Uncategorized

Atlas robots master courses almost flawlessly

The engineers at Boston Dynamics have enabled the Atlas robot to cope with a complex course in various ways. This works almost perfectly in tests. The “almost” bothers them.

The Atlas robots can easily offer predictable motion sequences. Now its developers want to take the next step. By training Atlas to move through complex obstacle course environments, Atlas learns new movements inspired by human behavior. The training pushes Atlas to its limits.

Atlas impresses on the course

The first of the two robots runs up a row of raised plywood panels, skips a gap and runs up and down stairs in the course set up on the second floor of the Boston Dynamics headquarters. The second robot jumps onto a balance beam and does the same steps in reverse order, while the first robot jumps over the beam. Both land after two synchronized backflips. It looks almost perfect.

The developers are satisfied, but not completely satisfied. During the course run, the robots repeatedly showed the smallest uncertainties and wobbles. After the robots have completed their backflips, have one of them pump his arm triumphantly like a striker after a spectacular goal. It works, but the atlas stumbles a little with this simple movement. It’s just a small stumbling step, something most people watching the video would not notice. Almost perfect is not enough for the Atlas team.

Human “form factor” optimal for many areas of application

The Atlas team wants to create a humanoid who can perform all movements precisely and perfectly. That is why it decided to use the course as a training environment. In this way, Atlas should learn to be able to perform the same range of movements and physical tasks as humans. The assumption is understandable. If a robot were ultimately able to react to its environment with the same degree of dexterity as the average adult human, the range of possible applications would be practically limitless.

Almost finished!

Please click on the link in the confirmation email to complete your registration.

Would you like more information about the newsletter? Find out more now

The Atlas team is constantly changing the control of the humanoid robot to perception. Every movement is no longer preprogrammed. Rather, Atlas actively reacts to its own perception and then selects a variant of one of the actions available to it. In this way, he can actively develop himself in practice. Currently, however, this further development still has to be manifested through code changes. Boston Dynamics has not yet launched the self-learning robot. For the time being, Skynet remains a vision of the future – fortunately.

You might be interested in that too

Leave a Reply

Your email address will not be published. Required fields are marked *