According to the Zhitong Finance APP, Western Securities has released a research report stating that the motion control of humanoid robots is the core technology for achieving dynamic gait, fine operations, and environmental adaptation. It also involves the comprehensive application of hardware devices, software tools, and development kits. The firm focuses on the research of motion control software algorithms and data training for humanoid robots, recommending attention to the following beneficial segments and targets: ① Motion control systems: GoogolTech (301510.SZ), Leadshine (002979.SZ). ② Software and simulation tools: NVIDIA Isaac Sim, PyBullet, Gazebo. ③ Motion capture devices: Xsens, Vicon, Lingyun Guang (688400.SH), NoriTech.

The main viewpoints of Western Securities are as follows:

Humanoid robots have wide applications, with both opportunities and challenges

Humanoid robots refer to robots that have human-like forms and functions. As the industry continues to develop, humanoid robots with highly anthropomorphic appearances, strong perception capabilities, and intelligent decision-making and learning abilities are expected to widely enter various ecological scenarios, such as industrial automation, medical rehabilitation, service and retail, hazardous rescue, home services, and educational research. With continuous influx of capital and talent, and rapid iteration of robot control and AI technologies, the humanoid robot industry is evolving rapidly. However, from the perspective of large-scale commercialization, the humanoid robot industry still faces many challenges in technology, economy, and society.

Precise complex motion control technology is the foundation for the widespread application of humanoid robots

Motion control of humanoid robots refers to the precise control of robots to simulate and execute various human actions and movements. This control includes not only precise control of the robot's joints but also planning and controlling the overall motion trajectory of the robot to ensure that it can complete complex tasks, laying the foundation for the widespread development of robots in practical applications.

Model-based control and learning-based control methods complement each other, aiding the iterative upgrade of humanoid robot motion control

At the software algorithm level, the motion control of humanoid robots can be divided into model-based control and data-driven control, each with its unique advantages to solve different control problems. 1) Model-based control: has unique advantages such as high interpretability, but mainly relies on accurate modeling and manual parameter adjustments, making optimization in multi-rigid body planning and control complex. 2) Learning-based control: refers to using data-driven techniques to enable robots to learn motion strategies from experience. Learning-based methods are particularly suitable for environments where traditional models may struggle to capture the complexities of interacting with unstructured or unpredictable terrains. The hybrid control approach that combines both provides a foundational control framework through mathematical models while enhancing adaptability and robustness using data-driven methods, balancing control precision and adaptability, significantly improving the motion and operational capabilities of humanoid robots.

Leading companies show outstanding motion control performance, with multiple segments expected to benefit

Humanoid robot products represented by Tesla Optimus, Yushu G1, and Boston Dynamics Atlas exhibit excellent motion control capabilities, although their areas of expertise differ due to variations in hardware solutions Generally speaking, the development of software algorithms related to motion control is completed by the robot manufacturers themselves. The division of development for motion control-related hardware (such as controllers, actuators, sensors, etc.) varies significantly, with both self-research and procurement from third-party suppliers being common. Additionally, hardware related to robot training, such as motion capture equipment, remote control devices, and simulation software tools, is mostly provided by third-party suppliers or open-source platforms.

Risk Warning: The development of the humanoid robot industry is not meeting expectations, the data training effects are not meeting expectations, and the iteration of motion control algorithms is not meeting expectations