Human-like robots occupy a particular niche in the psyche of not only America but the world at large.
After all, from comic books to Hollywood screenplays, robots, friend and foe alike, have firmly anchored works of fiction in the realm of the future and fantasy.
And while today, companies from Ocado to Amazon are exploring and increasingly using autonomous and automated solutions within the realm of fulfillment operations, the reality is that safe, reliable robots Building and deploying systems remains difficult and very expensive.
A major challenge for companies considering integrating warehouse automation is often the investment in a fundamental redesign of their current facilities and building an entirely new technology architecture to support robotic fulfillment solutions. investment is required.
To complicate matters, these same facilities are often leased rather than owned for industry-standard three- to five-year terms, rather narrowing the scope of ROI for automation efforts. is.
That's why humanoid robots that can be inserted into existing fulfillment processes present an opportunity. You don't need to redesign your warehouse to increase efficiency by enhancing and accelerating product sorting, retrieval, and delivery workflows, but you do need training. And it has historically been an inevitable obstacle to its development and adoption.
This is a pioneering technology that researchers at Google DeepMind are using to advance the field of robotics and show how future robots can learn new tasks by watching video tutorials, such as simple task-based video games. It includes.
read more: Humanoid robots take away the burden of warehouse automation
What is needed to operate robots on a large scale?
Google's Genie, announced last month (February 23), is described as “the first generative interactive environment trained in an unsupervised manner from unlabeled internet videos.”
According to the company's blog post, the Genie model allows users to learn what movements a robotic arm can perform to manipulate objects and how to control it simply by watching a video of the same real robotic arm in action. He said he was able to learn. The researchers noted that this could mean that future robots could learn new tasks by watching video tutorials.
Other future-friendly use cases cited by the researchers include coding the robot's central machine learning (ML) and artificial intelligence (AI) platforms to resemble video games, allowing it to move through 3D space and perform various tasks. It may be possible to train robots to perform specific tasks.
“AI allows robots to better understand their surroundings and better detect objects and people they encounter,” says a University of Chicago computer director who directs Human Robot Interaction (HRI). said Sarah Sebo, assistant professor of science. ) Lab said in an interview with PYMNTS.
This approach to training is particularly valuable in unstructured fulfillment and warehouse distribution environments, and is especially effective for humanoid bipedal robotic systems.
As reported last Monday (March 4), Amazon has already deployed a humanoid device called Digit, developed by a company called Agility Robotics, to move boxes in one of its warehouses.
The company recently announced (February 28) that its Innovation Fund is ramping up investments in robotics.
read more: Excitement and money flow at the intersection of AI and robotics
Simplify your work with fulfillment robots
Industrial and stationary robots have been around for decades, but artificial intelligence (AI) has the potential to give robots more freedom of movement. AI allows robots to do more than just memorize pre-programmed routines, his AI vision software for robotic baggage handling is his CEO and co-founder of Plus One Robotics, his solution. said Eric Nieves in his PYMNTS interview. AI allows robots to detect changes in their surroundings and use algorithms to make decisions on the fly.
“One of the keys to this achievement is 3D sensor technology that gives robots a deeper understanding of their environment,” Nieves said. “Robots are now not only operating in predictable use cases; with minimal guidance, they can intelligently recognize and adapt to new situations.”
The realm of opportunity is vast. GreyOrange CEO Akash Gupta told PYMNTS in January that only about 10-15% of warehouses have mechanized at least some processes, and the bulk of fulfillment optimization is still being captured. That means they haven't.
