The Gala Humanoids Weren’t Just a Show — They Were a Market Preview

Unitree Robotics spar with martial artists during the gala. (CCTV)

China Media Group’s (CMG) Spring Festival Gala 2026 — China’s annual televised new-year extravaganza with the most viewers — included robotics in its kung-fu choreography. The humanoid robots on stage came from four Chinese startup companies: Unitree, MagicLab, Galbot, and Noetix. These robots appeared in multiple performances, becoming key technological highlights of this year’s show.

One of the biggest challenges for humanoid robots performing at the Spring Festival Gala is that the show is live-stream — there is no room for accidents. Motion control, human interaction, balance algorithms, and countless other details must be thoroughly tested and verified before these robots step onto such a high-profile stage in front of a global audience. A critical focus is how these robots “see” the world: without reliable environmental perception, they can’t detect hazards, respond to unexpected changes, or execute safe emergency maneuvers in real time.

Humanoid robots rely heavily on sensing systems to understand their surroundings and maintain stable motion. Today, these systems typically combine cameras, LiDAR, and sometimes radar, paired with essential onboard proprioceptive sensors. Cameras are the most commonly used: they’re cost-effective and high-resolution. In simple terms, cameras capture images and algorithms estimate depth; this depth information supports obstacle avoidance, grasping, and foot placement. However, camera performance can degrade under harsh lighting, reflections, low light, smoke, or occlusion.

To mitigate these limitations, some systems incorporate radar. Radar performs better in challenging outdoor conditions — such as rain, fog, and glare. They can also detect objects at longer ranges. However, its drawback is lower spatial resolution than cameras, making it less suitable for fine-grained perception. In simple words, it can detect object but cannot see them clearly in details.

This is where LiDAR stands out: Unlike cameras, it doesn’t care much about lighting — it detects distance directly and builds a clean 3D map of what’s around the robot. For example, if the stage lights are too bright, and at the same time the floor is shiny, reflecting the stage lights, a camera-based system will have a high possibility to lose details or maybe misjudge the depth because they will be affected by lights and reflections which create many blind areas.

When radar is used in the same situation, it can still detect objects without being affected as much by stage lighting and visual interference. However, radar often struggles to identify exactly what an object is or define its edges precisely. In a real world situation, for example, if two dancers are moving very close together during a formation change, radar might interpret them as a single target, making it difficult for the robot to decide how to navigate on stage. So Radar can‘t provide detailed information needed. That’s why radar is usually more of a supporting sensor, while cameras and LiDAR do most of the heavy lifting for accurate perception and safe movement.

Finally, this is where LiDAR really helps. LiDAR actively measures distance and builds a real-time 3D map of the stage — almost like drawing a geometric blueprint of everything around the robot. With this map, robots can reliably track how far it is from the stage edge, a moving performer, or even when lighting is harsh or the background is visually blind. Some LiDAR systems, such as FMCW lidar can also measure motion and speed of objects, helping the robot understand how fast something is approaching and when to slow down, reroute, or stop. In a live show, that combination of accurate mapping, motion awareness, and strong resistance to visual interference can be the difference between a flawless performance and a safety risk.