paper applies to the conditions. This paper consists of five chapters. Chapter 2 describes a chromaticity-related theory of LED lights to help understand the context of the research. Chapter 3 outlines the system structure for measuring chromaticity of LED lights. Chapter 4 gives a description of designing fuzzy controller and the experiment result. Chapter 5 suggests future research direction and the conclusion on positioning on the road using chromaticity of LED lights, based on the experiment result.