臺灣博碩士論文加值系統

This thesis proposes an optimal repulsive summation vector algorithm to create an on-line multi-obstacle avoidance strategy for six degrees of freedom (6-DOF) robot manipulator while it performs a specific task. First, the environment, obstacles and the model of the 6-DOF robot, developed by Industrial Technology Research Institute, were built in the V-REP platform as well as the required sensors and API connections. Second, the robot follows a path created by using offline path planning. Third, the optimal online repulsive summation vector algorithm is implemented in order to achieve three reactions in the robot manipulator: (1) the movement to avoid the obstacles only considering the tip, (2) the movement further considering the arm plane to efficiently avoid the obstacles and (3) sustains the same pose while the arm successfully avoid obstacles in the above reactions. In addition, the Kinect 3D sensor is used to capture the real locations of obstacles in V-REP to perform the multi-obstacle avoidance. The proposed method is validated in simulation and experiment, the results are shown it is reliable and robust to support the operation of real robot manipulator.

This thesis proposes an optimal repulsive summation vector algorithm to create an on-line multi-obstacle avoidance strategy for six degrees of freedom (6-DOF) robot manipulator while it performs a specific task. First, the environment, obstacles and the model of the 6-DOF robot, developed by Industrial Technology Research Institute, were built in the V-REP platform as well as the required sensors and API connections. Second, the robot follows a path created by using offline path planning. Third, the optimal online repulsive summation vector algorithm is implemented in order to achieve three reactions in the robot manipulator: (1) the movement to avoid the obstacles only considering the tip, (2) the movement further considering the arm plane to efficiently avoid the obstacles and (3) sustains the same pose while the arm successfully avoid obstacles in the above reactions. In addition, the Kinect 3D sensor is used to capture the real locations of obstacles in V-REP to perform the multi-obstacle avoidance. The proposed method is validated in simulation and experiment, the results are shown it is reliable and robust to support the operation of real robot manipulator.

ABSTRACT i
Acknowledgement ii
Contents iii
Figures Index v
Table Index viii
Chapter 1 Introduction 1
1.1 Motivation 1
1.2 Objectives 2
1.3 Related Work 2
1.3.1 Path Planning 2
1.3.2 Robotics Simulation 3
1.3.3 Obstacle Avoidance 4
1.4 Proposed Approach 4
1.5 Thesis Organization 6
Chapter 2 Overall System Structure 7
2.1 Robot Manipulator 7
2.2 Simulation System 9
2.2.1 V-REP Environment 10
2.2.2 V-REP Simulator Features 10
2.3 Detection System 11
2.4 Obstacle Algorithm 12
2.5 Robot Reaction 13
2.6 Motion Trajectory Generation 13
Chapter 3 ITRI Robot manipulator 14
3.1 Robot Characteristics 14
3.1.1 DH – Parameters 14
3.1.2 Robot Simulation Model 15
3.2 Representation and Homogeneous Transformation. 16
3.2.1 Elementary Rotation Matrix 16
3.2.2 Vector Representation 18
3.2.3 Transformation Matrix 18
3.2.4 Three-Angle Representation 20
3.3 Kinematics of Manipulator 22
3.3.1 Analytical Jacobian 26
3.4 Robot Forward and Inverse Kinematics 27
3.4.1 Jacobian Pseudo- Inverse: 28
3.4.2 Damped Least Squares 29
3.4.3 Analytical Solution for ITRI Robot 30
3.4.4 The Inverse Kinematics for Six-Axis Robot Arm 33
3.4.5 The Representative Points in The ITRI Robot Using The CAD Model 36
Chapter 4 Real-time Object Detection with Kinect Sensor 37
4.1 Color Detection 39
4.2 Hand Detection 41
4.3 Sensor Range 42
4.3.1 Skeletal Tracking Precision and Multiple Kinect Sensors 42
4.3.2 Calibration Between the Manipulator and Kinect 42
Chapter 5 Obstacle Avoidance Algorithm 45
5.1 The BiTRRT 48
5.2 The Artificial Potential Field 48
5.3 Repulsive Summation Vector 52
5.4 Basic Translation Reaction 52
5.5 Arm Plane 54
5.5.1 Augmented Plane 57
5.6 Fast Rotation Vector 57
5.7 Translation Reaction 60
5.8 Fast Rotation Vector Reaction 60
5.9 Sustain The Same Pose 60
5.10 Total Summation Vector 61
Chapter 6 Motion Trajectory Generation 62
6.1 Robot Controller Interface System 62
6.1.1 IMP Mode 63
6.2 IMP Controller 66
6.3 Motion Control Library 68
6.4 Path Interpolation 68
6.5 The Online Trajectory 69
Chapter 7 Experimental Results 71
7.1 Offline Paths Created by BITRRT 71
7.2 Simulation Reaction to Hands 76
7.3 Reaction to Object Using HSL Color Detection 80
7.4 Robot Reaction Obstacle Avoidance 82
Chapter 8 Conclusions 91
References 92

Figures Index
Figure 2 1. Overall system structure 7
Figure 2 2 Conventional industrial robots: a) delta robot , b) SCARA robot and, c) 6-DOF manipulator. 8
Figure 2 3 ITRI robot manipulator 8
Figure 2 4 Simulators: a) V-REP robotics simulator developed by Coppelia, b) Gazebo and Robot DK simulators 9
Figure 2 5. Simulator platform framework [11] 11
Figure 2 6. Microsoft Kinect V2 Sensor 12
Figure 3 1. Robot representation Models: (a)The CAD model. (b) The real ITRI robot arm. (c) Coordinate frame of each joint. (d) DH frames 15
Figure 3 2. CAD Models: a) 125,000 triangles; b) 39,000 triangles; c) 18,000 triangles. 16
Figure 3 3. Rotation α about the Z – axis, from x, y and z to x’, y’ and z’. 17
Figure 3 4 The classic DH convention and its parameters. 23
Figure 3 5. Damping setting for DSL calculation in simulator 30
Figure 3 6. CAD Model ITRI Robot 31
Figure 3 7. Frames location for DH parameters relation 31
Figure 3 8. Representative points in the robot 36
Figure 4 1. Different 3D sensor models 37
Figure 4 2. Kinect v2 features 38
Figure 4 3. RGB geometric representation 39
Figure 4 4. HSL color representation 40
Figure 4 5.Detection of the skeleton in front of the sensor. [21] 42
Figure 4 6. Cartesian frames: vector representation from each other 43
Figure 5 1.Offline obstacle avoidance path calculation. 45
Figure 5 2 Online avoidance. 46
Figure 5 3 Overall flow diagram 47
Figure 5 4. Representation of the path using positive and negative charges for the star and goal point. 49
Figure 5 5. Representation of local minima inside concavity. The robots move into the concavity until the repulsive gradient balance out the attractive and repulsive forces. 51
Figure 5 6. Representation of local minimum without concave obstacles. The robot moves away from the two convex obstacles until it reaches a point where the gradient vanishes; at this point, the sum of the attractive gradient and the repulsive gradient is zero. 51
Figure 5 7. Example of S inverted function: repulsive magnitude graph with parameters c=3, a=0.1m, and b=0.6m. 53
Figure 5 8. Robot and repulsive vector: (a) obstacle projection Pp, creating the RVFT and the RVT; 54
Figure 5 9.Representative points of the manipulator for the arm plane: (a) CAD model; (b) sketch to describe vectors and 3D points. 56
Figure 5 10. Representation of the augmented arm - plane 57
Figure 5 11.Avoidance of robot by retracting back: a) left retraction , b) right retraction 58
Figure 5 12. Representation of the perpendicular plane, arm plane and Repulsive vector 59
Figure 6 1 Robot interface development diagram 62
Figure 6 2 Robot information tab item 63
Figure 6 3 IMP mode tab item 64
Figure 6 4 VREP item view 65
Figure 6 5 Robot controller interface connected with VREP simulator 65
Figure 6 6 V-REP mode item using the V-REP simulator and the real robot 66
Figure 6 7. Control loop for Position velocity controller 67
Figure 7 1. Example of path to go Start position 72
Figure 7 2. Cartesian Path Offline Example 1: a) Path Go to Goal b) Path Go back to start position 72
Figure 7 3. Cartesian Path Offline Example 2: a) Path Go to Goal b) Path Go back to start position 73
Figure 7 4. Building the path using BiTRRT: a) Go to goal path, b) way back path 73
Figure 7 5.The sequence of avoidance path from Start to end point to end point. 74
Figure 7 6 Graph : joints position - time of the offline path 74
Figure 7 7 Cartesian position of the tip 75
Figure 7 8. Sequence of avoidance path from end point to start (return back path) 75
Figure 7 9. Hands Detection: a) Real Image Body, b) Open Hand and, c) close hand 76
Figure 7 10. CAD models: a) ITRI Robot and static obstacles b) hands models (mobile obstacles) 76
Figure 7 11. Offline path planning: a) and b) Initial position of the hands parallels to the robot, c),d),e),f),g) and h) shows how is the performance of the obstacle avoidance 78
Figure 7 12.Open hand reaction: a) to f) sequence of right-hand reaction for obstacle avoidance 79
Figure 7 13. Close hand reaction: a) to, f) Reaction to the left hand for obstacle avoidance 80
Figure 7 14.Object Color Detection 81
Figure 7 15. Simulation: a) to f) sequence of obstacle avoidance 82
Figure 7 16. Overall bin picking process using the offline planning and online obstacle avoidance 83
Figure 7 17. Complete simulation task: a) to, i) Complete offline path; j) to ,l) Online Avoidance; m) to, o) recalculation of the new path to complete the task 85
Figure 7 18. Complete task: a) to f) Offline path; g) to j) Online avoidance using the RVFT; K) to, o) calculation and following the next path to complete the task. 86
Figure 7 19. Response without Fast rotation vector: a) to f) sequence of the simulation and real robot 87
Figure 7 20. Response with fast rotation vector: a) to f) the response using the fast rotation vector shows a smooth trajectory and avoid to reach dangerous positions, for this approach the inclination angle is θ=30°. 88
Figure 7 21 Tip position response 89
Figure 7 22 joint position response 89
Figure 7 23 Simulation and real response for the figure 7-22 and 7-23 90

Table Index
Table 1 D-H parameter of ITRI six-axis robot arm 14
Table 2. Technical features of Kinect V2 38
Table 3. Comparison table between planners 48
Table 4. Settings for choose red color. 81
Table 5. Labels description 81

[1] L. E. Kavraki, P. Svestka, J.-C. Latombe, and M. H. Overmars, “Probabilistic roadmaps for path planning in high-dimensional configuration spaces, ” IEEE transactions on Robotics and Automation, vol. 12, no. 4, pp. 566-580, 1996.
[2] G. Sánchez and J.-C. Latombe, “On delaying collision checking in PRM planning: Application to multi-robot coordination,” The International Journal of Robotics Research, vol. 21, no. 1, pp. 5-26, 2002.
[3] F. Flacco, T. Kröger, A. D. Luca, and O. Khatib, “A depth space approach to human-robot collision avoidance, ” in 2012 IEEE International Conference on Robotics and Automation, 2012, pp. 338-345.
[4] R. C. Luo, M.-C. Ko, Y.-T. Chung, and R. Chatila, “Repulsive reaction vector generator for whole-arm collision avoidance of 7-DoF redundant robot manipulator, ” in Advanced Intelligent Mechatronics (AIM), 2014 IEEE/ASME International Conference on, 2014, pp. 1036-1041: IEEE.
[5] L.-P. Ellekilde and H. G. Petersen, “Motion planning efficient trajectories for industrial bin-picking, ” The International Journal of Robotics Research, vol. 32, no. 9-10, pp. 991-1004, 2013.
[6] T. Kroeger. (2017). Reflexxes. Available: http://www.reflexxes.ws

[7] L. Nogueira, Comparative Analysis Between Gazebo and V-REP Robotic Simulators,” Seminario Interno de Cognicao Artificial-SICA 2014, p. 5, 2014.
[8] E. Rohmer, S. P. N. Singh, and M. Freese, “V-REP: A versatile and scalable robot simulation framework,” in 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2013, pp. 1321-1326.
[9] J. J. O. Barros, V. M. F. d. Santos, and F. M. T. P. d. Silva, “Bimanual Haptics for Humanoid Robot Teleoperation Using ROS and V-REP,” in 2015 IEEE International Conference on Autonomous Robot Systems and Competitions, 2015, pp. 174-179.
[10] M. A. Olivares-Mendez, S. Kannan, and H. Voos, “Setting up a testbed for UAV vision based control using V-REP & ROS: A case study on aerial visual inspection, ” in 2014 International Conference on Unmanned Aircraft Systems (ICUAS), 2014, pp. 447-458.
[11] A. D. Luca, F. Flacco, A. Bicchi, and R. Schiavi, “Nonlinear decoupled motion-stiffness control and collision detection/reaction for the VSA-II variable stiffness device, ” in 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2009, pp. 5487-5494.
[12] F. Flacco, A. D. Luca, and O. Khatib, “Motion control of redundant robots under joint constraints: Saturation in the Null Space, ” in 2012 IEEE International Conference on Robotics and Automation, 2012, pp. 285-292.
[13] V. J. Lumelsky and E. Cheung, “Real-time collision avoidance in teleoperated whole-sensitive robot arm manipulators,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 23, no. 1, pp. 194-203, 1993.
[14] E. Rohmer, S. P. N. Singh, and M. Freese, “V-REP: A versatile and scalable robot simulation framework, ” in 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2013, pp. 1321-1326.
[15] C. W. Wampler, “Manipulator Inverse Kinematic Solutions Based on Vector Formulations and Damped Least-Squares Methods, ” IEEE Transactions on Systems, Man, and Cybernetics, vol. 16, no. 1, pp. 93-101, 1986
[16] Y. Nakamura and H. Hanafusa, “Inverse kinematic solutions with singularity robustness for robot manipulator control,” ASME, Transactions, Journal of Dynamic Systems, Measurement, and Control, vol. 108, pp. 163-171, 1986.
[17] C. S. Chen, S. K. Chen, and C. H. Lai, “Real-time coplanar NURBS curve fitting and interpolation for 6-DOF robot arm,” in 2015 International Conference on Advanced Robotics and Intelligent Systems (ARIS), 2015, pp. 1-6.
[18] O. Khatib, “Real-time obstacle avoidance for manipulators and mobile robots, ” The international journal of robotics research, vol. 5, no. 1, pp. 90-98, 1986.

[19] N. Koenig and A. Howard, “Design and use paradigms for gazebo, an open-source multi-robot simulator, ” in Intelligent Robots and Systems, 2004.(IROS 2004). Proceedings. 2004 IEEE/RSJ International Conference on, vol. 3, pp. 2149-2154: IEEE.
[20] D. Devaurs, T. Siméon, and J. Cortés, “Enhancing the transition-based RRT to deal with complex cost spaces, ” in 2013 IEEE International Conference on Robotics and Automation, 2013, pp. 4120-4125.
[21]Microsoft. (2017). Skeletal Tracking. Available: https://msdn.microsoft.com