본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색

결과 내 검색

동의어 포함

고급검색

인명/단체명

저자정보 상세정보
인명/단체명을 입력하세요.

키워드

  • 대표어

  • 외국어

-

목차보기

Title Page

Abstract

Contents

Ⅰ. Introduction 15

1.1. Motivation 15

1.2. Background 16

1.3. Related work 17

1.4. Summary of contributions 19

1.5. Organization of the thesis 19

1.6. Publications 20

Ⅱ. Image-based visual servoing 21

2.1. Conventional IBVS 21

2.2. IBVS for under-actuated system 22

2.3. Adaptive-gain IBVS 24

2.4. Square compensation for landing pad oscillation 25

Ⅲ. Feed-forward IBVS 27

3.1. System description 27

3.2. Disturbance observer based target velocity estimation 27

3.3. Kalman filter based target velocity estimation 31

Ⅳ. Autonomous ship deck landing system 35

4.1. Marker design 35

4.2. Autonomous landing procedure 35

Ⅴ. Simulations 37

5.1. Adaptive-gain IBVS 43

5.2. Square compensation for landing pad oscillation 43

5.3. Autonomous landing with disturbance observer 43

5.4. Autonomous landing with Kalman filter 45

5.5. Comparison of autonomous landing performance with disturbance observer and Kalman filter 54

Ⅵ. Experiments 57

6.1. Autonomous landing with disturbance observer 59

6.2. Autonomous landing with Kalman filter 64

Ⅶ. Lessons learned 73

7.1. Selection of the camera and lens 73

7.2. The role of the time interval in autonomous landing procedure 73

7.3. Choosing AR tags over LED or color markers 76

Ⅷ. Conclusions and future work 78

8.1. Conclusions 78

8.2. Future work 79

References 81

List of Tables

Table 5.1. Parameters related to FF-IBVS for the simulations 39

Table 5.2. Simulated ship motion in the Sea State 2 environment. 39

Table 5.3. Simulated ship motion in the Sea State 4 environment. 42

Table 5.4. Comparison of ship velocity estimation error between disturbance observer-based and Kalman filter-based methods. 55

Table 6.1. Parameters related to FF-IBVS for the experiments 57

Table 6.2. Experiment results for autonomous landing with disturbance observer: touchdown error. 62

Table 6.3. Experiment results for situation (a): touchdown error. 65

Table 6.4. Experiment results for situation (b): touchdown error. 67

Table 6.5. Experiment results for situation (c): touchdown error. 70

Table 6.6. Experiment results for situation (d): touchdown error. 72

List of Figures

Figure 1.1. Autonomous landing of a VTOL UAV (TR-60 of Korea aerospace research institute (KARI)) 15

Figure 1.2. Comparison of PBVS and IBVS with block diagram: (a) PBVS; and (b) IBVS 16

Figure 1.3. IBVS illustration. 17

Figure 2.1. Pinhole camera model. 22

Figure 2.2. Projections of the target positions on the image plane and virtual image plane: (a) on the image plane; and (b) on the virtual image plane, respectively. 23

Figure 2.3. Features are out of FOV at a low altitude. 24

Figure 2.4. Parameter c in the adaptive IBVS gain. 25

Figure 2.5. Square fitting of the features: (a) before square fitting; and (b) after the square fitting. 26

Figure 3.1. System description of autonomous landing: (a) with disturbance observer; and (b) with Kalman filter. 28

Figure 3.2. Concept of DOBC 30

Figure 3.3. Flowchart of the FF-IBVS algorithm with disturbance observer. 30

Figure 3.4. Track-to-track fusion structure. 32

Figure 3.5. Flowchart of the FF-IBVS algorithm with Kalman filter. 34

Figure 4.1. Markers used for vision-based autonomous landing. 36

Figure 4.2. State machine structure for autonomous landing. 36

Figure 5.1. Simulation setup for autonomous landing. 37

Figure 5.2. PX4 simulation in a Gazebo environment. 38

Figure 5.3. Time history of the motion of the ship at Sea State 2: (a) linear motion; and (b) angular motion. 40

Figure 5.4. Time history of the motion of the ship at Sea State 4: (a) linear motion; and (b) angular motion. 41

Figure 5.5. Flowchart of the autonomous landing algorithm. 42

Figure 5.6. Distance between the center of the features and the center of the image plane: (a) without adaptive IBVS gain; and (b) with adaptive IBVS gain. 44

Figure 5.7. Comparison of effect of the square compensation for landing pad oscillation. 45

Figure 5.8. Images captured during the simulation for square compensation: (a) before square fitting; (b) after the square fitting. 46

Figure 5.9. Simulation result of autonomous landing with disturbance observer: velocity estimation result. 47

Figure 5.10. Simulation result of autonomous landing with disturbance observer: error of the velocity estimation. 47

Figure 5.11. Simulation result of autonomous landing with disturbance observer: time history of the altitude of the UAV and the ship. 48

Figure 5.12. Simulation result of autonomous landing with disturbance observer: time history of the position error. 48

Figure 5.13. Simulation result of autonomous landing with disturbance observer: touchdown error. 49

Figure 5.14. Simulation result of autonomous landing with Kalman filter: time history of the velocity of the ship and UAV. 50

Figure 5.15. Simulation result of autonomous landing with Kalman filter: time history of the altitudes of the ship and UAV. 50

Figure 5.16. Simulation result of autonomous landing with Kalman filter: time history of the horizontal position error. 51

Figure 5.17. Simulation result of autonomous landing with Kalman filter: time history of estimated velocity of the ship. 51

Figure 5.18. Simulation result of autonomous landing with Kalman filter: time history of orientation of the ship and UAV. 52

Figure 5.19. Simulation result of autonomous landing with Kalman filter: time history of the roll and pitch angle of the UAV. 52

Figure 5.20. Simulation result of autonomous landing with Kalman filter: touchdown error. 53

Figure 5.21. Velocity estimation result at Sea State 4 situation. 55

Figure 5.22. Altitude of the UAV and the ship at Sea State 4 situation. 55

Figure 5.23. markers detection result at Sea State 4 situation. 56

Figure 6.1. Experiment setup for autonomous landing algorithm. 58

Figure 6.2. UAVs for experiments: (a) Tarot X4 equipped with a gimbal camera; and (b) Tarot 650 pro without a gimbal camera. 58

Figure 6.3. Landing platforms for flight experiments: RC car and golf cart leading a landing pad. 59

Figure 6.4. Landing platforms for flight experiments: motion platform on a truck for simulating motion of the ship. 60

Figure 6.5. Images captured during the experiments for autonomous landing with disturbance observer. 61

Figure 6.6. Experiment result for autonomous landing with disturbance observer: the time history of the velocity of landing platform 62

Figure 6.7. Experiment result for autonomous landing with disturbance observer: the time history of the altitude of the UAV 62

Figure 6.8. Experiment result for autonomous landing with disturbance observer: time history of the observed velocity of the landing platform: (a) for x-dir.; and (b) for y-dir.. 63

Figure 6.9. Experiment result for autonomous landing with disturbance observer: Trajectory of the landing platform and the UAV 64

Figure 6.10. Images captured during the experiments for situation (a). 66

Figure 6.11. Experiment results for situation (a): time history of the velocity of the landing platform 67

Figure 6.12. Experiment results for situation (a): time history of the velocity of the landing platform 67

Figure 6.13. Images captured during the experiments for situation (b). 68

Figure 6.14. Experiment results for situation (b): time history of the velocity of the landing platform 69

Figure 6.15. Experiment results for situation (b): time history of the altitude of the UAV 69

Figure 6.16. Experiment results for situation (c): time history of the velocity of the landing platform 70

Figure 6.17. Experiment results for situation (c): time history of the altitude of the UAV 70

Figure 6.18. Images captured during the experiments for situation (d). 71

Figure 6.19. Experiment results for situation (d): time history of the velocity of the landing platform 72

Figure 6.20. Experiment results for situation (d): time history of the altitude of the UAV 72

Figure 7.1. Calculation of minimum FOV considering the GPS error and attitude of the UAV. 74

Figure 7.2. Calculation of minimum camera resolution considering the altitude of the UAV and marker size. 75

Figure 7.3. The unstable situation is caused by the rapid switching between the IBVS and hold states. 76

Figure 7.4. Example of color change of the LED light according to the attitude or altitude of the UAV. 77

Figure 7.5. Example of color change of the color marker according to the lighting condition. 77

초록보기

 Unmanned aerial vehicles (UAVs) have been widely utilized in various fields such as bridge and wind turbine inspection, as well as search and rescue missions. Their importance is particularly increasing in maritime environments, including offshore wind turbine inspection, marine rescue operations, and illegal fishery monitoring. In order to enable these missions to be conducted autonomously, the process of autonomous takeoff and landing is essential. While many UAVs rely on the global positioning system (GPS) for navigation and landing, the limitations of GPS, such as large error margins, necessitate the development of more accurate landing systems, especially in challenging maritime conditions where UAVs need to land on fast-moving and oscillating ship decks.

The objective of this research is to achieve accurate autonomous landing on moving ship decks under harsh sea wave conditions. To overcome the limitations of conventional navigation methods, particularly GPS, this thesis focuses on a vision-based approach. The UAV is guided using GPS while approaching the fast-moving and oscillating ship. Once the markers on the ship deck are detected by the onboard camera, the UAV initiates the landing procedure using the vision system.

For the vision-based autonomous landing method, the image-based visual servoing (IBVS) technique is employed. However, conventional IBVS schemes are designed for static targets, and their stability for moving targets has not been validated. In this research, the aim is to land the UAV on a moving ship deck. To adapt the IBVS scheme to autonomous landing on a moving ship, the velocity of the ship is estimated and added as a feed-forward term in the IBVS controller. Two methods are considered for ship velocity estimation: disturbance observer and Kalman filter. The disturbance observer approach focuses on simplifying the communication system by relying solely on image data for ship velocity estimation, without using GPS on the ship. The motion field model, which represents the relationship between the velocities of points in 3-D space and the corresponding velocities of the features in the image plane, is used to model the system dynamics of the disturbance observer. The second method, the Kalman filter, utilizes both ship velocity data and measurements from the camera mounted on the UAV to improve the accuracy of ship velocity estimation. The measurements, which include ship position, ship velocity, and relative position between the ship and UAV camera, are fused using a track-to-track fusion algorithm.

Furthermore, techniques such as virtual image plane transformation, adaptive gain, and square compensation of the features are employed to address the under-actuated system characteristics of quad-rotor UAVs and the rotational motion of the ship induced by sea waves. A comprehensive study on the landing system is conducted, encompassing the design of markers for autonomous landing and a decision-making algorithm.

This thesis presents the research on autonomous landing on a ship deck, followed by the performance verification of the autonomous landing algorithm through numerical simulations and real flight experiments. The simulations and experiments cover both disturbance observer-based and Kalman filter-based ship velocity estimation methods. In the experiments employing the disturbance observer-based method, the UAV successfully lands on the moving platform with an average touchdown error of 0.3m under Sea State 2 conditions, with the ship oscillating and moving at approximately 3m/s. Similarly, in the experiments utilizing the Kalman filter-based method, the UAV achieves an average touchdown error of 0.2m under more challenging Sea State 4 conditions, with the ship oscillating and moving at speeds exceeding 5m/s.

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차

권호

기사명 원문보기 기사목차
닫기