In context
In mid-2024, humanoid robot development was accelerating, with companies pushing beyond controlled lab environments to demonstrate real-world viability. Robot Era's outdoor test on the Great Wall of China highlighted the industry's focus on robust locomotion and perception for unstructured settings.
What was reported
Robot Era, a Chinese humanoid developer, claimed a first by having its full-sized XBot-L prototype (1.65 m tall) walk on the Great Wall. The video showed the robot waving, shadowboxing, and navigating weathered stone paths and stairs.
The test emphasized adaptability to complex outdoor terrain, leveraging perceptive reinforcement learning (RL) algorithms. These allow the robot to sense its environment, maintain balance, and adjust gait in real time, unlike 'blind walking' systems that lack perception.
Co-founder Yue Xi noted the RL algorithms strengthen perceptive and decision-making capacity, enabling the robot to recognize road conditions and adjust stance. The hardest part was designing end-to-end RL that transitions from sensory input to locomotion. The event followed the unveiling of XHand, a 12-DOF dexterous hand, showing the company's investment in perception and learning.
Why it mattered
This demonstration underscored the shift toward humanoids capable of operating in unstructured, real-world environments—a key requirement for future industrial applications beyond factory floors. It also highlighted the importance of perceptive RL over purely mechanical stability, pointing to more versatile automation solutions.
“Perceptive RL algorithms help to strengthen the robot’s perceptive and decision-making capacity in the face of unfamiliar terrains.”
Source: Robotics & Automation News (roboticsandautomationnews.com) · Published 2024-06-21 · “Robot Era’s humanoid robot walks on Great Wall of China”