During the 2026 World Artificial Intelligence Conference (WAIC), the most electrifying moment at the Shanghai World Expo Exhibition and Convention Center was undoubtedly the humanoid robot table tennis match.

Presented jointly by Agibot, CHINGMU, Peking University, Beijing Academy of Artificial Intelligence (BAAI), and UC Berkeley, the HOPE AI autonomous humanoid robot table tennis exhibition match captivated the audience. As the world's first full-sized bipedal humanoid robot capable of fully autonomous decision-making in a real-time table tennis rally, it engaged visitors in unscripted, untethered, human-robot matches with no human intervention. United Nations Deputy Secretary-General Alexander De Croo also visited the exhibit to experience this milestone firsthand-witnessing embodied intelligence evolve from "being able to move" to "being able to compete-and compete well."

Behind every precisely executed swing, CHINGMU played a critical role-providing high-speed 3D trajectory capture of both the ball and the robot itself.

Alexander De Croo, UN Deputy Secretary-General, Tries It Out On-site

01｜A Human-Robot Ping-Pong Match: The Real Challenge Isn't the Swing-It's "Seeing"

For a robot, playing table tennis is far more complex than it appears.

A ping-pong ball can travel at over 10 meters per second, reaching the opponent's side in just a few hundred milliseconds-with complex spin and unpredictable trajectories. The robot must complete ball detection, 3D positioning, trajectory prediction, stroke planning, and motion control within an extremely short time window-every step demanding ultra-low latency and ultra-high precision.

The most foundational-and most critical-step is accurately capturing the real-time spatial position and motion trajectory of both the ball and the robot itself. Only by "seeing" clearly can the robot truly "hit" accurately.

As Rong Shanyu, Master's student at Peking University's School of Software and Microelectronics, explained during the on-site demonstration:

"Using CHINGMU's motion capture system to track the high-speed ball's position and velocity in real time, we can compute where, in which direction, and at what speed the robot should strike-and then feed that data to Agibot's underlying motion control system to enable high-speed human-robot table tennis rallies."

02｜CHINGMU K26 Mocap System-Equipping the "Expedition A3" with "Intelligent Vision"

As the technical partner for the HOPE AI autonomous humanoid robot table tennis exhibition match, CHINGMU deployed its K (Kunpeng) Series K26 Motion Capture System to capture the real-time motion trajectories of both the ping-pong ball and the robot.

The K26 system builds a solid perceptual foundation with two core performance advantages:

• 300fps high-frame-rate capture: Continuously capturing spatial dynamics at extremely high frequency, ensuring not a single trajectory point of the high-speed ping-pong ball is missed;

• Sub-mm positioning precision: Locking onto the precise 3D coordinates and velocity vectors of the ball and robot within milliseconds-even under complex lighting and dynamic backgrounds.

Throughout the process, CHINGMU did far more than simply "record trajectories." It delivered continuous, accurate, low-latency spatial data inputs for the robot's real-time decision-making, enabling the robot to execute the full loop:

Real-time acquisition of spatial coordinates and velocity of ball & robot → Trajectory prediction and stroke timing → Algorithm-driven autonomous decision-making → Precise swing control via chassis and arm

To put it simply: CHINGMU's K26 handles the "seeing," while the Expedition A3 handles the "hitting." Without this high-precision pair of "eyes," the robot would be unable to respond to the dynamic physical world in real time.

03｜From Table Tennis to Tennis: Unlocking More Challenging Sports Scenarios

This is not the first time CHINGMU has empowered humanoid robots to take on demanding sports challenges.

Earlier, the Galaxy General robot tennis performance-praised by CCTV-was also powered by CHINGMU's optical mocap technology, providing end-to-end high-precision spatial positioning. Whether tracking large-court tennis bounces or capturing rapid-fire table tennis volleys, CHINGMU's mocap solution consistently demonstrated exceptional spatial tracking stability and high-frequency responsiveness for both the ball and the robot.

From tennis to table tennis, from single-sport demonstrations to more complex real-time human-robot interactions, each step places higher demands on robots-high-speed targets, continuous decision-making, real-time feedback, and intricate human-robot interaction-all testing the full capability of the Embodied AI system. CHINGMU is continuously equipping robots with high-precision 3D spatial perception, serving as a critical bridge between the physical world and intelligent robotic decision-making, and pushing the boundaries of robot movement in dynamic real-world environments.

Making Robots More Like Humans

When a robot can autonomously judge, react quickly, and execute stable high-speed movements, what we're witnessing is no longer just an impressive "performance"-it's a robot moving like a human: running across the court, making independent decisions, and striking with precision.

As a global leader in motion capture, CHINGMU remains committed to the path of 100% independently developed and fully controllable optical mocap technology. Beyond real-time competitive play, CHINGMU's technology has been deeply integrated into core areas including robot performance evaluation, teleoperation, dexterous hand manipulation, and multimodal data acquisition. Looking ahead, CHINGMU will continue to partner with leading enterprises and research institutions to empower robots with more complex, flexible, and human-like motion capabilities-helping robots truly step out of the lab and into the vast landscape of real-world applications.