A System For Real-Time Control Of Humanoid Robots

The system lets human-shaped robots move like people in real time, helping them do tasks in places that are hard to reach or unsafe.

Controlling robots remotely in real time, known as teleoperation, can be useful in many real-world situations. Some engineers have recently been working on systems that let users guide humanoid robots, which have a human-like body, to copy full-body movements accurately.

Researchers at Stanford University and Simon Fraser University have developed a new system called TWIST (teleoperated whole-body imitation system). It allows humanoid robots to closely follow a human’s full-body movements in real time and complete various real-world tasks.

– Advertisement –

The goal is to give humanoid robots the same level of full-body control as humans. In everyday situations, such as working in a messy kitchen, people naturally use both hands, feet, and even other parts of their bodies—like elbows or sides—to interact with their environment. The aim is for humanoid robots to perform in a similar way by directly imitating human movements.

The team tested their teleoperation system through a series of real-world experiments using the G1 humanoid robot developed by Unitree Robotics. The results showed that the system successfully enabled remote control of the robot and could also be adapted for use with other humanoid robots, such as the T1 model by Booster Robotics.

The key finding was that full-body human motion data is effective for controlling humanoid robots, including their legs, feet, waist, knees, and elbows. The system allows these body parts to move in a way that closely matches human motion, giving robots human-like, full-body coordination. One practical use of this capability is to have humans control humanoid robots to collect large-scale motion data for training robotic foundation models.

In the future, the TWIST system could be improved by reducing its dependence on motion capture systems, which are not portable and limit its use. With further development, the system could be deployed in real-world environments, such as enabling robots to perform manual tasks in hazardous areas or to automate industrial processes that demand high precision.

The next step in development involves expanding data collection and enabling robots to learn and perform tasks autonomously, aiming to make humanoid robots more capable and effective in real-world applications.