1. Industrial robots with an open-architecture controller xxvii

1.1. PUMA560 manipulator 5

1.2. Block diagram of the resolved acceleration control method, where x_r, and ctI*G the desired position/orientation, velocity and acceleration vectors in Cartesian coordinate system 6

1.3. Trajectory following control problem 7

1.4. Desired joint angle θ_r, velocity and acceleration in joint space 8

1.5. Genotype of K_v and K_p which means an individual 9

1.6. Evolutionary history in case that the resolved acceleration control law was employed 12

1.7. Simulation result in case of using the gains with minimum fitness 14

1.8. Block diagram of impedance model following force control with I-action 17

1.9. Block diagram of impedance model following force controller, in which x_r(k), and are transformed into the joint driving torque T 19

1.10. Antecedent membership functions for e_x(k). 21

1.11. Antecedent membership functions for Δe_x(k). 21

1.12. Profiling control situation 23

1.13. Time histories of environmental stiffness in x-and z-directions 23

1.14. Block diagram of the force control system using a fuzzy reasoning 24

1.15. Force control result using only IMFFC 25

1.16. Force control result using IMFFC with fuzzy force controller 25

1.17. Position compensations Δx(k) and Δz(k) generated from the fuzzy force controller 26

1.18. A desirable force response ^sE_fz (k) for a teaching signal obtained through simulation 28

1.19. Velocity of arm tip which is the teaching signal of output of neural network 29

1.20. Neural network for dynamics learning of contact motion 30

1.21. Learning history of error function E. 31

1.22. Output generated from the learned neural network 31

1.23. An example of force control result by feedforwardly using the learned NN 32

1.24. Block diagram of the force control system using a neural network 33

1.25. The upper, middle and lower figures show the outputs from IMFFC+NN, NN only and IMFFC only, respectively 34

2.1. Transformation from position command to joint driving torque 37

2.2. Example of force control under known environments 42

2.3. Stiff end-effector fixed to the tip of the robot arm 43

2.4. Force control results: (a)K_m1 = K_m3 = 5000 N/m, (b)K_m1 = K_m3 = 10000 N/m, (c)K_m1 = K_m3 = 15000 N/m, (d)K_m1 = K_m3 = 20000 N/m 44

2.5. Block diagram of the impedance model following force controller with the FEM 49

2.6. Fuzzy rules encoded into a chromosome 50

2.7. Estimated stiffness for each environment  52

2.8. Relationship between true stiffness of environment and desired damping coefficients 53

2.9. An example of evolutionary history in case of K_m1 = K_m3 = 17500 N/m 54

2.10. Learned antecedent membership functions (best individual) in the case of K_m1 = K_m3 = 17500 N/m 55

2.11. Simulation results using the best individual in the case of K_m1 = K_m3 = 17500 N/m: (a) force control result, (b), (c) time history of estimated stiffness , and desired damping ,  56

2.12. Generalized antecedent membership functions for GFEM 59

2.13. Hybrid position/force control problem under unlearned environments 60

2.14. Force control results in x- and z-directions under an unlearned environment 60

2.15. Position control result in y-direction 61

2.16. Desired damping coefficients in the x-direction 61

2.17. Desired damping coefficients in the z-direction 63

2.18. Force control results in case that a larger magnitude is given to the desired contact force 64

3.1. Desktop-size industrial robot RV1A 68

3.2. Relation between CL data cl(i) and desired trajectory ^wr(k), in which ^wO is the origin of work coordinate system 69

3.3. Relation between position component p(i) in CL data and desired position ^wx_d(k), in which p(i) – ^wx_d(k) and p(i + 1) – ^wx_d(k + 3) are called the fraction vectors 70

3.4. Block diagram of communication system by using UDP packet 70

3.5. Communication scheme by using UDP packet between PC and Industrial robot RV1A, in which the sampling period Δt is set to 10 ms 71

3.6. Orientation control of arm tip based on CL data shown in Fig. 3.2, in which the initial orientation is given with [ϕ(k) θ(k) ψ(k)]^T = [0 0 0]^T 74

3.7. CL data cl(i) = [p^T(i) n^T(i)]^T consisting of position and orientation components, which is used for desired trajectory of the tip of robot arm 75

3.8. Initial part of x-component X_d(k) of desired trajectory in robot absolute coordinate system 76

3.9. Initial part of y-component Y_d(k) of desired trajectory in robot absolute coordinate system 77

3.10. Initial part of z-component Z_d(k) of desired trajectory in robot absolute coordinate system 77

3.11. Initial part of desired roll angle ϕ_d(k) calculated from the desired trajectory shown in Fig. 3.7. 78

3.12. Initial part of desired pitch angle θ_d(k) calculated from the desired trajectory shown in Fig. 3.7. 78

3.13. Roll angles around x-axis, in which (a) is the initial angle represented by ϕ(k) = 0, (b) is the case of — π ≤ ϕ(k) < 0, (c) is the case of ϕ(k) = ± π, and (d) is the case of 0 < ϕ(k) ≤ π 85

3.14. Control result of x-directional position X(k) 86

3.15. Control result of y-directional position Y(k) 86

3.16. Control result of z-directional position Z(k) 86

3.17. Control result of roll angle ϕ(k) 87

3.18. Control result of pitch angle θ(k) 87

3.19. Timing chart between a robot controller and a PC through Ethernet, in which UDP packet with 196 bytes is used 88

3.20. Experimental scene, in which a student is pulling the arm tip with 10 N in x-direction 88

3.21. Experimental results when Eq. (3.27) is used in x-direction with several desired stiffness K_dx N/mm. The relations between the time and position are plotted 89

3.22. Experimental results when Eq. (3.30) is used in x-direction with several desired dampings B_dx Ns/mm. The relations between the time and position are plotted 89

3.23. Experimental results when Eq. (3.32) is used in x-direction with several damping cofficients ζ. The relations between the time and position are plotted 90

4.1. Handy air-driven sanding tools usually used by skilled workers 92

4.2. Example of a whirl path generated based on contour lines 94

4.3. Normalized tool vector n(i) represented by θ₁(i) and θ₂(i) in work coordinate system 95

4.4. Block diagram of the proposed hybrid position/force controller with weak coupling 96

4.5. Robotic sanding system developed based on an open-architecture industrial robot KAWASAKI FS30L 99

4.6. Hardware block diagram between a Windows PC and a PC-based controller 99

4.7. Polishing force F(k) composed of contact force f (k) and kinetic friction force F_r(k) 102

4.8. Zigzag path generated from the main-processor of a CAM system. Each through point has a position and an orientation components 103

4.9. Example of proposed hyper CL data. 105

4.10. Overview of the robot sander developed based on KAWASAKI FS20 107

4.11. Sanding scenes using the proposed 3D robot sander developed based on a KAWASAKI FS30L 109

4.12. Curved surfaces before and after polishing process 109

4.13. Wooden chair with an artistic curved surface (Courtesy of Furniture Workshop Nishida) 109

5.1. Conventional five-axis NC machine tool with a tilting head 120

5.2. Main- and post-processes to compute corrected NC data for five-axis NC machine tool 121

5.3. Artistic paint roller models designed by 3D CAD 121

5.4. Definition of the tool length of a tilting head 122

5.5. Inclination and rotation of main head 122

5.6. Motion of tip of ball-end mill generated from main-processor 123

5.7. Sample model with curved surface and the five-axis machining scene 123

5.8. Overview of the proposed machining scheme for cylindrical wooden workpieces 127

5.9. Three-axis NC machine tool MDX-650A with a rotary unit 127

5.10. Relief design illustrated by a 3D CAD 128

5.11. An example of zigzag path on a flat model (y_length = y_max y_min) 130

5.12. Typical edge chipping 130

5.13. An example of relation between the distance and the point density with respect to CL data 133

5.14. Antecedent membership functions for d(i) and Δd(i) 134

5.15. Carving scene of a wooden paint roller with an artistic design 135

5.16. An example of the feed rate values given to NC machine tool, in which the total number of steps is 400, F_max and F_min are set to 2000 mm/min. and 600 mm/min., respectively 137

5.17. Artistic design wooden paint rollers carved by the proposed system 138

6.1. Mold polishing robot developed based on YASKAWA MOTOMAN UP6 145

6.2. Generation of normalized tool vector n(i) at p(i) on a free-formed surface 147

6.3. Polishing strategy by efficiently using the contour of a ball-end abrasive tool 149

6.4. Force sensor to measure the polishing force ||F|| 150

6.5. Relation between the tool’s center and CL data 153

6.6. PET bottle blow mold before polishing process 155

6.7. Experimental setup 156

6.8. Desired trajectories composed of zigzag paths 157

6.9. Experimental scene using the proposed mold polishing robot 158

6.10. Comparison of CL data and actual trajectory 160

6.11. X-directional actual trajectory obtained through an experiment 160

6.12. Constriction part of mold surface after wiped with a cloth containing the polishing compound Cr₂O₃ 161

7.1. LED lens mold with several concave areas precisely machined within a tolerance, which is a typical axis-asymmetric workpiece 177

7.2. Polishing scene of a PET bottle blow mold using a conventional articulated-type industrial robot with a servo spindle 177

7.3. Static relation between position and contact force in case of a conventional articulated- type industrial robot 178

7.4. Proposed desktop orthogonal-type robot with compliance controllability 178

7.5. Hardware block diagram of the desktop orthogonal-type robot composed of three single-axis devices with a position resolution of 1 μm 179

7.6. Static relation between position and contact force in the case of the desktop orthogonal-type robot with a stiff ball-end abrasive tool 179

7.7. Static relation between position and contact force in the case of the desktop orthogonaltype robot with a rubber ball-end abrasive tool 179

7.8. Block diagram of the CAD/CAM-based hybrid position/force controller with a velocitypulse converter 180

7.9. Relation between the polishing force error E_f(k) and manipulated variable v_normal(k) calculated from Eq. (4.13) 180

7.10. Example of spiral path used for desired trajectory of abrasive tool 181

7.11. 3D model designed for a profiling control experiment 181

7.12. Profiling control scene along a spiral path 182

7.13. Position control result along the spiral path 182

7.14. Control result of the polishing force ||^SF(k)|| 183

7.15. Desired contact force in case of 1 Hz 184

7.16. Frequency characteristics of the force control system in case of using the elastic ball-end abrasive tool 185

7.17. Frequency characteristics of the force control system in case of using the small wood-stick tool 185

7.18. Finishing scene by using a wood-stick tool and diamond lapping paste 187

7.19. Surface of the LED lens mold before and after the finishing process. The diameter is 4 mm 188

7.20. Large-scale photo of the LED lens mold before finishing process 188

7.21. Image of the small stick-slip motion for an abrasive tool 189

7.22. Surface of the LED lens mold before and after the finishing process. The diameter is 4 mm 190

7.23. LED lens mold after the conventional finishing process 191

7.24. Undesirable small cusp marks still remain on the surface 192

7.25. Large scale photo of an LED lens mold after finishing process by using the proposed stickslip motion control 192

7.26. An example of the static relation between position and contact force in the case of a thin wood-stick tool. The tip diameter is 1 mm 193

7.27. Effective stiffness of the orthogonal-tpe robot with the thin wood-stick tool 196

7.28. Neural network for acquiring the nonlinearity of effective stiffness 197

7.29. Learning result of effective stiffness with non- linearity in press motion 197

7.30. Learning result of effective stiffness with non- linearity in unpress motion 198

7.31. Block diagram of the force feedback control law shown in Fig. 3.4 199

7.32. Detailed block diagram of the NN-based effective stiffness estimator shown in Fig. 7.31 199

7.33. Control resuilt of polishing force ||^SF(k)|| 200

7.34. Effective stiffness estimated by the neural networks illustrated in Fig. 7.32 200

7.35. Skilled worker’s hand lapping of an LED lens cavity with multiple concave areas to be finished 203

7.36. Automated lapping of a concave area on LED lens cavity 204

7.37. LED lens cavity after the lapping process 204

7.38. Finished surface of a concave area on an LED lens cavity, in which the controller shown in Fig. 7.32 is used 204

7.39. Simple idea for realizing an automatic tool truing 205

7.40. Truing of a wood-stick tool and its desired shape of contour 206

7.41. Software flowchart of the orthogonal-type robot for realizing a long-time automatic lapping process, in which on-line tool truing function is incorporated 207

7.42. Force input device with three-DOFs 208

7.43. CAD/CAM-based position/force controller with the force input device to manually regulate the desired polishing force F_d(k) 210

7.44. Control result of polishing force ||F(k)||, in which the desired polishing force F_d(k) is regulated by an operator with the force input device 211

7.45. Dialog window for viewing the variation of each component of force vector f(k) =[f_x(k) f_y(k) f_z(k)]^T 212