Why the time is ripe for collaborative robotics
Impressions from Hannover Messe 2016
Collaborative robots (cobots) are lightweight robots that work next to humans and assist them in repetitive, dangerous and tedious tasks. Unlike traditional robots, which work fenced off from humans and are pre-programmed to perform one task, cobots are small and flexible. They are relatively new — first models entered the market in 2009. So far around 15,000 have been installed globally. The reason for a relatively slow adoption has been mostly their price tag and safety concerns. Recent developments in robotics hardware and machine learning software solve these challenges.
Cobots pay for themselves in a couple of months
This week’s Hannover Messe, the world`s largest industrial fair, was filled with cobots. Every large equipment manufacturer showed one. The biggest news was the launch of FRANKA EMIKA from KBee, a Munich-based company owned to 40% by KUKA. At € 9,900 this robot changes the industry cost dynamics. Its gripper, motors and gears are not top quality and its load value is relatively low, but it will likely be sufficient for quite a few applications.
The price tag of comparative robots from KUKA, Fanuc or Bosch is around €40,000 and €100,000 and the unit economics (equivalent to man-hours) around €25/h. In September 2015 a Chinese company Ningbo Techmation released a €20,000 cobot. A further 2x price reduction from KBee opens up a range of new applications.
The most popular report on the robotics market by Boston Consulting Group (2015) talks about merely a 20% cost reduction (from around € 100,000 to € 80,000) in industrial robots between 2015 and 2025! This prognosis might need to be revised.
Human safety secured with software
The second largest inhibitor to wide scale adoption of cobots have been safety issues. Here is where I think the recent development in software, and specifically deep learning, will bring key answers. The idea is that a robot is able to learn the environment around it and constantly improve on the catalog of objects and tasks it knows. The same way it will understand how humans move and anticipate unexpected behavior such as a collision. Where a camera that feeds the software can’t see other sensors will help.
Deep learning software will also improve the unit economics, adding to the already significant cost reduction in hardware. A robot that knows how to operate in new environments and how to grab new objects will of course be more adaptable and multi-functional. During its lifetime it will be able to serve more roles and needs, and pay for itself faster.
Improved economics will bring more investments
$ 1.3 bn in venture capital went into robotics last year. Drones, robotic personal companions and surgical robots emerged as the three dominant investment categories. The classical applications for robots, like manufacturing and logistics, attracted only around 10% of the total capital. As the unit economics improve, startups jump into the classical applications and investors will pay more attention.
Automotive manufacturers already offer use cases where cobots can start working tomorrow, e.g. interior installations in automotive. Mercedes sets an example to other car manufacturers. By integrating a large number of cobots into its manufacturing it can customize its production lines. Tesla and BMW are doing it too.
At a cost of around €5 per hour mass adoption of collaborative robotics is imminent.
