News
News Categories

Conventional robot training not fast enough? NVIDIA’s robot simulator, Isaac, will make it faster and safer!

By Vijay Anand & John Law - on 11 May 2017, 1:01am

NVIDIA’s Isaac trains intelligent machines to tackle real-world situations in a safe and fast manner

During his keynote at GTC 2017, Jen-Hsun Huang, CEO of NVIDIA, made several announcements pertaining to the advancement of A.I., and machine learning. Of the announcements made, Huang also introduced Isaac, the company’s first robot simulator.

Typically, robots are currently trained via reinforcement learning to sense the world, learn from it based on feedback from its environment and take appropriate action to better reach the intended goal. However, NVIDIA put forth the notion that training in this manner could take a long time, more so if the robot in question has to perform a very complicated task such as performing surgery to shadow a surgeon's skill? This is something that cannot be trained readily by conventional methods and putting someone's life at risk. For this, we need to rely on an alternative universe, and thus Isaac the robot simulator was born. 

How does Isaac work, you ask? In a nutshell, robots training robots and evolving them in a virtual environment. Learning in a simulated environment and iterating that virtually is the key to a safer and faster AI evolution. One doesn’t have to wait to train in the physical world, which could be time consuming and dangerous.

Technically, Isaac utilizes sophisticated video game and graphics technologies to train intelligent machines and the accompanying artificial intelligence (A.I.) to handle real-world situations in simulated real-world conditions, all before the robot (and its AI in question) gets deployed to see real-life action.

“Robots based on A.I. hold enormous promise for improving our lives, but building and training them has posed significant challenges,” Huang said. “NVIDIA is now revolutionizing the robotics industry by applying our deep expertise in simulating the real world so that robots can be trained more precisely, more safely and more rapidly.”

Specs-wise, Isaac is built on an enhanced version of Epic Games’ Unreal Engine 4 (UE4) and NVIDIA’s advanced simulation, rendering, and deep learning technologies. These technologies take just minutes to simulate a performance, compared to other simulators that would take months to conduct the same simulation.

Besides Isaac, NVIDIA also took the chance to reiterate the proposition of its Jetson platform. The platform was designed by the company in order to enable fast, efficient processing of complex data of edge devices. As such, it allows these aforementioned edge devices to be capable of handling critical tasks, such as search and rescue, elder support, and industrial automation of tedious or potentially dangerous tasks.

If you’re already in possession of one of the company’s Jetson platform, you’ll be happy to know that NVIDIA partners are releasing open-source reference platforms that will include fabrication plans for drones, submersibles, robots on wheels and other devices. These partners include:

  • Fellow – Industrial robots
  • Starship, Marble, and Dispatch.AI – delivery robots
  • Enroute Lab and Aerialtronics – Search and Rescue drones
  • Toyota – Human-support robots
  • FIRST – High School robotics programs

Join HWZ's Telegram channel here and catch all the latest tech news!
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.