Covariant AI Infused Robots in Logistics

Covariant AI Infused Robots in Logistics

In the world of robotics, getting a robot to pick up objects of varying sizes has continued to stump even the largest of tech companies the likes of Google and Amazon. In the realm of manufacturing components within a warehouse where objects for robots to pick up are similar in size, hard coding the program works for this application but in retail logistics, that’s a different ball game. An AI robotics startup located in California once called Embodied Intelligence, now called Covariant, seems to have solved this problem through machine learning algorithms and soon to include neural networks.

Through an installation in Germany, Covariant has managed to prove their robots combined with AI can tackle varying shapes and successfully pick them up. The company first trains the robots using hours worth of picking data then the robot continues its education through reinforcement learning, where its goal is to pick up ‘X’ and drop it at ‘Y’ successfully. That is not the only perk of Covariant’s robots. These robots are capable of learning to pick up new objects but when one robot learns to pick up a new object, this information will be relayed to the other Covariant robots worldwide. This system is call the “Covariant Brain”, which works through the use of neural networks.

As stated by Peter Puchwein, vice president of innovation at Knapp to The Verge:

Non-AI robots can pick around 10 percent of the products used by our customers, but the AI robot can pick around 95 to 99 percent. It’s a huge difference.

As with any budding innovative industry, Covariant is not the only company working with AI based robots, there is competition within the space. With people like Jeff Dean, Google’s head of AI; Geoffrey Hinton needs no introduction in the AI arena, known for designing machine learning algorithms and his artificial neural network work; and Yann LeCun, Facebook’s head of AI research; they can make claims about being the best within the space. They provide robots that can work 24 hours a day and pick up anything thrown at it. Google has an in house lab which has been dubbed “the arm pit” by employees in hopes to achieve such accuracy with their robotics program. Amazon holds a robotics competition yearly where robots have to stock shelves in hopes of finding robots to work within their warehouses.

This is where most individuals begin their claims that robots are going to take over our jobs. Research shows 54% of employers are facing a shortage within the picking field. With low hourly rates, at times not ideal workloads, and some say a boring line of work, employee turnover is a continuous struggle and people just don’t apply as well. After spending a couple months as a picker within a large logistics warehouse putting packages onto a conveyor, not a day goes by that the picker lifestyle is missed. This mindset is shared by many individuals. Robots hopefully will take over the picking field in logistics companies, but as of today will not be replacing humans. Instead of replacing people, robots will be taking on roles that most are reluctant to fill. 

Every year, more and more shoppers shop online. Logistics used to be simple where manufacturers shipped directly to outlets that sold their products. Now products are being sent to logistics companies to be shipped to various warehouses in order to end up on your door steps. This creates a large demand for pickers than can be supplied, as companies like Amazon continue to open distribution centers.  Having robots within warehouses provides an opportunity for better paying and more fascinating jobs. Letting the robot handle the boring process of picking up and dropping packages, it will provide roles for individuals to become and develop as technicians within the world of robotics. 

While Covariant is currently a small company within the working robots field, the promise they show through the use of machine learning and neural networks is sure to aid their growth within the world of robotics. Using reinforcement learning to help robots gain knowledge as they pick then utilizing neural networks to communicate the data to other robots, is sure to be revolutionary in the logistics industry.

Leave a Reply