In recently released footage from KAIST’s Dynamic Robot Control & Design Laboratory, the bipedal robot demonstrates smooth movement as it navigates various actions in real time.
Weighing 75 kilograms, the robot is powered by Physical AI, a form of artificial intelligence that enables machines to perceive and act in the physical world, helping explain its lifelike motion, according to Digital Trends.
The team combined AI training with real human movement data to make the robot move more smoothly and naturally, avoiding the stiff motions often seen in other robots.
They also used simulations that match the robot’s physical limits and applied a hybrid learning approach to further improve control and performance.
Unlike many humanoid robots that rely on off-the-shelf components, the KAIST team developed all of v0.7’s core parts in-house, including motors, gearboxes, and motor drivers, making the system technically independent, Interesting Engineering reported.
Hae-Won Park, the project’s lead developer, said the robot can run at speeds of up to 12 kilometers per hour on flat ground and climb steps higher than 30 centimeters, with further performance improvements planned.
The robot can also navigate uneven terrain using proprioception alone, without relying on visual sensors, a capability suited to industrial environments where visibility is limited.
Park said the team aims to expand the system into a full humanoid with an upper body, enabling it to handle complex industrial tasks and operate alongside humans in real-world settings.