Title Page
ABSTRACT
국문 초록
Contents
CHAPTER 1. INTRODUCTION 16
1.1. Research Background 16
1.2. Related Work 18
1.3. Research Purpose 20
1.4. Thesis Overview 21
CHAPTER 2. HYDRAULIC EXCAVATOR CONTROL SYSTEM 22
2.1. Position Control for Excavator Loading Motion 22
2.1.1. Forward Kinematics 22
2.1.2. Inverse Kinematics 25
2.1.3. Cylinder Kinematics 27
2.1.4. PID Controller 33
2.2. Experimental Setup 34
2.2.1. Hardware Setup 34
2.2.2. Controller Setup 35
CHAPTER 3. SOIL VOLUME ESTIMATION ALGORITHM 37
3.1. Soil Filtering 38
3.2. Soil Volume Estimation 40
CHAPTER 4. LOADING AND FLATTENINING ALGORITHM 43
4.1. Determination of Loading Point 44
4.1.1. Camera Coordinate Transformation 44
4.1.2. Loading Point Decision 46
4.1.3. Loading Path Generation and Following 49
4.2. Loading and Flattening Algorithm 50
4.2.1. Loading Strategy 50
4.2.2. Loading Strategy When Blind Spot Occurs 53
4.3. Soil Flattening 55
CHAPTER 5. RESULT AND DISCUSSION 56
5.1. Assessing the Accuracy of Algorithms for Soil Volume Estimation 56
5.2. Comparative Analysis of Soil Loading Flatness 58
5.3. Comparative Analysis of Soil Loading Volumes 60
5.4. Field Test Verification and Results 61
CHAPTER 6. CONCLUSION 63
REFERENCES 65
Table 2.1. D-H parameters of excavator manipulator 23
Table 2.2. RealSense camera specifications 35
Table 5.1. Volume estimation algorithm standard deviation 57
Figure 2.1. Excavator reference coordinate system 24
Figure 2.2. Designed excavator coordinate system for inverse kinematics 25
Figure 2.3. Relation between cylinder length and boom joint angle 28
Figure 2.4. Relation between cylinder length and arm joint angle 29
Figure 2.5. Relation between cylinder length and bucket angle 31
Figure 2.6. Environment for excavator loading experiment 34
Figure 2.7. Sensors and data flow used in experiment 36
Figure 3.1. Soil volume estimation algorithm flow chart 37
Figure 3.2. Result of ICP algorithm 39
Figure 3.3. Result of DBSCAN 40
Figure 3.4. RANSAC flow chart 41
Figure 3.5. Result of RANSAC 41
Figure 3.6. Volume estimation using point cloud 42
Figure 4.1. Loading and flattening algorithm flow chart 43
Figure 4.2. Camera coordinate system 44
Figure 4.3. (a) Camera coordinate system (b) Excavator boom coordinate system 45
Figure 4.4. Convert from camera to boom coordinate system 45
Figure 4.5. Point cloud converted from camera coordinate system to boom coordinate system 46
Figure 4.6. Filtered point cloud on top of the truck 47
Figure 4.7. Extracted center point on top of the truck 47
Figure 4.8. Determination of loading target point 48
Figure 4.9. Procedure for loading trajectory generation 49
Figure 4.10. (a) First soil loading (b) Second soil loading 50
Figure 4.11. Two high peaks of deposited soil on a truck 51
Figure 4.12. Highest and lowest point of soil 51
Figure 4.13. High risk of collision when loading in front and rear part 52
Figure 4.14. Loading by pushing the high peak 52
Figure 4.15. Blind spots in the soil 53
Figure 4.16. Relationship between blind spots and volume estimation 54
Figure 4.17. Comparison before and after flattening 55
Figure 5.1. Comparison of the volume estimation algorithm accuracy 56
Figure 5.2. Comparison group 58
Figure 5.3. Comparison of the soil height deviation 59
Figure 5.4. Maximum load volume comparison 60
Figure 5.5. Experiment environment in field test 61
Figure 5.6. Result of loading algorithm in field test 62