To truly unleash the full potential of humanoid robots, the prerequisite is to have a sufficiently good robotic hand. The video introduces several relatively advanced robotic hands one by one according to their products. The following is a summary of the content in the order of appearance. (From ICRA 2025 International Conference on Robotics and Automation)
Aidin Robotics from South Korea: 15 degrees of freedom+fingertip force sensing
The robot hand developed by Aidin Robotics has 15 degrees of freedom, with each finger having 3 degrees of freedom. The fingertips also integrate force sensing function, and the overall weight is about 1.3 kilograms. During the demonstration, the force exerted on the fingertips will be displayed in real-time in the ROS simulation environment, and the simulated force magnitude will increase with the increase of actual pressing force. From the screen, it can be seen that the manual operation of the robot is very smooth, with fingers that can both deflect left and right, as well as bend up and down. The entire hand is remotely operated through a data glove, and the user wears the glove to control it. The robot hand moves synchronously with the glove.
UK Shadow Robot: 24 degrees of freedom high-end dexterous hand
The second model is Shadow Robot's dexterous hand, which is widely regarded as a typical high-end platform in the industry. The hand has 24 degrees of freedom and is independently driven by 20 motors. The fingertip is equipped with a force sensor, allowing users to observe the force changes when the fingertip comes into contact with an object in real time on the screen. This function is crucial for operating various objects, as the operator needs to accurately grasp the magnitude of the applied force. This dexterous hand is compatible with development environments such as ROS, Python, and C++, with a total weight of approximately 4.3 kilograms. It uses the EtherCAT communication protocol and supports multiple operating modes such as position control and torque control. Its working voltage is about 48V, current is about 2.5A, and it can operate stably at a control frequency of 1kHz, making it particularly suitable for fine operations and high-precision force control tasks that require high bandwidth response.
Singapore · Sharpa: 22 degrees of freedom+visual follow+scissors Paper Cuttings
The third one is a robotic hand developed by Sharpa company. This hand has 22 degrees of freedom, supports Python, C++, and ROS development environments, and integrates force sensing functionality. In the presentation video, the scene of tracking human hand movements through a camera and synchronously controlling the robot hand to follow in real time was first presented. Then the picture of the hand manipulating the scissors to complete the Paper Cuttings task was demonstrated. Its thumb showed a high degree of dexterity and could perform complex actions relying on fine control, which showed that its thumb design and overall operating performance were very suitable for application scenarios requiring high-precision manipulation.
Alt Bionics: A Lightweight Basic Gripper with 6 Degrees of Freedom from the United States
Next, we will introduce the robotic hand launched by Alt Bionics. This hand has 6 degrees of freedom (6 DoF) and weighs approximately 500 grams. It is controlled remotely by the user and has a good response speed. Unlike the previously introduced high degree of freedom dexterous hand, its four fingers each have only 1 degree of freedom, while the thumb has 2 degrees of freedom. Additionally, some finger movements are designed to be coupled, requiring multiple fingers to move together to complete specific actions. Due to relatively few degrees of freedom and motion coupling, the overall dexterity of this hand is limited, making it more suitable for performing basic grasping tasks rather than complex and fine operations. The demonstration video also demonstrated its application limitations when some finger functions are limited.
Kobe Robotics from Japan: 6 actuators+Ethernet+fingertip force sensing
Next up is Kobe Robotics' robotic arm. The hand is driven by 6 actuators, communicates via Ethernet, is compatible with ROS and Python development environments, and integrates force sensors at the fingertips. Users can control their hands through the data glove, which focuses on engineering and scientific research application scenarios. Its core advantage lies in the effective combination of programmable control capabilities and force feedback functions.
China · Yinshi Technology: An entry-level hand with 6 degrees of freedom and clear parameters
Then there is the robotic hand from Inspur Technology. This hand also has 6 degrees of freedom, weighs approximately 540 grams, and typical operating parameters are 24V voltage and 0.2A current. Similar to the low degree of freedom models mentioned earlier, due to only having 6 degrees of freedom, its finger flexibility is relatively limited, making it more suitable as an end effector for performing simple grasping tasks rather than a highly agile operating platform. In terms of dexterity, the hand did not show particularly outstanding advantages.
Wonik Robotics Allegro from South Korea: 16 degrees of freedom+optical force sensing
The Allegro hand introduced in the video is launched by Wonik Robotics. This hand has been publicly demonstrated and used by Meta, with 16 degrees of freedom and a four finger structure, with each finger having 4 degrees of freedom. The fingertip adopts force sensing technology based on optical principles - when the pressing force increases, the light signal change of the fingertip will be significantly enhanced, thus intuitively reflecting the force state. This design makes the hand very suitable for high-sensitivity tactile experiments and human-computer interaction tasks, while also providing good safety for contact operations during robot hand movement.
China · Aoyi: 11 degrees of freedom+visual driven remote operation
Next is the robotic hand developed by Aoyi Company. This hand has 11 degrees of freedom and is equipped with 6 active actuators. It weighs approximately 540 grams and supports ROS. It uses RS-485 and CAN bus for communication, with typical operating parameters of 24V voltage and 0.25A current. In the demonstration, its movements are driven by capturing the hand posture of a real person through a camera, which belongs to a visual based teleoperation method. From the actual scene, it can be seen that the flexibility of fingers is somewhat limited. When bending, it usually manifests as the entire finger moving simultaneously, which belongs to a typical overall bending structure. A single joint lacks independent control ability, so it is not suitable as a platform for ultra-high dexterity operation.
Portugal · Seed Robotics: 19 degrees of freedom+independent fingers+impact resistant design
Afterwards, Seed Robotics introduced a robotic hand. This hand has 19 degrees of freedom, equipped with 8 active drive channels, supports ROS and Python, and communicates with the controller through a bus. The demonstration screen shows that the wrist can move in the front back and left and right directions, and each finger can be independently controlled, which brings high flexibility to multi finger collaborative operation. The base adopts a decoupled/detachable structural design, which can absorb the impact and reduce the risk of damage when the hand collides with a rigid object such as a desktop through structural buffering. In addition, it also offers multiple preset interaction modes, such as simulating the experience of shaking hands with robots. The force sensing accuracy of this hand is very high, and it can even sense slight contact between paper and fingertips. Overall, it achieves a good balance in dexterity, tactile accuracy, and impact resistance robustness.
Italy · Prensilia: 11 degree of freedom robotic hand and prosthetic solution
Next is Prensilia's robotic hand. The company's product line includes a robotic hand with 11 degrees of freedom and a model specifically designed for prosthetic applications. Both of these hands have fingertip force sensing function, and in the demonstration, it can be observed that the sensor signal changes in real time when the thumb is pressed. In addition, they also integrate physical buttons on the arms, which can switch between different preset action modes, allowing users to directly trigger corresponding grasping or posture actions. This design is particularly suitable for practical and wearable scenarios that require high operational simplicity and output stability.
Psionic, USA: 6 brushless motors+3D haptic+Mujoco simulation
Finally, we will introduce Psionic's robotic hand. The hand is internally driven by 6 brushless DC motors, which can operate in position, speed, or torque control modes and support communication through I ? C and UART. It is compatible with Python and C++programming, and can complete modeling and simulation in the Mujoco simulation environment. Its sensing system integrates 3D tactile sensors, with a typical operating voltage of about 12V and a current of about 74mA, and overall low power consumption. Thanks to the friendliness and scalability of the software and hardware interfaces, this robotic hand is highly suitable as a validation platform for grasping/manipulation algorithms, as well as an experimental tool for tactile interaction research.
Summary: Degrees of freedom, force perception, and interface determine the upper limit
Overall, the core message of this video is that the performance limits of humanoid robots largely depend on the degrees of freedom in their "hands," the precision of force/tactile sensing, and the capabilities of their control systems. The video covers a diverse range of robotic hands across various technical approaches—from highly dexterous hands like Aidin, Shadow, Sharpa, and Seed, which feature full-finger independent control and fine force regulation, to simplified designs such as Alt Bionics and Inspire with only 6 degrees of freedom suitable for basic grasping tasks. It also includes specialized robotic hands like Wanick, Aiyi, Prensilia, and Psionic, designed for specific applications (e.g., visual teleoperation, prosthetics, and tactile research), encompassing nearly all mainstream products from scientific research and teaching demonstrations to engineering implementations. For selecting or comparing dexterous hands, the key parameters highlighted in this video—degrees of freedom, weight, power supply specifications, communication interfaces, software ecosystem support, and force/tactile sensing configurations—are the most critical information worth extracting and tabulated for systematic analysis.
