The virtual worlds where robots are trained

The virtual worlds where robots are trained. This year Nvidia has started using AI agents to help build virtual environments to train robots and validate whether the solutions from training work or not.

What happened

Such virtual simulations are a common way to train robots. Photograph: Vive Arts and Atlas VReviewPlaying With Fire review – like Disney’s Fantasia on steroids as a virtual Yuja Wang plays amid luminous dreamworldsSouthbank Centre, London Wang’s avatar performs music from Ravel and Debussy to Prokofiev and Liszt while dizzying virtual worlds take shape around her in this virtual reality installationIt might be the first time that a soloist at the Last Night of the Proms is simultaneously giving another performance in another London venue.

Training systems that allow robots to negotiate the real world are getting more sophisticated. But the speed by which Freddo has been trained to walk, recognise the bottle and grasp it is impressive. Founders Michelle Lu and Kier Storey hope one day their software will control robots that can navigate and do useful tasks in the home and workplace. "It's a weird situation with robotics because actually the stuff that we find as humans to be incredibly difficult, like gymnastics, you can get robots to do reasonably well.

The stuff that humans are really good at, like fine dexterity, is really hard in robots," Storey says.

The wider picture

Image caption, Vsim founders Michelle Lu and Kier StoreyFreddo's skills were honed in a virtual environment, where a task can be performed in a computer simulation millions of times. "Things outside of the robot's control, like humans, animals or even other robots, could do things that require a change of strategy. Image caption, Robot training systems use virtual environments where tasks are practisedVsim is a start-up with 10 engineers working on its tech. It does not build robots, instead it has a suite of software designed to let organisations train and control robots.

That includes virtual simulation training systems and a so-called world model, external, called Cosmos, which gives a robot an understanding of the physics of the real world and how its environment might change as it moves around. "When we talk about creating the [virtual] world and actually scanning it in – a lot of that is actually manual labour. Simulation is not the only method for training robots. They can also be trained by watching human or video demonstrations.

What has been reported

Her research is focused on, external developing software and hardware that can aid robots to learn complex behaviour. But those simulated environments are still rough approximations of the real world, which limits what can be trained. Nine pieces of music conjure nine luminous dreamworlds for an audience wearing VR headsets: Disney’s Fantasia on steroids. Photograph: Pete WoodheadWorlds morph into each other with distant seismic rumbles. Explore more on these topicsClassical musicSouthbank CentreYuja WangVirtual realityInstallationreviewsShareReuse this contentMost viewedMost viewed

Figure caption, Freddo the robot in action at VsimByBen MorrisTechnology of Business EditorPublished18 September 2026, 00:00 BSTFreddo the robot walks across the office and takes a plastic bottled offered by a staff member. Given that a robot recently beat Usain Bolt's 100m sprint record, it's not the most startling achievement. It took just a few minutes to develop those skills and upload them to Freddo.

What happens next

His developers say rival systems could take days to attain such skills. I'm at Vsim, a British start-up based in Cambridge. Once the optimum solution (known as a policy) is found, it can be uploaded and used by the hardware – in this case Freddo. Tech giant Nvidia has a system called Isaac Sim which works that way – Lu and Storey both worked on an early version of it. In 2022 they decided to set up Vsim, to build the their own training system environment and other tools.

As they were starting from scratch Lu and Storey could optimise the software to exploit the powerful computer chips used in AI, known as graphics processing units or GPUs. "The underlying algorithms that we were using for most of these robotic simulations they hark back to the 1970s and 1980s, but those algorithms are not really brilliant fits for GPUs," Storey says.

The report has been compiled by The Daily Waves using information reported across bbc.co.uk, theguardian.com. Details are presented according to the information available at the time of publication and may change as authorities, organisers or other relevant parties provide updates.