While other robots are still learning how to run, jump upside down or run through obstacles, let's see how this robot learns to move on a tight, cramped surface without falling over. somersault.
In a video shared by Florida Institute of Human and Machine Awareness (IHMC) recently, scientists have successfully developed an automated step planning system for Boston Dynamics Atlas and NASA Valkyrie robots .
The system uses sensors to find the most effective way to reach a human-selected location. It can even operate on rugged and tight terrain.
Previously, IHMC often relied on manual methods, so that robots controlled people moved their steps accordingly through a special user interface. But that method makes robots move quite slowly and seem bulky. That's why IHMC decided to work out a way to help robots actively move without the need for a human hand to control.
Indeed, IHMC said that the movement based on human control is one of the reasons that Atlas robot once fell while moving in the robot challenge competition run by DARPA, the agency of the US Department of Defense held in 2015.
To overcome human error, the new system allows the operator to choose the desired moving position for the robot. But to the last step is not human control but the algorithm that navigates the robot to move through obstacles by creating a map of the steps to move to reach the desired destination.
Simulation algorithm for navigating robots on computer screens
By creating an imaginary point in space, the robot can determine where to place the moves
However, the new method will need more time to improve because now the robot is only 100% stable for flat terrain. Meanwhile, the success rate of moving through narrow terrain is only 50% and roughness is 90%.
In the future, researchers expect to increase processing speed for the algorithm, teaching robots how to navigate in a maze or other more complex terrain.
IHMC is not a new organization in the field of robot development. Earlier this year, the institute successfully developed a hydraulic-powered humanoid robot, promising to enhance the robot's power and range.
Impressive technologies help robots move firmly on rough terrain
Refer to The Next Web