DexForce trains humanoid robots on virtual data to cut costs

Chinese startup DexForce is training humanoid robots using virtual data, a method it says slashes the cost of development. According to Reuters, the company’s system simulates tasks like object manipulation, adjusting variables such as arm angles and object colors in a digital environment. This approach avoids the expense of physical prototypes and real-world trial-and-error.

The strategy reflects a broader push among robotics startups to overcome the high costs of hardware training. While competitors like Maven Robotics and Apptronik have advanced real-world deployments—including Maven’s reported UK factory pilot—DexForce’s focus on simulation could offer a faster, cheaper path to commercial viability. UniX AI, another Chinese humanoid robotics firm, is scaling sales ahead of an IPO, suggesting investor appetite for the sector remains strong despite recent cooling in the hype cycle, as seen with Flourish’s pivot to a lower-cost home robot.

DexForce’s virtual training could prove particularly valuable for industrial applications, where precision and adaptability are critical. If successful, the method may help narrow the gap between early-stage prototypes and mass-market adoption. The next test will be whether the company can translate its simulated results into reliable physical performance.

Sources: msn.com