This "mechanical claw" can easily grab eggs, cut paper, and clamp chips, just like a human hand!

If a robotic hand could perform the same functions as a human hand, it would have a high degree of dexterity in performing tasks.

However, it is challenging to develop integrated robotic hands that do not require additional actuation components while maintaining important human-like functions such as dexterity and gripping strength. Actuation components make these hands difficult to integrate into existing robotic arms, limiting their widespread applicability.

Now, a new solution has arrived. A research team from South Korea has developed a dexterous anthropomorphic robotic hand with integrated link drive based on the link drive mechanism, called ILDA. This robotic hand has 15 degrees of freedom (20 joints), 34N fingertip force, compact size (maximum length: 218 mm), no additional parts required, low weight of 1.1 kg and tactile sensing capabilities.

Interestingly, the robot hand can be mounted directly on existing commercial robotic arms to perform a wide variety of tasks, from grasping eggs to using scissors and tweezers, the paper reports in Nature Communications.

Figure｜ILDA robot hand finger flexibility demonstration (Source: Nature Communications)

The robot solution complements its strengths

In order to achieve effective grasping movements, the industry has actually developed many relatively dexterous anthropomorphic robotic hands that can adaptively grasp some objects. In this paper, the researchers focused on analyzing and developing multi-degree-of-freedom hands with high dexterity. The representative core elements of dexterous robotic hands are divided into the following aspects: direct motor drive, tendon drive, and connecting rod drive mechanism.

The hand developed based on the motor direct drive mechanism is a common structure that can intuitively position the motor relative to the joint and drive the joint directly or using gears or timing pulleys. This structure can have high joint drive efficiency and easily arrange the joints in the desired position.

Figure | MPL robotic arm developed by Johns Hopkins APL Laboratory (Source: Johns Hopkins University)

As a specific example, the MPL v2.0 robot hand developed by the Johns Hopkins APL laboratory has shown high dexterity, active with 22 degrees of freedom, and a compact design. It integrates actuators and electronics and is capable of natural human-level movement and tactile feedback. However, the size and performance of the hand are highly dependent on the motor, especially the fingers. The use of high-end motors or drive transmission components will increase costs. In addition, due to the weight of the motor, the inertia at the fingers is high, so a complex control mechanism is required. Without innovation in actuator technology, it is difficult to achieve compactness, lightness, and high performance.

Figure｜Shadow Robot robotic arm (Source: Shadow Robot)

Hands based on tendon-driven mechanisms are most similar to human hand-driven mechanisms. Typically, their actuators are located on the forearm and connected to the joints through tendons to transmit driving forces. The robotic hand developed by NASA, the David hand developed by DLR, and the Shadow dexterous hand developed by Shadow Robot can be regarded as representatives with this mechanism. This is a very suitable approach for developing a single humanoid robot, but the actuators and electrical components of this type of robot hand are quite large, making it difficult to combine these robots with many existing commercial robot arms.

Link drive mechanisms are commonly used in our daily lives. Hands developed based on this mechanism facilitate the movement of joints in the desired direction through a structure that combines multiple links to transmit power from the actuator. This type of robot hand has the advantages of bidirectional control of joints, robustness, and easy manufacturing and maintenance. However, it is difficult for them to achieve multi-degree-of-freedom movement and maintain a large workspace, especially in serial robot hands such as fingers. Tendons are thin and flexible, so each joint can be driven independently through the rotation axis, but the links are relatively thick and stiff, making this configuration difficult to achieve.

Through the analysis of existing robotic hand solutions, the researchers concluded that the robotic hand must have the following advantages: flexibility, fingertip force, controllability, robustness, low cost, low maintenance and compactness. In addition, all components should be embedded in the hand itself and include all the above functions, thereby developing an integrated link-driven dexterous anthropomorphic robotic hand (ILDA).

Figure | ILDA overview, configuration includes five robot fingers with fingertip sensors, palm side with integrated actuators, controller and accessories (Source: Nature Communications)

The new scheme is constructed by the fusion of parallel and serial mechanisms, and realizes 2-DOF motion of the metacarpophalangeal joint (MCP) and 1-DOF motion of the proximal interphalangeal joint (PIP) through a combination of connecting rods. The selection, placement and configuration of small parts that play a role in each joint can achieve the required DOF motion and drive angle, as well as an efficient power transmission structure to obtain high fingertip force and its back-driving capability. The force sensing capability of the hand is ensured by connecting a six-axis force/torque (F/T) sensor to the fingertips. Using the designed fingers, the researchers developed a five-fingered robotic hand with 15 degrees of freedom and 20 joints.

Figure｜Structure of robot fingers (Source: Nature Communications)

In practical application, it is built by solving the circuit board layout and wiring problems, ensuring the compactness of the electronics. All motors are integrated in the palm, and there are five fingers and fingertip sensors, which can be easily connected to a general robot arm through a simple connection configuration.

In the experiment, this new robotic hand was able to grasp objects of various shapes, provide strong grasping force, and ensure precision during grasping. Finally, through experiments such as cutting paper with scissors and picking up small objects with tweezers, the high utilization rate of the hand was verified, replicating the tool operations performed by human hands in daily life.

Figure | ILDA robot hand's flexible grasping ability (Source: Nature Communications)

A new level of performance

The researchers conducted some analysis on the performance of the ILDA robot hand. In terms of the link drive mechanism, the key to the design is to achieve a link-driven robot finger mechanism with 3-DOF motion similar to human fingers and a narrow finger-sized workspace to ensure the dexterity of the robot hand.

Most link-driven robot fingers only achieve 1 or 2 degrees of freedom motion dependent on two joints. Through linear displacements at three prismatic joints, the researchers developed a combination of 3-DOF motion for the finger, generating three linear displacements through a combination of rotary motors and ball screws. The three motors can simultaneously generate three degrees of freedom motion and produce high force output.

Figure | Kinematic structure of the robot finger mechanism (Source: Nature Communications)

To achieve the set target requirements, the researchers mainly considered the following factors: (1) Selection and configuration of parts of appropriate size to achieve the required degrees of freedom of movement: To realize the functions of the above kinematic model in the narrow space of the finger shape, it should be properly arranged in the configuration of the model. Therefore, it is very important to select small parts of appropriate size from a design perspective. (2) Efficient power transmission structure that minimizes friction between assembled parts. In order to obtain high fingertip force, a compact structure is required while minimizing friction in the power transmission part. (3) Easy to manufacture and assemble. In order to increase the market penetration of the developed manipulator, it is also important to evaluate it from the aspects of cost and maintenance. Therefore, it is very important to design a simple and robust manipulator structure.

Figure | Detailed dimensions of the robotic hand (Source: Nature Communications)

Finally, all power transmission components and motors were integrated into the palm side of the hand, five F/T sensors were mounted on each fingertip of the configured finger assembly, and sensor wiring was completed so it would not interfere with finger movement, resulting in an integrated robotic hand with a maximum length of 218 mm and a weight of only 1.1 kg.

To verify the performance of the ILDA hand, the researchers evaluated it from three dimensions: (1) dexterity within the workspace; (2) fingertip force; and (3) tactile perception.

In the experiment, the MCP joint can be driven from 0° to 90°, and the PIP joint can also be operated from 0° to 90°; in addition, the PIP joint can operate independently of the MCP joint, with the finger abduction and adduction angles of ±35°.

Figure｜Performance analysis (Source: Nature Communications)

The magnitude of the contact force at the contact point is determined by the fingertip sensor, and the same force is applied to the fingertip and the reference sensor. The force applied by the finger is increased in sequence, and the 25 mA current is increased every 2 seconds. The maximum force applied by this finger is 28 N in the extended position and 34 N in the bent position, verifying the accuracy of the static force applied by the finger, with an average error of 0.9 N. The responses are generally well matched, without critical errors, and have great potential for achieving force control when performing delicate tasks.

Figure | The robot performs various grasping tests and fine operations (Source: Nature Communications)

During the testing phase, the developed hand was used to crush an aluminum can, with the maximum force measured on each finger being 25 N. The hand can also be used to safely grasp an egg without crushing it. In order to confirm the feasibility of using the robotic hand to manipulate tools, the hand was connected to a commercial robotic hand to conduct a paper cutting experiment, as cutting paper with scissors in daily life is a task that requires a high degree of dexterity.

The final test involved grasping and moving a small object with the tweezers. The manipulator moved so that the tip of the tweezers could hold the small chip. The hand performed a grasping motion, causing the tweezers to peel back the chip's cover and grab the chip. Next, the object was moved to another location and the tweezers were released to complete the operation. Everything was done smoothly.

Figure | The robot picks up the chip with tweezers (Source: Nature Communications)

Lowering the threshold for commercial applications

The researchers said that the dexterous anthropomorphic manipulator ILDA based on the link drive mechanism ensures the original advantages of the link drive mechanism, such as bidirectional control of the joints, robustness, and easy manufacturing and maintenance. At the same time, it ensures 15 degrees of freedom active movement with 20 joints, sufficient working space between the fingers, and higher fingertip force. It is also lighter, compact in size, and provides space for sensor integration.

The ILDA robot hand can be easily attached to existing commercial robotic arms or robotic arms under development without the need for additional parts. The key advantage is that the hand exhibits high performance and the parts configuration is integrated with the hand itself.

This robotic hand can perform different types of grasping depending on the shape of the object. Scissors and tweezers are used to determine the possibility of manipulating tools in daily life. Although it is difficult to accurately quantify the effectiveness of the hand in manipulating tools using scissors, they use multiple degrees of freedom of the hand and perform combined movements through bidirectional control of the joints.

The development of a dexterous anthropomorphic robotic hand with ultra-high degrees of freedom remains an open problem that requires ongoing research from both a scientific and engineering perspective, but in this study, the researchers sought to maximize the performance of the robotic hand in all dimensions.

So far, the industry has developed many relatively dexterous robotic hands, but the high costs caused by complex manufacturing processes and maintenance difficulties have limited their commercial use. The applicability of this ILDA robotic hand will be able to be expanded to practical research fields and many industry applications through comprehensive optimization of functions and costs, promoting further research on robotic hands.

Can such a flexible robot hand be used on humans in the future to help disabled people complete some daily operations? We can wait and see.

References:

https://www.nature.com/articles/s41467-021-27261-0

Written by: Cooper Edited by: Kou Jianchao Layout by: Li Xuewei

Source: Academic Headlines