Jan 10, 2026 12:00:00

'Robot Kindergarten' evolves AI-equipped robots

Pantograph , a company developing 'general intelligent robots' that learn through exploration, failure, and continuous improvement, has unveiled its concept for a 'kindergarten for robots.' By allowing robots to learn through various real-world environments, rather than just virtual training, the company aims to solve the problem of data depletion for AI training and enable continuous model improvement.

Pantograph: Building a Preschool for Robots
https://pantograph.com/blog/building-a-preschool-for-robots.html

While AI is rapidly developing and is trained on any dataset available on the internet, robotics has no such datasets and must be created from scratch, Pantograph said.

Image and video data are also important when training robots. However, for robots that directly interact with the real world, it is difficult to infer things like material properties, textures, and the feel of interaction from video. For this reason, Pantograph is creating a learning environment in which thousands of small, inexpensive robots build a world model by touching, throwing, and rubbing everything they can reach. Pantograph describes this as a 'robot kindergarten.'

Pantograph's robots are small and strong, with tracks instead of wheels for improved stability and motor efficiency. Furthermore, their design is material-efficient and easy to mass-produce, making the training phase, where data size matters, less costly and highly scalable.

Below, Pantograph's robot is pushing a sofa. Although it is a small robot, it can move objects heavier than itself, weighing about 1 kg per arm.

Also, achieving fine movements is the most difficult part of robot development. While Pantograph's robot can perform complex manipulation tasks, its gripper is designed to be simple. Below, the robot assembles wooden blocks.

The following video shows the robot inserting a USB cable, demonstrating careful and delicate movements.

These actions are performed remotely as follows:

Pantograph said, 'General intelligent robots will fundamentally change the way we work and our ability to build. This technology is fundamental to how society works, including labor, the economy, and what it means to make something. We want to use robotics to further expand the possibilities of what humans can do. We are aiming to reduce hardware costs not only to enable large-scale training, but also to enable more people to build robots and expand what they can make.'