In the world of technology, the evolution of humanoid robots continues rapidly. Figure AI has developed a new artificial intelligence -backed controller that makes the humanoid robot Figure 02 more natural. The company created an artificial intelligence system that learns to walk like a human being, using Reinforcement Learning – RL) to enable Figure 02 robots to exhibit a more natural walk.

It can be defined as an artificial intelligence approach that enables a model of artificial intelligence to optimize a model of artificial intelligence through trial and error based on rewards and penalties. Using this method, Figure taught Figure 02 robots to walk like a human. In this way, robots can learn about years of data in the simulation environment in just a few hours.

From simulation to real life

The reinforcement learning process enables the robots to learn the stylistic features necessary to make the walk -in -like -like – as synchronizing the collars with leg movements. This process helps robots adopt the closest way to human walking.

Those learned in simulation can be transferred to the real world with zero intervention. Figure says that by randomly changing the physical properties of the robot in simulation, the robot enables him to learn by modeling various scenarios that he may encounter in the real world.

Figure 02 is primarily designed for use in industrial and production areas. While these robots help to eliminate labor deficiencies, they aim to reduce the number of people working in dangerous jobs. However, we are in a period when humans will not be limited to factories, but started to enter our homes.

In this area, a large number of companies such as Tesla, Figure, Agility Robotics, 1x, Apptrronics and Unitree operate. In addition, Nvidia’s big investment in robot technology makes the GR00T N1 and Omniverse platform to use it. Figure also introduced a new machine learning model for humanoid robots in recent weeks. Helix came to the fore as a Vision-Linguage-Action (VLA) model.