Led by ETH Zurich robotics professor Marco Hutter, the team’s machine learning technology, based on a neural network, allows the robot to combine its visual perception of the environment with its sense of touch (proprioception) for the first time, based on direct leg contact.
The above video shows how ANYmal can move through a series of challenging environments, including the path up the 1,098m high Mount Etzel at the southern end of Lake Zurich.
Overcoming obstacles such as slippery ground, high steps and forest trails, the quadrupedal robot from ETH Zurich’s Robotic Systems Lab navigated the 120 vertical metres ‘effortlessly’ in a 31 minute hike, the team said — four minutes faster than the estimated duration for human hikers — with no falls or missteps.
Published in Science Robotics, the team’s findings could make ANYmal suitable for use in situations too dangerous for humans or too impassable for other robots, such as forest fires or earthquake aftermath.
To navigate difficult terrain, humans and animals quite automatically combine the visual perception of their environment with the proprioception of their legs and hands. This allows them to easily handle slippery or soft ground and move confidently, even in low visibility. Until now, legged robots’ ability to do this has been limited.
“The reason is that the information about the immediate environment recorded by laser sensors and cameras is often incomplete and ambiguous,” explained Takahiro Miki, a doctoral student in Hutter’s group and lead author of the study.
For example, tall grass, shallow puddles or snow appear as insurmountable obstacles or are partially invisible. In addition, the robot’s view can be obscured in the field by difficult lighting conditions, dust or fog.
“That’s why robots like ANYmal have to be able to decide for themselves when to trust the visual perception of their environment and move forward briskly, and when it is better to proceed cautiously and with small steps,” Miki said. “And that’s the big challenge.”
ANYmal, developed by ETH Zurich researchers and commercialised by ETH spin-off ANYbotics, was exposed to numerous obstacles and error sources in a virtual training camp before facing real-world challenges.
This let the network learn the ideal way for the robot to overcome obstacles, as well as when it can rely on environmental data and when it would do better to ignore it.
According to Hutter, this training allowed the robot to master difficult natural terrain without having seen it before, even if the sensor data in the environment is ambiguous.
The researchers won a $2m first prize in September 2021 in the DARPA Subterranean Challenge robotics competition, demonstrating the robot’s efficiency and capability to autonomously explore an underground system of narrow tunnels, caves and urban infrastructure.
英國鐵路公司如何推動凈零排放
It would be better if the trains had good coverage of the country. Large areas have no easy connection and so cars (or buses?) and lorries are still...