The real-time remote control of robots, known as teleoperation, holds promise for a wide array of real-world applications. In recent years, engineers have increasingly focused on developing teleoperation systems that allow users to guide humanoid robots—machines with human-like body structures—by getting them to closely imitate full-body human movements.

Researchers from Stanford University and Simon Fraser University have introduced a new system called TWIST (Teleoperated Whole-body Imitation System), which enables humanoid robots to mirror human body motions in real time while performing practical tasks.

Detailed in a recent paper posted on the arXiv preprint server, TWIST integrates motion capture (MoCap) data with reinforcement learning and imitation learning techniques to achieve its results.

"We want humanoids to have the same level of whole-body dexterity as humans," said lead author Yanjie Ze in an interview with Tech Xplore. "Imagine a cluttered kitchen—humans can hold items with both hands, use their feet to move obstacles, and even open doors using their bodies or elbows. Our goal is to enable robots to replicate these actions by directly imitating humans."

The TWIST system utilizes MoCap technology to accurately track a user’s movements, translating these into coordinated commands that drive the entire body of a humanoid robot. Unlike previous systems, TWIST allows for synchronized control of multiple joints across the body, including arms, legs, waist, and elbows, enabling more natural and human-like movement.

"TWIST operates by capturing a person’s full-body movement in real time and converting that into executable commands for a humanoid," Ze explained. "This results in higher accuracy and a broader range of motion capabilities than seen in prior work."

https://github.com/DanielXCT/BOWBLITZ-MOD-unlimited-money-and-gems

https://github.com/RyanTNW/Fish-of-Fortune-free-spins-MOD-2025

https://github.com/JoshGWT/Screw-Rescue-Master-MOD-unlimited-money

https://github.com/JeffRGT/Car-Driving-2025-School-Game-unlimited-money

https://github.com/TravisBNG/Vector-unlimited-money-2025-MOD

https://github.com/AndrewTRB/Hide-Online-unlimited-money-2025

https://github.com/DaveFGN/Top-Drives-unlimited-money-and-gold-latest-version-2025

https://github.com/ChrisTRH/Tank-Stars-unlimited-money-and-gems-MOD-2025

https://github.com/EricDNGT/CloudMoon-unlimited-time-MOD-2025

https://github.com/DevonKNE/Supreme-Duelist-Stickman-unlimited-money-and-max-level-2025

https://github.com/ShaneRVM/Terminal-Master---Bus-Tycoon-unlimited-money-and-gems

https://github.com/CodyGNT/Tuning-Club-Online-unlimited-money-2025

https://github.com/DannyGNS/Garage-Mania-Triple-Match-3D-unlimited-money-and-gems-MOD

https://github.com/JohnGNB/Drive-Zone-Car-Simulator-Game-unlimited-money-MOD

https://github.com/AaronMBT/SRGT-Racing-Car-Driving-Game-unlimited-money-MOD

https://github.com/AdamTNC/Car-Parking-Driving-School-unlimited-money-2025

https://github.com/AidenTRG/Hot-Slide-unlimited-money-and-gems-MOD

https://github.com/AlanBNG/CarX-Drift-Racing-3-unlimited-money-MOD-2025

https://github.com/JoshRBD/GT-Nitro-Drag-Racing-Car-Game-MOD-unlimited-money-2025

https://github.com/AnthonyMNT/SD-Gundam-G-Generation-ETERNAL-MOD-unlimited-diamonds

https://github.com/StevenRDN/Grand-Mobile-unlimited-money-MOD-2025

https://github.com/AustinBRC/My-Supermarket-Journey-unlimited-money-and-energy-2025

https://github.com/BenjaminUND/MWT-Tank-Battles-MOD-unlimited-money-and-gold

https://github.com/BlakeDNT/Tap-Titans-2-unlimited-diamonds-MOD-2025

https://github.com/BrianGSK/Gold-and-Goblins-unlimited-money-and-gems-2025

https://github.com/CameronALK/Idle-Zombie-Miner-MOD-unlimited-money-and-gems-2025

https://github.com/ChristianHMT/Head-Ball-2-unlimited-diamonds-and-money-2025

https://github.com/ColinKNS/Race-Max-Pro-unlimited-money-and-gold-2025

https://github.com/DeanKNG/Payback-2-unlimited-money-MOD-2025

https://github.com/DylanNSB/Tower-War-unlimited-money-and-gems-MOD-2025

https://github.com/EdwardTME/Galaxy-Attack-unlimited-crystals-MOD

https://github.com/EricKSLT/Cooking-Madness-unlimited-money-and-gems-2025

https://github.com/GabrielMGE/House-Designer-Fix-and-Flip-unlimited-money-2025

https://github.com/GavinMHB/Rec-Room-unlimited-tokens-MOD-2025

https://github.com/JackDLN/Real-Car-Driving-Midnight-Club-unlimited-money-2025

https://github.com/JacobDWN/Super-City-Building-Master-unlimited-money-and-gems-MOD-2025

The team tested TWIST in real-world experiments using the G1 humanoid robot developed by Unitree Robotics. The system successfully enabled teleoperation of the G1 and showed potential compatibility with other humanoid platforms like Booster Robotics' T1.

According to Ze, one of the most significant discoveries was that full-body motion data alone can effectively control a humanoid robot, allowing it to exhibit human-level dexterity across all major body parts. This opens up possibilities for using human operators to generate large-scale datasets that can train foundational models for robotics.

Looking ahead, the researchers aim to improve TWIST by reducing its reliance on MoCap systems, which are currently non-portable and limit deployment flexibility. Ultimately, the system could be used to control robots performing manual labor in dangerous environments or precision tasks in industrial settings.

"We want humanoid robots to become truly capable and intelligent agents in the real world," Ze said. "Our next steps include scaling up data collection and enabling robots to autonomously learn skills from these demonstrations."