Robot Era’s humanoid XBot-L robot navigates the ancient, uneven Great Wall of China using reinforcement learning algorithms.
Soldiers of imperial China patrolled the 13,000 miles of the Great Wall to keep out nomadic tribes. Thousands of years later, humanoid robots are walking along it, expertly navigating the uneven terrain.
Chinese robotics developer Robot Era released a video showing its XBot-L humanoid prototype walking along sections of the Great Wall.
Standing at 5 feet, 4 inches tall, the XBot-L was able to walk along the weathered stone pavements, maintaining its balance and sensing the environment in front of it.
https://youtube.com/watch?v=nx2YO3twZYs%3Fenablejsapi%3D1%26origin%3Dhttps%253A%252F%252Fwww.iotworldtoday.com
The earliest sections of the Wall date back to the Qin Dynasty around 220 B.C.. Over thousands of years, the condition of the fortress’s walls has deteriorated due to time and human activities.
The XBot-L, however, successfully navigated the terrain without assistance. It practiced shadowboxing atop the wall, climbed uneven steps and even made its way through dimly lit archways in guard towers.
Photo Credit: Robot Era
The Beijing-based startup was founded in 2023 and previously open-sourced the training software it uses to power its robots, Humanoid-Gym.
Powered by reinforcement learning algorithms, the XBot-L was able to better perceive the environment around it.
The robot was able to bypass obstacles using real-time path planning and motion control, making it able to navigate the Wall’s withered path.
Related:Humanoid Robot Maker Debuts Robotic Hand for Object Manipulation
“Perceptive reinforcement learning algorithms help to strengthen the robot’s perceptive and decision-making capacity in the face of unfamiliar terrains,” said Yue Xi, Robot Era’s co-founder. “The robot thus can recognize complex road conditions and adjust its walking stance in a timely manner.”
The Great Wall video follows Robot Era’s May debut of XHand, a dexterous hand designed for humanoids to perform various object-grasping tasks.
