We enabled touchless gas pumps for a global network in under 3 weeks with our low-code middleware Interplay®. The COVID-19 pandemic has forced people to be more cognizant of what they are touching and the potential spread of germs. All kinds of places had to adapt. Restaurants had to rely on take-out. Grocery stores enforced utilizing masks. Though many people opted to avoid unnecessary risks, some things are still essential, and gasoline is one of them.

Typically, when stopping to get gasoline, there is a fair amount of risk in terms of touching things. A person has to touch a screen to activate it, enter in their credit card, key in their PIN, press a button to select their type of gasoline, then finally use the gas pump by holding the pump handle. Now, as people are more aware of the spread of germs and viruses, everyone knows that all of this touching is less than ideal.

A global convenience store/gas network began asking thoughtful questions: What are ways to minimize the number of hand touches during a fuel pumping experience? What if we could handle the money transaction via the customer’s mobile phone before a customer even gets out of their car?

Many solutions emerged. First, the gas station needs to know the car's identity from the license plate. The system can then connect to the mobile phone of that car’s owner to authorize the transaction, all while you’re still just sitting in your car. Using the mobile phone, the transaction can be verified, along with an opportunity to browse and choose some curbside snacks from the gas station. With this system, over half of the possible risk from touching the pump parts is mitigated. BONUS: this also mitigates the risk of scamming credit card scanners.

How to Build a Touchless Gas Pump - So, there are a few steps here.

• First things first: get this platform running on an edge server at the gas station-- no reason to pay for extra data back and forth to some central cloud
• We need to find the license plate of the customer’s car in order to trigger the AI recognition. This meant connecting and maintaining camera connections across thousands of cameras in a low latency network. The typical deployment has 16 simultaneous video streams on a single Interplay edge system.
• It also meant ignoring the cars that simply drove through the fuel bay and didn’t stop. Bonus difficulty: this is happening with fog, snow, rain, and other weather in unpredictable conditions in Norway and Sweden.
• Once we’ve captured an image, the image recognition necessity requires a good AI engine, as well as a sufficient amount of training data to account for various angle geometries and even different plate formats (varied by county/region).
• We adapted an Open Source engine for reading license plates, avoiding a paid service that would have cost several million dollars annually.
• Once a license plate is read, it needs to be compared to the list of registered customers (who, as part of this program, put in their information and license plate number as part of the sign-up process). We need to connect to the enterprise-level customer database. From this, we know the customer’s mobile phone for messaging and even know which side of the car has the gas inlet.
• Once a match is found, the system will then ping the customer’s mobile phone (also part of the sign-up) to confirm: “Hi, I see you just pulled in for gasoline. What grade?” This all requires a further connection to a messaging platform
• Once the customer confirms the transaction and gasoline type, the system now needs to unlock the pump, which requires an IoT connection to the pump’s physical mechanisms for the handle and gasoline type.
• The customer can now fill-er-up. Once completed, the system can now confirm the transaction, which means completing the transaction using the stored credentials and a connection to the payment gateway and order management system
• Once the transaction is completed, the system once again confirms everything back to the customer’s phone via the messaging platform.

That sounds complex, and it is. Moreover, we have to package all of that into a simple plug-n-play server so gas station managers can deploy this with minimal IT support required-- no AI knowledge required. We’re now also helping our client file patents around this revolutionary award-winning system.

Indeed, we got it done in about 3 weeks, from conception to actual production roll-out. This system is now in place at hundreds of convenience stores, on its way to be rolled out to thousands of more stores in the coming months.

If anything, the description here is a very simplified version of the actual data and logic flow. Despite this complexity, we estimated that we could do it in less than a month and at a very affordable price. Why were we so confident? Because we have Interplay.