It's not every day that I stumble upon an innovation that completely upends my understanding of a particular type of technology. However, when it comes to this IoT offering, I am excited.
German startup COMPREDICT builds data-driven virtual sensors that replace hardware sensors across vehicles.
They are purely software-based and can be deployed in the cloud or embedded in any vehicle: scooters, passenger cars, trucks, buses, and other heavy vehicles.
I spoke to Stefan Werner, head of global sales at COMPREDICT, to find out more.
What the heck is a digital sensor?
If you're a fellow IoT aficionado, this is your first question. And well, this is where it gets interesting.
A quick reminder, hardware sensors in connected devices use physical or chemical effects to measure attributes like temperature, pressure, weight, distance, force, torque, speed, etc.
Virtual sensors are software-based. They use machine learning algorithms to precisely estimate the targeted physical quantity from other already available signals.
COMPREDICT's virtual sensors use the data from a vehicle's CAN-bus (Controller Area Network) to provide deep insights into the behaviour of the vehicle's subsystems like steering, drivetrain, tires, brakes, battery, electronics and more.
This includes health and usage monitoring and paves the way for predictive maintenance management.
Stefan Werner explained:
"Extracting data from the CAN-bus combined with machine learning algorithms provides an environment where a software sensor can replace the hardware centre.
The virtual sensor itself is something tested, redone, and redeveloped. We made it faster, with less code. But we also proved that the sensor replaces the hardware sensor with 95% accuracy."
COMPREDICT's virtual sensors can:
- Reduce the bill of material (BOM) costs.
- Increase measurement capabilities (without the need to add hardware sensors.
- Monitor vehicle component health and usage in the field – this will be a game changer for Level 4 and 5 autonomous vehicles.
- Provide predictive maintenance detection by forecasting wear and tear, ageing and failure.
These are big solutions to industry pain points, and from the get-go the company has captured industry attention, with early investors including tire makers Michelin and Flexbus.
Breaking the vehicular ceiling
The company, founded in 2016, spun out of thesis work from its founders at the University of Darmstadt with the collective culmination across the team of over 20 scientific publications focusing on durability engineering and modelling of vehicle dynamics and 10 patents and patent applications
I wanted to know how the company was able to embed itself so successfully across the vehicle sector, which is traditionally slow to innovate, and startups can find themselves in a loop of pilots that never transcend to commercial product offerings.
According to Werner, rather than simply presenting the tech to a company,
"Automotive manufacturers were asking us how we could help them to reduce their bill of materials. Because it becomes extremely compelling commercially, if you guys want to replace $50 of hardware sensors within the vehicle, and gain better insights, we can do that with software."
Digital sensors not only save the cost of purchasing a physical sensor but also the service costs over a lifetime.
Werned shared: "We've been seeing sensors that fail after a year or have to be replaced every three years in the vehicle."
Virtual sensors have a data-gathering digital twin, which can be used for the test beds in vehicle makers' simulations. This can potentially reduce the number of testing vehicles on the road.
With these capacities:
"the manufacturer could learn about the driving patterns, the driving behaviour, of its clients or customers, and also about the wear and tear components in the vehicle on a mass scale, not only on ten testing vehicles on the streets."
Furthermore, a fleet owner could present health certificates to customers in a subscription model.
Fleets also benefit from predictive maintenance capabilities such as anticipating the need for tire replacements, purchasing tires at scale, and then presenting them to the customer.
Earlier this month, the company announced a partnership with AI software company Palantir to provide virtual sensors and other services to partners worldwide for better cloud-based solutions for connected vehicles.
Wait, you said Blackberry?
Today BlackBerry is a global provider of automotive software, with its QNX technology used by 24 of the top 25 EV OEMs and embedded in more than 215 million vehicles worldwide.
The BlackBerry IVY platform, built in partnership with AWS, simplifies COMPREDICT's ability to manage large-scale OEM deployments of its in-vehicle predictive maintenance and diagnostics algorithms.
This makes the COMPREDICT hardware and OS platform agnostic while reducing connectivity costs for the automaker.
Blackberry brings the company down to the embedded environment, which is a big deal for growth capacity across a myriad of use cases.
Furthermore, this is just the beginning for the company, with plans to use virtual sensors to facilitate continuous improvement and enhancement with new software functionalities and over-the-air updates. This company is one to watch.