What happens if you put a robot in a baby carrier and use it to control humanoid robots? CHILD, the University of Illinois's most curious project, revolutionizing robotic teleoperation. Instead of complicated exoskeletons or expensive VR controllers, simply wear this fanny pack and manipulate its joints to control robots remotely.
The inspiration comes from the parent-child relationship: when you carry a baby in a baby carrier, their movements are naturally limited to yours. The same principle, applied to robots. And does it work? A lot.
How the CHILD robotic baby carrier works
The system CHILD (Controller for Humanoid Imitation and Live Demonstration) is the idea of Joohyung Kim, associate professor of electrical and computer engineering at theUniversity of Illinois at Urbana-ChampaignThe concept is disarming in its simplicity, as I wrote: a miniature robot that can be worn like a baby in a sling and that serves as a control interface for life-size humanoid robots.
The operator moves the arms, legs, and joints of the pouch robot, and the remote humanoid robot replicates the exact same movements. It's a form of robotic puppeteering that eliminates the need complex motion capture systems or cumbersome exoskeletons. Kim explains that during his time at Disney Research, he often wondered how to make creating movements for animatronic robots more intuitive. The Robotic Carrier aspires to be just that intuitive interface.
The CHILD robot can be built for less than dollars 1000 Using 3D-printable components and readily available hardware, it's incredibly more affordable than traditional exoskeletons, which can cost tens of thousands of dollars.
The baby carrier that teaches robots
The analogy with the parent-child relationship is not accidental. As the researchers explain in the published paperWhen parents naturally carry a baby in a baby carrier, the baby's range of motion is completely contained within that of the adult. This makes it intuitive to scale movements in puppeteering mode.
Noboru Myers, first author of the study together with Obin Kwon, Sankalp Yamsani e Joohyung Kim, demonstrated that the system enables full joint-level control of humanoid robots. CHILD supports both direct joint control for full-body movements and loco-manipulation, i.e., the ability to coordinate movement and object manipulation.
Teleoperation has become a popular approach to collect data and solve robotic tasks in the context of Physical and Artificial Intelligence. EmbodiedCHILD is one of the solutions we offer to allow users to provide direct commands via teleoperation for specific target tasks.
Practical applications
The CHILD system integrates with PAPRLE (Plug-And-Play Robotic Limb Environment), a robotic teleoperation pipeline for connecting various input devices such as puppets, game controllers, and VR-based interfaces. This integration enables portable, wireless control of target robots, supporting humanoid shapes and dual-arm configurations.
The tests were conducted using robots such as Unitree G1, which costs more than $16.000, demonstrating that a $1000 controller can effectively manage professional-grade robots. Adaptive force feedback is incorporated to improve the operator experience and prevent unsafe joint movements.
The design of the CHILD is completely open-source, with hardware specifications and source code available on GitHubThis approach promotes accessibility and reproducibility in robotics research, allowing other laboratories to build and improve the system.
The future of teleoperation with a pouch
Kim argues that, in the long term, humanoid robots will become more common than wheeled ones. Wheeled or vehicle-based robots currently dominate the industry and will continue to play an essential role in various industries., but if the goal is to deploy more robots in intimate human environments like our homes, the humanoid form will probably be the only viable solution.
The CHILD system is a democratic approach to advanced robotics. As I pointed out in this article on assistant robotsThe real challenge is no longer technological but social: convincing people to accept robots in their daily lives. An intuitive controller like the robotic baby carrier could be just the bridge needed between technological complexity and human usability.
The paper presentation is scheduled for the IEEE RAS HUMANOIDS Conference 2025 in Seoul, where it will be interesting to see the reaction of the international robotics community. In the meantime, anyone can download the plans and build their own pouch robot. The only thing left to do is decide which humanoid robot to have it control.
