New AI system helps robots transfer virtual training into real-world tasks

UK researchers developed an AI training method that improves how robots transfer skills from simulation to real-world industrial tasks.

The approach uses AI-generated variations during training, helping robots adapt more reliably when conditions change outside the simulator.

By reducing the need for extensive physical testing, the method could lower cost, time, and risk in deployment.

It may be especially useful in hard-to-test sectors such as manufacturing, recycling, and hazardous operations.