Boosting sensor performance

How? Solved in Software!

Last week, we were guests at the headquarters of one of the largest global discount retailers to discuss the application of the MOOS system.

As one would expect, they had a clearly outlined strategy for store automation and digitization. They were also extremely well-informed about the technological possibilities and challenges on the path towards digitizing shelves.

Connecting Shelves as Cornerstone

They completely subscribe to a vision of an ecosystem of different sensor and analytical solutions for different application areas, e.g., measuring, predicting, and optimizing bakery schedules, capturing products in the basket/cart, creating an alert for high-risk shrinkage items (tobacco, alcohol, care, or OTC products), that jointly cover all of the relevant use-cases. So, imagine cameras, tagging, loyalty/scan apps, analytics, and other sensors and actuators (e.g., ESL or interactive displays) all chipping in. They recognize that connecting shelves will be a ‘cornerstone’ in the tech stack to recognize transactions and create wider solutions.

As a result, we could very quickly skip selling the concept or selling the benefit and go into tech directions. While intrigued by our paper-based approach to creating dynamic range, they started probing for utility, cost, durability, and ease of deployment and operation – at a (massive) scale.

As they had run various experiments with incumbents and alternatives, we could go deep into the rabbit hole of technical challenges with FSR pressure matrix sensors. How do you solve for drift? For long-term decay? For shock absorbance? For latency in stabilizing reads? For disturbances due to temperature or humidity?

What about dynamic range for stacking versus noise? How do you spread the force across cells with the seal or cover? How do you correct for the surface of cover protections? Do you need spacers or foams? And how do you optimize all of these to get a good, robust, and reliable sensor read-out and consistent performance?

Broken Record

At some point, our line of answering became a broken record. For all of these points, rather than investing in (costly and lengthy!) hardware innovations, we solve these issues in our software layer. This is far more effective.

Software can address understood and predictable issues with sensor solutions, allowing the possibility to overcome some persistent drawbacks of sensors. The process involves creating enough understanding of a factor, its interplay with others, then collecting enough data to add it to our engine as a predictive factor – and correct for it in live operations.

While we have addressed a significant chunk of factors in this way, at the same time, the list of application and situational factors is also growing. For instance, recognizing fallen or misplaced products. While we haven’t addressed this yet, the path towards doing so is very clear. More important than having a nice approach to boost the current performance of sensors, we’re on a path of continuous improvement.

Next to a quicker cycle time, the approach of ‘solving in software’ creates a basis for continuous improvement. We simply push an update to our sensors to make them a bit better or a bit more versatile.

The Next Frontier

To unlock the next levels of performance or cost-effectiveness, some constraints of the physical world need to be addressed. The physical properties, e.g., operating conditions or response range, of a sensor won’t change, no matter how much software you throw at it.

Truly tapping into the interplay between hardware and software requires a more fundamental understanding of what can be solved in hardware, what in software, or what in the unique combination. For instance, a path toward modularity may require a hardware click, link, or docking solution and a software part of recognizing the neighboring sensors. There are many examples like these. Designing or redesigning sensor components with this in mind can unlock a pathway towards cheaper, better, and more flexible systems.

At MOOS, we have the approach of ‘solving in software’ and are fully on the path of continuous improvement. At the same time, we are working on the ‘next frontier’ by redesigning hardware components to better facilitate the software boost.

Together, we’re well-positioned to create the cornerstone for connected shelves. Not only today, but also ready for tomorrow.