ByteWrist: A Parallel Robotic Wrist Enabling Flexible and Anthropomorphic Motion for Confined Spaces

Jiawen Tian*,
Liqun Huang,
Zhongren Cui,
Jingchao Qiao,
Jiafeng Xu,
Xiao Ma,
Zeyu Ren*

ByteDance Seed

Abstract

This research introduces ByteWrist, a novel highly-flexible and anthropomorphic parallel wrist for robotic manipulation. ByteWrist addresses the critical limitations of existing serial and parallel wrists in narrow-space operations through a compact three-stage parallel drive mechanism integrated with arc-shaped end linkages. The design achieves precise RPY (Roll-Pitch-Yaw) motion while maintaining exceptional compactness, making it particularly suitable for complex unstructured environments such as home services, medical assistance, and precision assembly. The key innovations include: (1) a nested three-stage motor-driven linkages that minimize volume while enabling independent multi-DOF control, (2) arc-shaped end linkages that optimize force transmission and expand motion range, and (3) a central supporting ball functioning as a spherical joint that enhances structural stiffness without compromising flexibility. Meanwhile, we present comprehensive kinematic modeling including forward / inverse kinematics and a numerical Jacobian solution for precise control. Empirically, we observe ByteWrist demonstrates strong performance in narrow-space maneuverability and dual-arm cooperative manipulation tasks, outperforming Kinova-based systems. Results indicate significant improvements in compactness, efficiency, and stiffness compared to traditional designs, establishing ByteWrist as a promising solution for next-generation robotic manipulation in constrained environments.

4s per circle.
2s per circle.
1s per circle.

Design and Prototype of ByteWrist

Design and Prototype of ByteWrist.
Figure 1 Design and Prototype of ByteWrist. (a) Sturcture Design of ByteWrist, (b) Prototype of ByteWrist.

ByteWrist is illustrated in Fig. 1a, which is driven by three stage motors. The output of the first-stage motor is connected to the first-stage driving linkage, which is further linked to the parallel platform via an arc-shaped end linkage. Meanwhile, the second-stage motor is mounted inside the first-stage driving linkage, its output connects to the second-stage driving linkage, which is also linked to the parallel platform through an arc-shaped end linkage. Similarly, the third-stage motor is fixed within the second-stage driving linkage, with its output attached to the third-stage driving linkage that connects to the parallel platform via an arc-shaped end linkage.

All three stage driving linkages and arc-shaped end linkages, as well as arc-shaped end linkages and parallel platform, are connected via revolute pairs. All these six revolute pairs are oriented toward the center of the parallel platform.

To enhance the stiffness of the parallel platform, a supporting ball is mounted at its center and connected to the platform via a spherical joint. By controlling the movement of three stage motors, the end parallel platform can achieve precise RPY motion.

As illustrated in Fig. 1b, the prototype of ByteWrist adopts Quasi-Direct Drive based actuators manufactured by RobStride Dynamics.

Kinematics of ByteWrist

Coordinate System Definition of ByteWrist.
Figure 2 Coordinate System Definition of ByteWrist.

The kinematics of the parallel wrist involves solving the relationship between θ1\theta_1, θ2\theta_2, θ3\theta_3 and the RPY angles of the parallel platform. For the forward kinematics part, the inputs are θ1\theta_1, θ2\theta_2, θ3\theta_3, and the outputs are RPY angles of the parallel platform.

Forward kinematics can be solved by the following system of equations.

{O1P4O1P5O1P4O1P5=cos2π3O1P4O1P6O1P4O1P6=cos2π3O1P5O1P6O1P5O1P6=cos2π3\begin{cases} \dfrac{\overrightarrow{O_1P_4} \cdot \overrightarrow{O_1P_5}}{\left| \overrightarrow{O_1P_4} \right| \cdot \left| \overrightarrow{O_1P_5} \right|} = \cos \dfrac{2\pi}{3} \\ \dfrac{\overrightarrow{O_1P_4} \cdot \overrightarrow{O_1P_6}}{\left| \overrightarrow{O_1P_4} \right| \cdot \left| \overrightarrow{O_1P_6} \right|} = \cos \dfrac{2\pi}{3} \\ \dfrac{\overrightarrow{O_1P_5} \cdot \overrightarrow{O_1P_6}}{\left| \overrightarrow{O_1P_5} \right| \cdot \left| \overrightarrow{O_1P_6} \right|} = \cos \dfrac{2\pi}{3} \end{cases}

Given that this system of equations is nonlinear, the Newton-Raphson method is adopted herein for iterative solution. This research adopts the numerical method to solve the Jacobian matrix.

Experiments

Prototype of ByteMini

To verify ByteWrist functionality, we integrate them into our 22-DoF dual-arm mobile robot, ByteMini. Key design features include:

1. Arms: 7-DoF in SRS (Spherical-Revolute-Spherical) configuration with ByteWrist as wrist modules.

2. Grippers: 1-DoF grippers paired with RealSense D405 cameras for close-range vision.

3. Waist: 1-DoF lifting mechanism in high stiffness for height adjustment.

4. Chassis: 3-DoF omnidirectional mobile platform for flexible movement.

5. Head: 2-DoF in Pitch and Yaw motion, integrated with a RealSense D457 camera for primary vision.

6. Computing and Power: Dell T3280 computer as the main controller and 4.08 kWh battery for power supply.

Description of the ByteMini Robot.
Figure 3 Description of the ByteMini Robot.

ByteWrist Motion Range

To verify the wrist range motion and high flexibility, this study designs a circular trajectory, enabling the wrist to move in accordance with following equations.

{β2+γ2=0.68α=0\begin{cases} \beta^2 + \gamma^2 = 0.68 \\ \alpha = 0 \end{cases}

As illustrated in Fig. 4(a-c), the input of wrist movement and the actual position feedback are plotted for cycles T = 4s, 2s and 1s respectively. Postures of ByteWrist at 8 distinct moments are presented in Fig. 4(d-k), each corresponding to different pitch and yaw angles. Throughout the wrist movement, the yaw direction remains constant. Experimental results demonstrate that the robotic wrist exhibits flexible motion within this operational region.

Motion Range of ByteWrist.
Figure 4 Motion Range of ByteWrist.

Confined-Space Maneuverability

To verify the flexibility of ByteWrist in confined spaces, this study designs a experiment of objects grasping inside a glove box.

Confined-Space Grasping Experiment.

Motion jitter is induced by the operator’s instructional jitter, which poses considerable challenges to operators.

Dual-Arm Manipulation of Deformable Objects

The dexterous clothes-hanging task in GR-3 imposes high requirements on the capabilities of the robot, where the robot is required to achieve dual-arm collaboration in the chest area and perform manipulation of deformable objects with high precision and high dexterity.

ByteWrist Based Dual-Arm Manipulation of Deformable Objects.

Conclusions

1. Innovative Structural Design: ByteWrist adopts a three-stage motor-driven parallel mechanism, combined with arc-shaped end linkages and a central supporting ball. This design achieves a balance between compactness and stiffness. The prototype test confirms that the wrist can stably realize continuous RPY motion, meeting the requirements of anthropomorphic manipulation in confined spaces.

2. Complete Kinematic Modeling: The forward and inverse kinematic models of ByteWrist are established. For forward kinematics, the Newton-Raphson method is used to solve the nonlinear equations, realizing the mapping from driving linkage angles θ1\theta_1, θ2\theta_2, θ3\theta_3 to parallel platform RPY angles. For inverse kinematics, the rotation matrix is derived based on given RPY angles to calculate the required driving linkage angles. Additionally, a numerical method with an optimized step size (Δθ=1×103\Delta \theta = 1 \times 10^{-3}) is proposed to solve the Jacobian matrix, providing a theoretical basis for high-precision motion control.

3. Excellent Performance Validation: Experimental results demonstrate ByteWrist can achieve high- dynamic and large-angle motion. Compared with the Kinova dual-arm robot with serial wrists, ByteMini (integrated with ByteWrist) features higher integration and greater flexibility. ByteMini completes 116 hours of data collection for dexterous clothes-hanging tasks and realizes fully automated opera- tion, verifying ByteWrist’s ability to cooperate with dual arms for high-precision deformable object manipulation.

Citation

@misc{bytedance2025bytewrist,
      title={ByteWrist: A Parallel Robotic Wrist Enabling Flexible and Anthropomorphic Motion for Confined Spaces}, 
      author={Jiawen Tian, Liqun Huang, Zhongren Cui, Jingchao Qiao, Jiafeng Xu, Xiao Ma, Zeyu Ren},
      year={2025},
      eprint={2509.18084},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2509.18084}, 
}