Covariant: The AI Robotics Company That Just Raised $40m

The advent of industrial robotics has been ground-breaking in the world of manufacturing. After their inception in the 1960s, robots quickly flooded factories around the world due to the great benefits they brought to the production process. They were faster than us. They could carry out dangerous tasks. They were simply better. Today, robots are more effective than ever and remain the backbone of manufacturing globally. However, there still exists vast amounts of untapped potential we could derive from these machines. Incapable of carrying out processes unaided, robots are yet to become fully autonomous. But what if they could perceive, learn and adapt?

Covariant is the latest company striving to overcome these hurdles. A mere 3 years after its birth, the start-up have come out of Round 2 of funding with 67 million dollars raised and are now making waves within the robotics industry by partnering with automation giant ABB. Their aim? To produce highly robust, reliable and performant cyber physical systems. So, what’s all the fuss about?

There are many fundamental issues with traditional robots. Firstly, conventional designs lack human-like dexterity limiting accuracy in task completion. A typical model may occasionally misplace or break objects, slowing down the production process. Covariant, however, state that their systems are capable of picking and packing some 10,000 items with 99% accuracy. This is achieved through AI software which enhances the robot’s consistency. As opposed to mere repetition of pre-programmed tasks, Covariant’s machines use camera systems to capture information about individual objects. This data is then interpreted by machine learning algorithms meaning the robot is constantly getting smarter and getting better at what it does.

More impressively, Covariant powered systems are able to manipulate new objects in real time. Typically, robots are only capable of handling things they have been programmed to interact with. However, the use of reinforcement learning and neural networks allow Covariant robots to move objects they are unfamiliar with, without damaging or losing the product- the primary use for this being in the automated sorting and organising of products in a warehouse.  Additionally, all Covariant software is transferrable, meaning that when one robot learns, they all do.

In recent times, amidst the Covid-19 pandemic, this technology has become more relevant than. This crisis has exposed the fragility in supply chains globally by creating a drought of labour, halting production for many. However, with Covariant’s new alternative to human effort, companies might just get back on track sooner than we think.

Thumbnail Credit: IEEE Spectrum