Boston Dynamics releases video of its Atlas humanoid robot lifting and carrying a mini-fridge using AI-driven whole-body coordination in a laboratory demonstration
Demonstration relies on reinforcement learning trained for millions of hours in simulation.
in time people will come to understand "why humanoid"
You can’t lift a fridge with just your hands. Your whole body needs to conform to its shape, and bear the load between your arms and torso. Here, @BostonDynamics' Atlas uses proprioception to manage the whole-body interaction and adapt to a shifting 100+ lb load. Enabling this type of high performance manipulation is exactly why we walked away from what was arguably the world’s best implementation of MPC for humanoids, and shifted entirely to RL without looking back. This level of whole-body controls is a fundamental building block of physical intelligence and key to the value proposition of humanoids. More technical details in: Blog: https://bostondynamics.com/blog/training-a-humanoid-robot-for-hard-work/ Behind the scenes video: https://youtu.be/xKK5ze3FukQ
@Jason Eventually, the robots and goods & services will be free, but the dollar price of robots will be high in the beginning
We're going to have a massive surplus of extremely strong and dexterous robots to do extremely unnecessary tasks for us. I can't wait! ... predict these will be $1 an hour -- which is basically $8,760 a year. Insane. In fact, they'll probably sell them for $1 an hour with a minimum of 5,000 hours.
Great demo, living up to the name. From the blog post:
'Atlas uses reinforcement learning (RL) to learn how to lift a fridge by practicing the move with an absurdly large number of variations of the fridge in simulation. The hardest part is not seeing the fridge or knowing how to lift it, but learning to adapt to whatever version of the fridge that Atlas will encounter in the real world. This is a combined control and perception problem, where perception is done implicitly from body proprioception. The policy driving the behavior has learned to adapt to variations like the location of the fridge, its mass, the amount of grip on the ground and with the fridge, or the configuration where the fridge settles in between the torso, arms, and hands. That level of adaptation is one of the most fundamental building blocks of physical intelligence.'

Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
Boston Dynamics is hiring for RL roles.
Boston Dynamics is hiring for RL roles.
This is very similar to our work from last year https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model
So need this.
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
Atlas is truly the frontier of American engineering. A work of art, some might say ❤️
You can’t lift a fridge with just your hands. Your whole body needs to conform to its shape, and bear the load between your arms and torso. Here, @BostonDynamics' Atlas uses proprioception to manage the whole-body interaction and adapt to a shifting 100+ lb load. Enabling this type of high performance manipulation is exactly why we walked away from what was arguably the world’s best implementation of MPC for humanoids, and shifted entirely to RL without looking back. This level of whole-body controls is a fundamental building block of physical intelligence and key to the value proposition of humanoids. More technical details in: Blog: https://bostondynamics.com/blog/training-a-humanoid-robot-for-hard-work/ Behind the scenes video: https://youtu.be/xKK5ze3FukQ
Humanoids are happening, it is only a matter of time.
Everyone building a “data moat” w/ grippers in the meantime is stocking up on data with low shelf life. Not all robotics tokens are created equally.
in time people will come to understand "why humanoid"
New/mass production atlas carrying an entire mini fridge like it's nothing. Payload capacity for humanoids is often very low, when it really should be one of their defining attributes.
As an aside, BD has a long history of these gimmicky one off demos, but they've been churning them out pretty quickly lately. I think that's a good sign that the company is shortening development times and getting closer to something they can ship.
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
"Robot bring me a beer" task failed successfully
A nice summary of how humanoid robots are incredibly valuable
You can’t lift a fridge with just your hands. Your whole body needs to conform to its shape, and bear the load between your arms and torso. Here, @BostonDynamics' Atlas uses proprioception to manage the whole-body interaction and adapt to a shifting 100+ lb load. Enabling this type of high performance manipulation is exactly why we walked away from what was arguably the world’s best implementation of MPC for humanoids, and shifted entirely to RL without looking back. This level of whole-body controls is a fundamental building block of physical intelligence and key to the value proposition of humanoids. More technical details in: Blog: https://bostondynamics.com/blog/training-a-humanoid-robot-for-hard-work/ Behind the scenes video: https://youtu.be/xKK5ze3FukQ
This is very similar to our work from last year https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model
Great demo, living up to the name. From the blog post: 'Atlas uses reinforcement learning (RL) to learn how to lift a fridge by practicing the move with an absurdly large number of variations of the fridge in simulation. The hardest part is not seeing the fridge or knowing how to lift it, but learning to adapt to whatever version of the fridge that Atlas will encounter in the real world. This is a combined control and perception problem, where perception is done implicitly from body proprioception. The policy driving the behavior has learned to adapt to variations like the location of the fridge, its mass, the amount of grip on the ground and with the fridge, or the configuration where the fridge settles in between the torso, arms, and hands. That level of adaptation is one of the most fundamental building blocks of physical intelligence.'
Boston Dynamics showed Atlas lifting and carrying a 100+ lb mini-fridge, using reinforcement learning to handle weight, grip, position, and balance through body proprioception.
shows how humanoids may handle hard labor: not by seeing objects better, but by adapting through contact, body feedback, domain-randomized training, and hardware built for strength and repairability.
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
We're going to have a massive surplus of extremely strong and dexterous robots to do extremely unnecessary tasks for us.
I can't wait!
... predict these will be $1 an hour -- which is basically $8,760 a year. Insane.
In fact, they'll probably sell them for $1 an hour with a minimum of 5,000 hours.
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
@ericjang11 @chris_j_paxton It will be sad sad world if the answer is — for picking up fridges with bare hands 🤣
in time people will come to understand "why humanoid"