Unitree’s humanoid robot is learning how to understand natural instructions, recognize colors, and complete real-world tasks inside homes and hospitals.
In this demo, the robot opens drawers, identifies the correct medicine box by color, sorts clothes into a laundry basket, loads a dishwasher, closes the door, moves cushions, and even adjusts a hospital bed.
What makes this especially interesting is that the robot is not only moving objects. It is combining vision, language, memory, and physical control to understand what people mean and respond to follow-up instructions like “put this one in too” or “stop now.”
This is the direction humanoid robotics is moving toward: machines that can see their surroundings, understand everyday speech, and help with tasks that normally require human judgment.
Could robots like this become normal helpers in homes, hospitals, and care facilities?
Unitree has demonstrated a new generation of robotic home assistants capable of understanding natural language and executing complex, multi-step tasks in real-time. By integrating vision and language, these robots can navigate diverse environments like homes and hospitals, interacting with objects and responding to human commands without pre-programmed scripts.
What if your next roommate wasn't another person, but a robot that actually understood what you
were asking? Well, Unitry just showed exactly that. And honestly, this feels less like a robot demo,
and more like watching the first generation of robotic home assistants.
It starts with something incredibly simple. A person tells the robot to pick up a cushion
from the floor and place it on the sofa. No remote control. No pre-programmed sequence.
The robot listens, figures out which cushion you're talking about, walks over, grabs it,
and places it exactly where it belongs. Then comes the interesting part. The person asks,
what color is the cushion? Instead of just guessing, the robot actually looks at it
through its cameras before answering, white. That's vision and language working together
in real time. Then, the robot moves into a hospital room. It opens the correct drawer
on a medical card, counts the medicine boxes inside, identifies their colors,
picks out only the blue one, and places it on the counter. Think about that for a second.
It isn't just recognizing objects. It's following multi-step instructions while
understanding what it's seeing. Next, it heads into a bedroom. Someone throws clothes onto a bed
and asks the robot to put them into the laundry basket. The robot grabs the first shirt, drops
it into the basket, then immediately understands when the person points to another shirt and says,
put this one in too. No complicated commands. Just normal conversation.
Then, probably my favorite demo. In the kitchen, someone asks it to load a plate into the
dishwasher. The robot picks up the plate, looks inside the dishwasher, notices there's already a
glass cup in there, carefully places the plate onto the rack, and even closes the dishwasher
door afterward. That's several different actions chained together without needing step-by-step
guidance. And finally, it moves back into a hospital environment. A patient asks the robot
to lower the bed. The robot walks over, grabs the manual crank, starts turning it, and the moment
the patient says, that's enough, it immediately stops. That may sound simple, but responding
naturally to spoken commands while physically interacting with the world is one of the biggest
challenges in robotics. We're watching robots move beyond carefully controlled factory tasks
and into homes, hospitals, and everyday environments where every situation is different.
They're beginning to see, understand, and act more like human assistants than machines following
scripts. And if this is what they can already do today, just imagine what they'll be doing a few
years from now. If you want to keep up with the biggest breakthroughs in AI and robotics,
make sure to subscribe to AIx. You won't want to miss what's coming next.