GTC 2017 | Not just Tesla V100, what can the Nvidia Isaac robot simulator mentioned by Huang do?

According to Nvidia, the Isaac robot simulator uses advanced video game and graphics technology to simulate real-world conditions to train intelligent machines before deployment. The Nvidia Jetson platform can accelerate the construction of machines. Huang Renxun said that robots from more than 50 companies around the world currently use this platform to process complex data from terminal devices to complete important tasks such as search and rescue, as well as tedious or potentially dangerous tasks such as elderly assistance and industrial automation.

“AI-powered robots have the potential to dramatically improve people’s lives, but building and training them is incredibly challenging,” Huang said. “Nvidia is revolutionizing the robotics industry by leveraging its deep expertise in simulating the real world to train robots more accurately, safely and quickly.”

Isaac Robotics Simulator: Provides virtual environment for training and testing

Isaac robot simulator (Image source: Leifeng.com)

Leifeng.com previously reported , "The name Isaac comes from two people: one is Isaac Newton, who symbolizes physics, and the other is Isaac Asimov, who symbolizes AI. This is the ultimate robot AI virtual training environment created by Nvidia." From this, it can be seen that Huang Renxun wants to use the GTC conference to clearly inform the world that Nvidia will completely transform into an AI company.

It is reported that before a robot is put into actual deployment, it must be fully trained and tested. If a physical prototype is used every time, the cost is not only high but also impractical. Creating a complete environment that can interact with the robot may not be safe or very complicated, and modeling all possible interactions between the robot and its surrounding environment is also very time-consuming.

In response to this situation, Nvidia released the Isaac robot simulator. According to Huang Renxun:

Isaac's goal is to create an "alternate universe" that obeys the laws of physics. The virtual training space will fully obey the laws of physics in the real world, except for time.

The Isaac robot simulator is built on an enhanced version of Epic Games Unreal Engine 4, using simulation, rendering and deep learning technologies to provide a virtual environment for robots to train and test AI. In the virtual environment it builds, developers can build various test scenarios through deep learning training, simulate them in minutes, and apply the knowledge gained to the real unit after the simulation training.

Jetson Platform: Accelerating Machine Building

In addition, at the GTC conference, Huang Renxun also introduced the application of the Nvidia Jetson platform. At present, companies using this platform are all over the world, including Fellow, which develops industrial robots, Starship, Marble and Dispatch.AI, which develop delivery robots, Enroute Lab and Aerialtronics, which develop search and rescue drones, Toyota, which develops life-assisting robots, and the American FIRST non-profit organization, which focuses on the development of high school student robots.

At the same time, Nvidia also said that in order to accelerate the development of advanced robots based on Jetson, its partners are also working on launching open source reference platforms for drones, submersible drones, wheeled robots and other devices. These platforms will help developers save costs and expenses, create models and modules, and so on.