Fire rescue simulation exposes limitations of humanoid robots in real-world scenarios

The second edition of the World Humanoid Robot Games included a fire rescue simulation competition where 23 teams tested their machines in a real-world scenario inside a fire station. Each robot had 30 minutes to identify hazardous substances, shut down randomly designated valves, and extinguish fire sources using an extinguisher. Of the 12 teams that participated in this specific test, only 3 managed to complete all challenges. Factors such as rain, lighting variations, and outdoor conditions proved to be significant obstacles to the robots' vision and mechanical precision. While UniX AI's robot succeeded in completing the tasks within the stipulated timeframe, its performance was considerably slower than that of a human firefighter, and its gripper required two attempts to properly activate the extinguisher. The results demonstrate that humanoid robots still face difficulties in visual perception, movement planning, and coordinating multiple components simultaneously in uncontrolled situations. Experts point out that the failure data collected in real-world environments is essential for improving AI and robotics systems, since any error at any stage of the process could compromise the entire mission.