**4.1 Alleviation of impact force**

#### **4.1.1 Experimental setup**

In this experiment, the protection of drive train is verified. The outline of this experiment is shown in Fig. 11. The DIP and the MP2 are fixed with a bump against the mechanical stopper. The PIP is flexed at a tilt, 45 [deg.], and the skidding mechanism is tried and enabled to operate with tuning the adjustment nut. The adjustment nut of MP1 is clenched up to the disabled angle. Drive the MP1 and contact hardly the fingertip to a rigid object. Keep sliding the PIP to a bump against the stopper of the PIP joint and fastening the fingertip on the object for a few seconds. Extend the MP1 to a bump against stopper of the MP1 joint. Values of the fingertip force and the encoder (MP1) are measured during this experiment.

Fig. 11. Impact force experiment

Development of Multi-Fingered Universal Robot Hand withTorque Limiter Mechanism 105

with the real angle in the skidding case. Thus, in this section, compensating method for this

1

where, *f* is the fingertip force and *Fthreshold* is the threshold of the fingertip force. *t* is the motor

step. The fingertip force *f* is over the constant value *Fthreshold*, in other words, the fingertip contacts an object. In addition, the motor torque *t* is over the constant value *Tthreshold*, in other words, the torque is able to operate the skidding mechanism. In this instance, as operating the skidding mechanism, the angle is not counted up (down). In other instance, the angle is

As shown in Fig. 13, the finger hits a rigid object three times from the position of 0 [deg.] using the above method. Meanwhile, the rigid object is set at 91 [deg.], and the finger is extended back to 12.5 [deg.] by measuring the experimental movie. The finger is controlled

Fig. 13. Outline of experiment about mismatch between joint angle and counted pulse

*threshold threshold*

is the angle of the joint and *i* is the control

(2)

*f F t T otherwise*

0

 & 

*i i*

In this section, this gap is compensated in the following equation.

torque and *Tthreshold* is the threshold of the torque.

gap is validated the evidence.

**4.2.1 Experimental setup** 

counted up (down).

by the time-control method.

In addition, the adjustment nut of PIP is clenched up and the similar experiment without Torque Limiter Mechanism is conducted for comparison.

#### **4.1.2 Experimental results**

The result of this experiment is shown in Fig. 12. The blue line represents the fingertip force in the case of "Torque Limiter Mechanism is active", and the black line represents the inactive case. The red line is the value of encoder in the MP1. At 130[step], the fingertip force increases drastically. There is strong evidence that the fingertip touched on a rigid object. At 600[step], the fingertip force decreases precipitously. The fingertip pulled away from the object at this time. As shown by this graph, the case with active skidding mechanism has the lower impact force than the case with inactive one. After that, the fingertip force is kept low during the joint is skidding. The fingertip force in active case converges to the force in inactive case in accordance with the joint is skidded to a bump against the stopper. The fingertip force during this period is lower than converged value. By the result of multiple experiments, the average peak of the force is 0.33 [kgf] in inactive case and 0.20 [kgf] in active case. The peak in active case is drop by about 40% from in inactive case. As identified above, this Torque Limiter Mechanism protects the finger against the accidental overload. Meanwhile, a transition at 700 [step] is impact force by a bump against the stopper at the MP1 joint.

Fig. 12. Transition of encoder contacted against rigid object

#### **4.2 Mismatch between joint angle and counted pulse**

This skidding mechanism protects the finger against the accidental overload by the experiment in Section 4.1. However, this skidding mechanism has one problem. In the case of the joint is driving and skidding, this problem must be considerable. In this case, the joint angle recognized by the encoder is different from the real joint angle. The encoder is set in every motor, and the joint angle is recognized indirectly by the number of rotations. The motor drives the joint through the skidding mechanism, and the recognized angle has a gap with the real angle in the skidding case. Thus, in this section, compensating method for this gap is validated the evidence.

### **4.2.1 Experimental setup**

104 The Future of Humanoid Robots – Research and Applications

In addition, the adjustment nut of PIP is clenched up and the similar experiment without

The result of this experiment is shown in Fig. 12. The blue line represents the fingertip force in the case of "Torque Limiter Mechanism is active", and the black line represents the inactive case. The red line is the value of encoder in the MP1. At 130[step], the fingertip force increases drastically. There is strong evidence that the fingertip touched on a rigid object. At 600[step], the fingertip force decreases precipitously. The fingertip pulled away from the object at this time. As shown by this graph, the case with active skidding mechanism has the lower impact force than the case with inactive one. After that, the fingertip force is kept low during the joint is skidding. The fingertip force in active case converges to the force in inactive case in accordance with the joint is skidded to a bump against the stopper. The fingertip force during this period is lower than converged value. By the result of multiple experiments, the average peak of the force is 0.33 [kgf] in inactive case and 0.20 [kgf] in active case. The peak in active case is drop by about 40% from in inactive case. As identified above, this Torque Limiter Mechanism protects the finger against the accidental overload. Meanwhile, a transition at 700

Torque Limiter Mechanism is conducted for comparison.

[step] is impact force by a bump against the stopper at the MP1 joint.

Fig. 12. Transition of encoder contacted against rigid object

**4.2 Mismatch between joint angle and counted pulse** 

This skidding mechanism protects the finger against the accidental overload by the experiment in Section 4.1. However, this skidding mechanism has one problem. In the case of the joint is driving and skidding, this problem must be considerable. In this case, the joint angle recognized by the encoder is different from the real joint angle. The encoder is set in every motor, and the joint angle is recognized indirectly by the number of rotations. The motor drives the joint through the skidding mechanism, and the recognized angle has a gap

**4.1.2 Experimental results** 

In this section, this gap is compensated in the following equation.

$$\theta + = \begin{cases} \quad 0 & \begin{pmatrix} f > F\_{\text{flres} \text{cloud}} \\ \& t > T\_{\text{flres} \text{shell}} \end{pmatrix} \\ \theta\_i - \theta\_{i-1} & \text{otherwise} \end{cases} \tag{2}$$

where, *f* is the fingertip force and *Fthreshold* is the threshold of the fingertip force. *t* is the motor torque and *Tthreshold* is the threshold of the torque. is the angle of the joint and *i* is the control step. The fingertip force *f* is over the constant value *Fthreshold*, in other words, the fingertip contacts an object. In addition, the motor torque *t* is over the constant value *Tthreshold*, in other words, the torque is able to operate the skidding mechanism. In this instance, as operating the skidding mechanism, the angle is not counted up (down). In other instance, the angle is counted up (down).

As shown in Fig. 13, the finger hits a rigid object three times from the position of 0 [deg.] using the above method. Meanwhile, the rigid object is set at 91 [deg.], and the finger is extended back to 12.5 [deg.] by measuring the experimental movie. The finger is controlled by the time-control method.

Fig. 13. Outline of experiment about mismatch between joint angle and counted pulse

Development of Multi-Fingered Universal Robot Hand withTorque Limiter Mechanism 107

clutch brake system. The drive mechanism in the joint can be protected against overload by using the skidding mechanism. At the skidding time, the joint angle recognized by the encoder is different from the real joint angle. However the difference can be corrected by the

This work was supported by the robot study group in the Advanced Materials Processing Institute Kinki Japan. The following researchers participate this study group: Hiroyuki Nakamoto (Hyogo Prefectural Institute of Technology), Tadashi Maeda (Maeda Precision Manufacturing Limited Kobe), Nobuaki Imamura (Hiroshima International University), Kazuhiro Sasabe (The Kansai Electric Power Co., Inc.), Hidenori Shirasawa (The

Fukui, W.; Nakamoto, H.; Kobayashi, F.; Kojima, F.; Imamura, N.; Maeda, T.; Sasabe, K.; &

pp.2225-2230, ISBN 978-1424446490, Porto, Portugal, November 3-5, 2009 Imamura, N.; Nakamura, Y.; Yamaoka, S.; Shirasawa, H. & Nakamoto, H. (2007).

Jacobsen, S. C.; Iversen, E. K.; Knutti, D. F.; Johnson, R. T.; & Biggers, K. B. (1986). Design of

Lovchic, C. S. & Diftler, M. A. (1999). The Robonaut Hand : A Dexterous Robot Hand for

Morita, T.; Iwata, H. & Sugano. S. (2000). Human Symbiotic Robot Design based on Division

Mouri, T.; Kawasaki, H.; Nishimoto, Y.; Aoki, T.; Ishigure, Y. & Tanahashi. M. (2008). Robot

*of Robotics and Mechatronics*, Vol.20 No.2, pp.280-288, ISSN 1883-8049 Nakamoto, H.; Kobayashi, F.; Imamura, N. & Shirasawa, H. (2006). Universal Robot Hand

0780358867, San Francisco, California, USA, April 24-28, 2000

1934272015, Orelando, Floroda, USA, July 16-19, 2006

Shirasawa, H. (2008). Development of Multi-Fingered Universal Robot Hand, *Proceedings of the 35th Annual Conference of the IEEE Industrial Electronics Society*,

Development of an Articulated Mechanical Hand with Enveloping Grasp Capability, *Journal of Robotics and Mechatronics*, Vol.19, No.3, pp.308-314, ISSN 1883-

the Utah/MIT Dexterous Hand, *Proceedings of the 1986 IEEE International Conference on Robotics and Automation*, pp.1520-1532, San Francisco, California, USA, April 7-

Space, *Proceedings of the 1999 IEEE International Conference on Robotics and Automation*, pp. 907-912, ISBN 978-0780351806, Detroit, Michigan, USA, May 10-15,

and Unification of Functional Requirements, *Proceedings of the 2000 IEEE International Conference on Robotics and Automation*, pp.2229-2234, ISBN 978-

Hand Imitating Disabled Person for Education/Training of Rehabilitation, *Journal* 

Equipped with Tactile and Joint Torque Sensors -Development and Experiments on Stiffness Control and Object Recognition-, *Proceedings of The 10th World Multi-Conference on Systemics, Cybernetics and Informatics*, pp.347-352, ISBN 978-

software method.

**7. References** 

8049

11, 1986

1999

**6. Acknowledgment** 

Advanced Materials Processing Institute Kinki Japan).
