Competitive physical human

Training a robot in a simulation that allows it lớn rethành viên how khổng lồ get out of sticky situations lets it traverse difficult terrain more smoothly in real life.

Bạn đang xem: Competitive physical human

Joonho Lee at ETH Zurich in Switzerlvà & his colleagues trained a neural network algorithm, which was designed to control a four-legged robot, in a simulated environment similar to a video clip game, which was full of hills, steps and stairs.

The researchers told the algorithm which direction it should be trying khổng lồ move sầu in, as well limiting how quickly it could turn, reflecting the capabilities of the real robot. They then started the algorithm making random movements in the simulation, rewarding it for moving in the right way và penalising it otherwise. By accumulating rewards, the neural network learned how khổng lồ move sầu over a variety of terrain.


Advertisement

Currently, most robots respond in real time to lớn measurements of their surroundings using preprogrammed reactions, encountering every problem for the first time. A neural network allows the robot to lớn lean from its mistakes, và by first training the algorithm in a simulation, the team is able khổng lồ limit risks and reduce costs.

Xem thêm: Sao Biển Hoạt Hình - Spongebob & Patrick


Read more: Robot dog taught itself to lớn get baông chồng up when people kiông xã it over

“We can’t bring it lớn the real terrains we want to lớn train it on because it might damage the robot, which is very expensive,” says Lee.

The researchers initially used a small neural network that was preprogrammed with knowledge about the simulated environment, enabling the algorithm to lớn learn quickly by taking inputs from virtual sensors & remembering them. They then transferred this knowledge to a large network used lớn control the real robot.

Xem thêm: Chi Tiết Cập Nhật 10.11 - Thông Tin Chi Tiết Bản Cập Nhật 10


Using this experience, the robot was able to move 0.452 metres per second on mossy land – more than twice as fast as it can move sầu with its mặc định programming.

Journal reference: Science Robotics, DOI: 10.1126/scirobotics.abc5986

More on these topics:


Advertisement
Trending Lachạy thử Video Free

Advertisement

MORE FROM NEW SCIENTIST



Puzzle #116: Can you figure out all the scores for the 1991 season?



Eye-tracking software could make video calls feel more lifelike


Technology

Astronomers have sầu created the largest ever map of dark matter


Physics

Drones may have sầu attacked humans fully autonomously for the first time


Technology

Sign up to our newsletters

Enter your tin nhắn address to lớn get started

Get the app


FOLLOW US


Back to top

Chuyên mục: Tin Tức