'Training' successful two-legged robot over complicated terrain

Built with the ability to travel on stone steps of different heights without falling, Berkeley robots can be used for rescue work.

Today's movable robots often have problems in uneven terrain because each of their moves will vary in height and distance.

Berkeley robot shows impressive movement on complex terrain - (Video: Dailymail).

Scientists at the University of California, USA have overcome this problem when using machine learning technology to "teach" robots to adapt to bumpy surfaces.

Complicated AI algorithms will guide the robot to move according to strict rules of foot position, so the distance and height of each step becomes quite rigid. This leads to the robot falling when walking on rough terrain because every step must be not the same to avoid falling, sliding, slowing motion.

The team overcame this difficulty by using a machine to "teach" two-legged robots how to handle more logical steps.

Their system allows robots to move more comfortably, making them more flexible than any previous robot version.

Complex terrain forms that the team created for robot testing - (Photo: University of California).

They create tests, using boulders of varying heights and randomly changed distances.

Incidentally, each robot in the test was able to move on artificial terrain easily without falling, even if they had never met the same terrain before.

They don't "know" what the terrain will be, only the next step is provided to the robot, a scenario similar to what they will face in reality.

Currently, they are completely "blind" and cannot use image data to calculate the next step. Instead, the team provides the distance and height of each stone and the AI ​​system will use complex math to calculate the next step.

Robots do not have difficulty moving on blocks of different heights - (Photo: University of California)

However, many feared that the military world would soon adopt this technology and create a killing robot capable of chasing targets on the battlefield without fear of any kind of terrain.

In contrast, the engineering team said their technology could be used in rescue and search missions when natural disasters occur. In the future, rescue robots need to move on crumbling areas and difficult terrain quickly in search of people who are trapped or injured.

They share on the blog: "They are great machines in navigating on terrain with no structure and uneven. In addition, they are much more flexible than wheeled robots. It is inherently difficult to navigate over a wide range of terrain or constantly changing height.

The ability of bipedal robots on discrete and unpredictable terrain helps them become potential candidates for space exploration and disaster response. '

•