For decades, robotics has been an essential component of factories, manufacturing plants and warehouses around the world. However, these are “robots” in the simplest sense of the term – mindless machines that serve a singular purpose. Companies are now beginning to invest in autonomous robots that can think for themselves. It sounds like the beginning of a dystonian sci-fi movie, but there’s actually a very good reason why robot workers need a brain.
If you spend much time on the internet, you’ve probably come across some of the astonishing developments in artificial intelligence (AI) and robotics. Over the past few years, robots have learned to walk, talk, play musical instruments, engage in sports and even beat humans at video games – all with minimal input from their human creators.
While these are undoubtedly impressive achievements, it’s not immediately obvious how any of this benefits existing businesses – particularly in the manufacturing and packing industries where robots are predominately used. In short, why would a factory or warehouse want a robot that can make its own decisions?
At GTC 2017, this questions was posed to Jesse Clayton, senior manager of product for Nvidia’s Intelligent Machines division. Clayton provided a simple yet highly logical answer:
The thing about [current] factory robots is that they are largely limited to very fixed function capabilities. They can pick this thing up and move it over here but that’s about the extent of their capabilities. They don’t deal well with dynamic situations like changes in lighting [or being] set up in another part of the factory that it doesn’t already have an understanding of.
The benefits of autonomous robots equipped with deep learning is obvious: it means you don’t have to rebuild or replace your robotic line whenever a new task presents itself to the company. You can simply deploy your existing fleet to learn new skills and functions.
However, there is one significant hurdle to cross: AI robots take a very long time to master new tasks. As Clayton explained, it simply doesn’t make economic sense to waste thousands of hours waiting for factory robots to work out what you want them to do. Compounding the problem is the need to build complex training labs that mimic the environments they will be deployed in. It’s simply too expensive to be worthwhile.
Nvidia is attempting to solve this problem with a new robot simulator dubbed Isaac. Unveiled at GTC 2017, Isaac is a deep learning software platform that trains intelligent machines in simulated real-world conditions before they get deployed. The platform is built on an enhanced version of Epic Games’ Unreal Engine 4 and uses NVIDIA’s advanced simulation, rendering and deep learning technologies. It’s essentially a virtual reality school for autonomous robots.
Working within this virtual environment, developers can set up extensive test scenarios using deep learning training, and then simulate them in minutes — which would otherwise take months to perform. Once the desired skills have been learned and mastered in Isaac, the knowledge can be transferred into the “brains” of real-world units.
With the cost and logistics of training AI robots drastically diminished, the human factory workers of the world may want to start dusting off their resumes.