Skip filling alert: myth or reality? By Nghia Phan, CTO at ffly4u

SUEZ use case: our CTO gives you an overview of the technical challenges to compute the skip filling rate, real KPI of the work in progress in the recycling industry.


What were the technical challenges to meet SUEZ business needs?

Even though one may not  call it as a technical challenge, the most important for us in the beginning of this project was to fully and deeply understand the complete operations of a waste collection center and understand the gain in operational performance that our customer was looking for. Without such deep understanding, it was pointless to go any further. A recycling business operates a limited number of skips, manages different type of wastes/items to be recycled, has limited number of docks, and also shares common resources such as collecting trucks and special compacting equipments. A limited number of skips means that you have to rotate them between the waste collection center (where individuals or businesses would drop off their wastes) and the actual waste recycling/processing center where the wastes are further sorted and processed.

Eventually we came to the conclusion that the Return On Investment of this project was really based on our ability to optimize the collection of only fully filled and compacted skips and hence save thousands of gallons of gasoline.

So the overall technical challenge was to be able  to determine where and foremost when a skip is ready for collection.

What available information could you use to develop a customized solution?

The short answer is that there were no readily usable inputs. As explained earlier, we spent a fair amount of time with the operators in the field to understand the business operations: what information the operators was collecting manually? How this information was then processed and used? For which purposes? What were the decisions they would take based on the manually collected information?

Once we had this understanding, we had to figure out the right technologies and the right “IOT” device to collect this same information but in an automated and systematic manner and then, how to process and report the data and associated analytics to the operators so that the correct decisions could be effectively taken.

The information to be collected was:

  • Very accurate location of each skip, especially when they are located on a specific dock in a waste collection center or when they are located in a waste processing center
  • Type of waste collected in each skip
  • Number of compacting events
  • Number of loading and unloading of skips
  • Duration of stay of a skip on a given dock
  • Number of emptying of skips. Emptying a skip allows us to reset the number of compacting events
  • Number of trips of a skip

The key differentiating and proprietary technology that ffly4u has developed and has applied in this specific use case is called EDGE AI Low Power®: we are processing the data locally within the device (aka “on the EDGE”) using Machine Learning / Deep Learning while minimizing the power consumption of such processing. In standard IOT devices, the processing usually takes place in the cloud with limited and poor data set because the IOT devices usually connect to the cloud using LPWAN network which has very limited bandwidth.


How do you compute a skip filling rate based on this combination of information?

Getting the exact filling level of a skip is the ultimate information that everyone in IOT is looking for. But the reality is that there is currently no affordable technology which is economically compatible with this type of low-value waste recycling business.

Nevertheless, at ffly4u, we managed to get a sufficiently accurate estimation of this filling level by combining the following information, obtained and enriched via EDGE AI LP®:

  • Type of waste: one can hence determine the volume density of the waste. Each dock is dedicated to a specific type of waste. Thanks to our technology, we are able to also assign / associate dynamically in real-time the type of waste to each skip.
  • Number of compacting events for a given skip: the lower the volume density of the waste is, the more you need to perform compacting to optimize the filling of the skip. This data is continuously adjusted by machine learning and our EDGE AI LP® technology.
  • Duration of stay of a skip on a given dock for a given type of waste. This data is continuously adjusted by machine learning and our EDGE AI LP® technology.
  • Number of loading, unloading and emptying operations which helps determining when to reset the number of compacting events between each rotation of skips between the waste collection center and the waste processing center.

The skip filling rate is a key indicator for Suez and more generally for the recycling industry as it has a direct impact on the skips rotation and eventually, on the quality of customer service. 

Write a Comment

Your email address will not be published.