The second World Humanoid Robot Games, held on August 16 in Beijing, presented a realistic simulation of firefighting and rescue operations. Twenty‑three teams competed to assess the genuine capabilities of robots in challenging, unpredictable conditions.
Real‑World Testing Environment
Unlike controlled indoor simulations, this stage was moved to an actual fire station. Robots were required to operate autonomously, performing tasks under conditions that closely mirrored real‑world firefighting scenarios. The primary goal was to evaluate how reliably robots can execute their functions when faced with unexpected variables.
“Real‑World Simulation” at a New Level
According to IT之家, this year’s scenarios were significantly more realistic than last year’s. Each robot had 30 minutes to complete three tasks:
- Identify hazardous substances.
- Close three randomly placed valves.
- Locate and extinguish a fire using a fire extinguisher.
Two types of simulated hazardous materials were randomly placed on site for the robots to detect and report via images. Afterward, the machines had to locate the valves, shut them off, and finally use the extinguisher to fully suppress the fire.
Real‑World Challenges
Human firefighters oversaw the ignition of the flames while monitoring how the robots deployed their extinguishers. The competition revealed considerable difficulties robots encounter when transitioning from a controlled lab environment to the real world. Rain, changing lighting, and outdoor conditions heavily affected visual perception and mechanical operations.
For example, a unit from UniX AI (优理奇) completed its assignment within the allotted time but operated noticeably slower than human firefighters. Its manipulator required two attempts to accurately aim at the extinguisher. Of the twelve teams that participated that day, only three managed to finish all challenges.
Hardware and Planning
Data from IT之家 indicates that rain, fluctuating illumination, and the outdoor setting negatively impact visual recognition and motion control. Any error in perception or movement planning can cause the entire task to fail. Robots must not only correctly identify objects and select appropriate actions but also precisely coordinate multiple joints while maintaining balance.
These trials also highlighted the complexity of humanoid robot hardware systems. Delicate arm movements demand high‑precision joints, while powerful actuators provide the strength, stability, and mobility needed for demanding tasks. Many missions require the combined effort of several hardware subsystems.
Some robots employed omnidirectional wheels instead of traditional bipedal legs. Multi‑joint manipulators were used for finer operations. Remote‑control remains a vital part of the humanoid robotics technology stack; several teams used VR headsets to steer the robots directly before or even during task execution, receiving real‑time footage from onboard cameras.
Future Prospects
The use of domestically developed Chinese joint modules and servo drives — covering a broad torque range from low‑force precision control to high‑force actions — was also evident. Experts note that conducting such trials in real environments yields valuable failure data that will improve perception, motion‑planning, and AI systems for future robots.
