Development of multi-fingered universal robot hand with torque limiter mechanism

A multi-fingered universal robot hand has been developed in order to construct the platform of humanoid hand study. We also have developed a small and five-fingered robot hand. The robot hand is designed to protect the small driving system from a large external force. This protection mechanism is small enough to be installed in the joint driving mechanism and adaptable enough to deal with various load. This paper describes basic and unique specifications of the robot hand, and the effectiveness is confirmed by fundamental experiments.

In this paper, a new robot hand is developed on resulting knowledge for advancing our study. Universal Robot Hand II has actuators, transmission gears, reduction gears, and Torque Limiter Mechanisms in the fingers. Using the Torque Limiter Mechanisms, the fingers can sustain overload not by the gears but by the structure. This is the imitative behavior of a human finger. This paper describes that new small robot hand mechanism has five fingers at the first. At the second, this Torque Limiter Mechanism is introduced. At the third, the effects of Torque Limiter Mechanism are verified in experiments. At the last, results of these experiments are summarized and concluded.

Specifications of developed robot hand 2.1 Basic design
This section describes the basic design of developed robot hand. Fig. 2 (left) shows "Universal Robot Hand II". The height between the lower limit of the palm and the upper limit of the middle finger is 290mm. The width of the robot hand opened up between the thumb and the little finger is 416mm. The size of this robot hand is a little larger than human hand. Thus, this size is enough to imitate human hand workings. The weight of the robot thumb is 0.262kg, the weight of the every other finger is 0.250kg, and the total weight of the robot hand without the pedestal is 1.323kg.

Fig. 2. Universal robot hand II and configuration of DOFs
This robot hand has 16 DOFs. Thumb has four DOFs (the IP, the MP, the CM1 and the CM2 joints), and the other fingers have three DOFs (the PIP, the MP1 and the MP2 joint). Every DIP joint is interlocked with the PIP joint. These DOFs and the movable directions of joints are shown in Fig. 2 (right). This robot hand has the multi-axis force/torque sensors in every fingertip and tactile sensors on every finger pad. The multi-axis force/torque sensor is able to measure the force and torque at fingertip. This sensor in every fingertip is as shown in Fig.3 (upper right). Tactile sensor is able to measure the pressure distribution on the finger pad. This sensor on every finger pad is as shown in Fig. 3 (lower right). The overview of the control system for this robot hand is shown in Fig. 4. This control computer gets the pulse from the encoders in every motor, the value from multi-axis force/torque sensors in every fingertip and the pressure distribution from the array-type tactile sensors on every finger pad. The fingers are controlled through driver circuits according to these data.

Basic performance
It is shown that the basic performance of the developed robot hand. The movable range of joints is as shown in Table 1

Superior function
Typically, the robot finger is classified into a hard finger and an elastic finger. In the hard finger, the rotation of the actuator responds plainly to the angle of the joint. In the elastic finger, the fingertip can be moved with elastic members depending on the external force. However, a human finger acts as both a hard finger and an elastic finger depending on a situation. Thus, Torque Limiter Mechanism is fitted into the joint of this Universal Robot Hand II. With this mechanism, driving mechanism in joints is started to skid from setup skidding torque. By implementation with this mechanism, the driving mechanism can be protected against overload, and the robot hand may grasp objects flexibly.

Mechanism
Torque Limiter Mechanism is constructed by a fixed plate, a rotating plate and rollers held between these plates as shown in Fig. 6. These rollers are tilted on an angle of  degrees. The where,  is the coefficient of the friction between rollers, plates. r is the radius of rollers, and P is the pressure by the adjustment nut. Every 20 joints of this robot hand have this mechanism as shown in Fig. 7.  In other word, this finger is able to be adjusted as a hard finger or a passive finger.

Advantages in finger behavior
Torque Limiter Mechanism has some advantages on the operation of the Universal Robot Hand II. The behavior of the finger with the mechanism is shown in Fig. 9 and Fig. 10. The PIP joint and the DIP joint are normally located as shown in Fig. 9 (a). The DIP joint is flexed in conjunction with PIP joint as same degrees. It is thought that excess overload is operated at the distal phalanx. As shown in Fig. 9 (b). If torque by the external force exceeds setup skidding torque, modules of drive gearing are turned over toward the direction of fending off the force to the mechanical stopper as one structure. This mechanism doesn't only make the actuators to be protected against the overload, but also support the external force mechanistically without output power of actuators by the position of particularity. Flexible grasp with Torque Limiter Mechanism is as shown by Fig. 10. Typically, a robot finger takes the form in Fig. 10 (a) in case of grasping a thin object. This is because that the commonly-used robot hand has engaged DIP and PIP joints in imitation of human joints. On the other hand, in case of grasping a thin object with human fingers, these fingertips are collimated, and increases area of contact between these finger pads. Thus, developed robot hand operates skidding mechanism in the DIP joint. Robot fingertips are collimated, and increases area of contact in Fig. 10 (b). Herewith, developed robot hand doesn't pinch a thin object with a point contact but with plane contact. In addition, the adjustment nut of PIP is clenched up and the similar experiment without Torque Limiter Mechanism is conducted for comparison.

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.

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.

Experimental setup
In this section, this gap is compensated in the following equation.
where, f is the fingertip force and F threshold is the threshold of the fingertip force. t is the motor torque and T threshold 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 F threshold , in other words, the fingertip contacts an object. In addition, the motor torque t is over the constant value T threshold , 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.

Experimental results
The result of this experiment is shown by Fig. 14. The red line is the fingertip force. The black line and the blue line is the value of the encoder in MP1. The black line is the compensated data, and the blue line is the raw data. As shown in Fig. 14, the impact force is measured. This has a reason that the DIP and PIP joints of the finger extended to the stopper and the skidding mechanism of the DIP and PIP joints are inactive. Thus, the impact force is not alleviated.

Fig. 14. Transition of fingertip force contacted against rigid object
The angular difference is 1.5 [deg.] between the first tap and the second tap in the compensated data. The angular difference is 0.5[deg.] between the second and the third. After multi-cycle experiments, the angular difference is 2 [deg.] at a maximum by one tap. This compensating method has cumulative difference but is practical. Depending on the desired accuracy of control system, the values of encoders should be reset with a bump against the stopper. The results show that this compensating method is effective.

Conclusions
In this paper, it is declared that the multi-fingered universal robot hand is developed. This robot hand is named "Universal Robot Hand." This robot hand has 5 fingers, 20 joints and 16 DOFs. This robot hand is a little bigger than a human hand. Every DOF is driven by the DC motor in the finger. Every joint has Torque Limiter Mechanism. This mechanism is the 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 software method.

Acknowledgment
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   This book provides state of the art scientific and engineering research findings and developments in the field of humanoid robotics and its applications. It is expected that humanoids will change the way we interact with machines, and will have the ability to blend perfectly into an environment already designed for humans. The book contains chapters that aim to discover the future abilities of humanoid robots by presenting a variety of integrated research in various scientific and engineering fields, such as locomotion, perception, adaptive behavior, human-robot interaction, neuroscience and machine learning. The book is designed to be accessible and practical, with an emphasis on useful information to those working in the fields of robotics, cognitive science, artificial intelligence, computational methods and other fields of science directly or indirectly related to the development and usage of future humanoid robots. The editor of the book has extensive R&D experience, patents, and publications in the area of humanoid robotics, and his experience is reflected in editing the content of the book.