Where is there room for startup companies who can disrupt industrial technology?
I was recently pondering this problem—and indeed, I consider it a problem.
Inductive Automation did it years ago partly with a state-of-the-art technology, but mostly it was by building a company that could be profitable with a disruptive pricing model.
ThingWorx did some cool things. But it exited via acquisition by PTC, which combined it with Kepware into a new division. I’m not sure, yet, the extent of disruption that has caused.
Then I had an introduction to two CEOs of very interesting start ups. Check out yesterday’s thoughts on a new take of AI, voice applications, and maintenance.
This post is about UrsaLeo freshly off a chat with CEO John Burton. It’s a story about software, science, and Raspberry Pi.
The story begins with a Google cloud based platform which is enterprise ready (can support 500K messages a second on a mid-range server). It’s also highly secure—separate servers for user login and data which communicate via a secure certificate, so a hacker can’t get into the data side with a user name and login.
Burton told me, “Today we take in sensor data (mostly environmental), store it, display on dashboards, and make it available for download. We also have a robust events / alerts engine almost ready for rollout that will generate email, text, Slack messages and also can be used to control equipment (turn things on/off etc.). Our provisioning process includes create an account on our website, enter a serial number (type or scan), and boot the device. Certificates are installed automatically and a virtual server spun listening for data from the device.”
In the area of object recognition, they are working with a new class of AI chips (Google, NVIDIA) making it possible to run machine learning algorithms at the edge. Combine with a $30 camera and a Raspberry Pi and you have a 1-2 frames / second object recognition system that can be trained to recognize 20-30 separate objects. “We’re using this to look at nuts welded to a steel plate for a customer in the UK (are the nuts the right size, in right locations?). This is their biggest source of customer returns.”
When you combine 3 or more cameras doing object recognition you can triangulate to do 3D positioning with reasonable accuracy. This is a great edge application.
Not to be ignored is another current digital trend—Digital Twinning. For UrsaLeo, this means 3D models combined with real time data. “We have a unique way of creating the 3D models utilizing technology one of the founders developed. You can click and drag objects together while wearing a VR headset. The approach being used everywhere else is to utilize gaming technology (Unreal engine or Unity engine) which is very expensive. You need a game developer to drive them as well. So add the 3D positioning with a digital twin and we can start to show goods moving through a facility in real time.”
An Industrial Internet of Things (IIoT) project can easily run into the millions of dollars. UrsaLeo is developing its technology with a combination of low cost single board computers, cameras, and AI devices combined with inexpensive digital twinning technology. This means a mid-sized manufacturer can get a sophisticated IIoT system for maybe $30K (and then ongoing cloud charges).
“We also think it’s important to make the system as open as possible so it can be connected to any hardware and also send it’s data to other applications. We’ve defined API’s for the sensors and on the cloud side we have a REST API, also a web hook, and a web streaming API. We think this approach is going to give the big monolithic players a tough time and also open up many new customers.”
There is a lot here to digest. I can think of a large number of my contacts in the industry who would be very interested in this. Stay tuned, I’m sure.