NVIDIA founder and CEO Jensen Huang recently outlined how robotics simulation offers virtual training and programming that can leverage physics-based digital representations of environments, robots, machines, objects and other assets.
While robots can be used to serve lattes and flip hamburgers, they also are valuable in packaging food, moving items in warehouses and in motor vehicle assembly, according to a June 29 NVIDIA blog post. Through robotics modeling, the machines learn to master skills that can impact a number of industries as simulators and software produce information used to develop machine learning models applicable to robots.
“Car companies employ nearly 14 million people,” NVIDIA CEO Jensen Huang said during a GTC keynote address, according to the blog. “Digitalization will enhance the industry’s efficiency, productivity and speed.”
To refine a robot, the developers construct digital representations of the object, locations and other items that a robot could potentially encounter in a virtual setting, which can adhere to physical rules and duplicate physical properties including illumination, friction, materials and gravity, the blog reported.
Moreover, the NVIDIA blog post detailed how robots enhance large scale operations, and the robotic simulations are valuable to some of the biggest companies in the robotics industry. These simulations help ensure fulfillment operations can handle the volume of tens of millions of shipments daily, including Amazon Robotics’ fulfillment center.
BMW uses robotic simulations for the planning of its vehicle assembly, other automakers are using robotics to hone their operations and Soft Robotics uses the technology to fine-tune its food packaging processes, the blog post reported.
A robotics simulator uses basic physics equations, such as Newton’s equations of motion to determine how items move, and it can detail some of a robot’s limitations as a result of its design with joints and hinges, according to the blog post
The simulators reportedly use a range of techniques to detect possible collisions between items, locate the points of contact and calculate the distances while outlining the forces needed to ensure they don’t collide, the blog reported. The company’s website also noted the simulations can also decipher sensor signals from users and specify the number of steps needed to repeat a movement, while other simulators can offer visual simulators that are output at timesteps along the way, such as the NVIDIA Isaac Sim, an application built on NVIDIA Omniverse.
