How do you teach a robot to move objects, avoid obstacles, and perform tasks without training it countless times in a real room? One answer is to first give it a digital world to practice in.
DexForce’s Different Approach
The Chinese company DexForce, based in Shenzhen, uses synthetic data and simulations to train robots in virtual environments. According to a Reuters report on October 15, 2026, this approach aims to reduce the cost and time required to collect real-world training data.
Instead of asking a person to repeat the same movement thousands of times in front of sensors, the company can generate many virtual variations of a task.
What is “sim-to-real”?
The main challenge is transferring knowledge from simulation to the real world, known as “simulation-to-reality.” On a computer, everything seems measurable, but in practice there are unpredictable frictions, lighting conditions, objects, and human movements.
The challenge is for the robot to apply what it has learned digitally without failing as soon as its environment changes.
Why is this a concern for the industry?
Robotics requires large amounts of data and many hours of training. If training becomes more effective through simulations, it may accelerate the development of robots for manufacturing, warehouses, and other settings. This doesn’t mean that a humanoid robot is already capable of taking over any human job.
When will we see them in our homes?
Home use requires high reliability, safety around people, adaptability to constantly changing objects, and affordability. There is currently no guarantee that this particular technology has overcome all these obstacles.
What’s interesting is the training method: the robots of the future may have undergone millions of “tests” before they ever touch their first real object.
Source: Reuters, October 15, 2026. Stock photo: Gabriele Malaspina / Unsplash.