Freddo the robot is strutting his stuff at Vsim, a British start-up nestled in the heart of Cambridge. This isn’t just any run-of-the-mill robot; he recently broke Usain Bolt’s 100m sprint record! Imagine that! As Freddo walks across the office, confidently grabbing a plastic bottle from a staff member, it’s clear that the future of robot training is not just bright, it’s thriving.
The brains behind this exciting venture are founders Michelle Lu and Kier Storey. They’re on a mission to develop software that will empower robots to navigate our homes and workplaces seamlessly. Storey notes, “The underlying algorithms we’ve been using for robotic simulations date back to the 70s and 80s, but they don’t really mesh well with modern GPUs.” After a year and a half, they’ve crafted a high-performance simulator that can predict about a second ahead, analyzing an impressive 20,000 potential outcomes… outside of the robot’s immediate actions!
Vsim isn’t alone in this race. Nvidia, the tech giant, is at the forefront of AI and robotics, boasting hundreds of engineers dedicated to creating software that allows organizations to train and control robots. Their suite includes virtual simulation systems and a world model known as Cosmos, which equips robots with an understanding of real-world physics and the dynamics of their changing environment. But let’s be real, even with Nvidia’s muscle, it only provides a basic grasp of reality. Storey emphasizes, “When it comes to manipulation—like grabbing a bottle and pouring it out—progress is definitely being made.”
In a twist of innovation, Nvidia has started deploying AI agents to help design these virtual training environments, which has effectively expanded their workforce. “We’re just throwing agents at it,” says Lu, who highlights the user-friendliness of their system, making it accessible for research groups and smaller start-ups. “If your simulator is fast, you can mimic hundreds of millions of scenarios in the brief moments your robot is deciding how to move.”
Now, let’s not forget the challenges. Some tasks—like manipulating highly deformable objects or cutting—are still tricky to simulate accurately. But this isn’t stopping their momentum. Vsim is gearing up to introduce a second robot, Nacho, which will aid in refining their technology and ensuring compatibility across various machines. And who can blame Freddo for wanting some company?
So, what’s next for these pioneers in robotics? Will they be able to conquer the complexities of real-world tasks? Only time will tell, but one thing is for sure: the robot revolution is just getting started…
Kaynak: Orijinal Haber