WORLD NEWS VIEWDexForce uses simulation data to train humanoid robots for real-world tasks
The Shenzhen company says virtual training could reduce the expense of gathering physical robot demonstrations.
Published 16 October 2026 (IST); sources checked 15тАУ16 October 2026.
DexForce trains robots in simulated settings
Chinese robotics startup DexForce is developing a training system in which a simulated robot repeatedly completes tasks in a virtual environment. The software varies object placement, angles and other conditions to help the machine learn skills that can later transfer to a physical humanoid robot, Reuters reported on 15 October.
Why virtual practice could lower costs
Robot makers have long relied on collecting real-world demonstrations from people and machines. That data can be costly and slow to gather. Simulation offers scale, but a skill learned on a screen must still work on hardware with real-world friction, sensing errors and unpredictable objects.
The gap between simulation and physical work
DexForce has attracted investors and says its systems are being used in industrial and service settings. The claim that virtual data will make robots broadly capable remains a commercial and technical hypothesis to be tested through reliable field performance.
Sources and verification
Read the report and check for later revisions: Primary reporting source
Photo is an archival image, not a photo of this event.