RFID + Raspberry PI
RFID Eval Kit
#How does it work?
When we first saw the Coke coolers, we noticed that Coke uses spring-loaded rows to neatly line up the products in their coolers. We affixed an RFID tag to the mobile plastic piece in each row. We then used the RFID signal strength measured by an AS3993 FERMI RFID reader mounted in the rear of the machine to identify how many drink bottles are in each row. If there are many drink bottles in the row, the tag would be closer to the reader and the signal strength would be higher. If there is only a single drink in each row, the tag would be farther away and the signal strength would decrease.
# What does it measure?
A raspberry pi does the necessary computation on the raw RFID signal strength detected by the RFID reader and sends that information through WIFI to a firebase web server which then displays this information on a webpage. The web page is hosted at https://cokehackrfid.firebaseapp.com/ and the data is hosted at https://cokehackrfid.firebaseio.com/.
Does the retailer/bottler/distributor have to do anything different to make your solution work? (if yes, please explain)
No. One significant feature of this solution is that it is extremely modular and easy to modify. If the retailer wanted to monitor their own products, they could use their own RFID tags..
What materials does your solution require?
A raspberry pi, any suitable RFID reader(either a long range high frequency solution or any ultra high frequency solution should work), and RFID tags.
Best guest on cost to implement
The RFID component of this would cost anywhere from $8 to $40 to implement at scale. The wireless component of this would cost anywhere from $5 to $20 to implement at scale. The RFID tags themselves carry a negligible cost so the entire cost of hardware for the system could be between $13-$60.
A high frequency RFID system would be very cheap to implement. An ultra high frequency RFID system would be costlier to implement, however the freestyle drink dispensers implement a similar system. So the solution that is implemented here using ultra high frequency RFID could find its way in other coke products and they could enjoy the same economies of scale. Additionally, some day soon, ultra high frequency tags will be cheap enough to embed in the drinks themselves. Once that happens, if Coke uses this solution, no additional equipment would be needed in these legacy coolers.
Challenges you/your team ran into
Since the RFID system isn't designed for this specific application, the signal was somewhat finicky to work with. We 3D printed a part that holds the tag at an adjustable position to deal with this. We ran into a premature hardware failure near the end, but we'll post the work on github so it can be replicated.
What you/your team learned
RFID is extremely useful and an incredibly difficult technology to beat for inventory tracking.
Should Coke be interested in this solution. We would be able to prototype and deploy several of these modules to test the solution's functionality in the field.
Anything else you want to add
We appreciate AMS' generous support by providing us with an eval kit for their module.
EDIT: Budgetary pricing for the UHF RFID implementation was verified at scale and the costs decreased significantly. A UHF system could be very easily implemented at scale for $20, according to the figures I received.