This humanoid robot learnt to walk with the help of AI and reinforcement learning, and it made numerous changes on its own. Here are all of the specifics.
With the aid of AI and reinforcement learning, this humanoid robot learned to walk on its own.
Since the dawn of robotics, humanoid robots that would live among us without posing a threat to mankind have been a far-fetched dream. Thanks to a number of advancements, robotics firms have been able to build high-end robots for a variety of applications over the years. We now have a pair of robotic legs that can walk independently.
Cassie, created by researchers at the University of California, Berkeley, is essentially a pair of robotic legs without a torso. It seems creepy at first glance, but when you see it trying to walk by trial and error, it resembles newborn learning to walk for the first time (at least in terms of movements).
What distinguishes Cassie the Robot from other robots?
You might argue that we've already seen robots walk in today's world. Robots like Boston Dynamics' Spot and Atlas have received a lot of exposure on the internet as a result of a viral video released earlier this year. So, walking robots aren't a joke anymore! So, what distinguishes Cassie from the rest of the pack?
In robotics, choreographing an orchestrated sequence of movements proved to be much more effective than teaching a robot to walk on its own. In Boston Dynamics' robot dance video, we saw the robots perform in a tiny room inside a sophisticated laboratory.
As one would imagine, robotics experts spent a lot of time fine-tuning those dance steps in the robots.
Consider what it would be like if the robots were taught to dance independently. That would have been a tough egg to break, wouldn't it?
What were the researchers' methods for cracking the code?
What were the researchers' methods for cracking the code?
In Cassie's case, the researchers used reinforcement learning to teach the robot how to walk independently. It is used by researchers to teach an AI's complex behaviour using a trial-and-error process
Cassie used the technique to perform a series of movements, including crouching walking, and walking from the ground up with an unpredictable box.
Cassie was educated by the researchers in two levels of virtual environments. Initially, they used a large library of robot movements to teach a robotic clone of Cassie to walk on her own.
The simulation was then transferred to a new virtual environment. The second simulation setting, SimMechanics, essentially replicates real-world dynamics to a high degree of accuracy.
After the simulated prototype of the robot was able to walk in SimMechanics, the researchers developed the walking model into the actual robot.
After being dressed, Cassie was able to learn to walk on her own without any further alterations. The robotic legs were able to walk on slippery and rocky ground, carry unforeseen loads, and prevent collapsing during exercise while being dragged.
Cassie managed to avoid dropping during the exercises, despite the fact that two motors in its right leg were damaged.
Cassie's project is only in the early stages of growth, but this all sounds intriguing. Experts in robotics from Stanford University, Imperial College in London, and Zhongyu Li, who worked on Cassie with his team, agree that these are the first steps toward creating advanced humanoid robots that can seamlessly integrate into future human societies.
To help their work, Newsmusk allows writers to use primary sources. White papers, government data, initial reporting, and interviews with industry experts are only a few examples. Where relevant, we also cite original research from others respected publishers.
source: techstory.in
留言