The Education of a Dumpster
Testing a Real-Time Resource-Fill Thing of the Internet
On Sunday, April 13, 2014, I gave a short talk on the trials and tribulations of creating what is most likely the first fully-operational Smart Dumpster. Although we may have been the first, we are no longer the industry leaders in this evolving technology, an honor that is held by the small hardware upstart Compology out of San Francisco, California.
In its earliest iteration, our Smart Dumpster was intended to simply alert our resource recovery (a term formerly known as trash) manager of the real-time fill levels of our campus dumpsters. If properly utilized, teaching a dumpster to communicate when it needs to be serviced allows managers to reduce the labor and wear on service trucks, as well as to avoid the nuisance of heavy equipment unnecessarily frequenting the neighborhood.
What began simply as an idea to install a fill switch on a dumpster evolved into a complex hardware-cloud-software apparatus. As our test version of the Smart Dumpster relied on sonar-based technology, I installed a Massa M3 that came with a developers kit. Built for the petro-chemical industry, this sonar was a great match against the noxious materials that could be present in a dumpster environment. After physically installing the M3, I used Digi’s platform to link the sonar to a cloud application. Within a week’s troubleshooting, I could personally monitor the fill level of our test dumpster. Immediately, I set up automatic “alerts” for our dumpster, indicating when it was full or emptied based on parameters set for observation. Even so, this was not the dashboard interface I had hoped to provide our resource recovery manager upon project development. I wanted to create a dashboard that was legible to any user, one that could be accessed on the internet using normal computing technique. My original concept is illustrated in FIGURE 1, shown below using images of the actual dumpster under evaluation:
Based on investigation and the support of both Massa and Digi engineers, I was led to Python programming. Months after learning this computational language would be the key to my dashboard, I found one of our student workers (Dylan van Krieken, pictured in FIGURE 3) had the interest and capacity to program the “back-end” of a small, effective demonstration dashboard. Our project was best executed as illustrated in FIGURE 2, shown below:
Dylan, our department’s student worker, did a great job of taking the sonar’s readings from the cloud and creating two different files. First, he updated a preordained text file that could be read using HTML. Second, he created an Microsoft Excel spreadsheet that kept a running record of all the sonar’s readings over the life of the project. Once Dylan successfully programmed these two endpoints for our data from the sonar, I wrote a relatively simple HTML website that acted as the actual dashboard interface for our department’s resource recovery managers, shown below in FIGURE 3. On an hourly basis, our department’s manager could read the precise depth of the test dumpster PHY-PLNT-N via this website. On the website, you can see that I was planning on adding three additional sonars to nearby dumpsters in the next iteration of the test (PH-PLNT-CRBD, PAINTSHP and CARR-HSE).
Important for an analyst, the records of dumpster fill were illuminating as to the patterns of user deposit and operational management. Here the green line shows the depth of the dumpster over a three-month period. Never did the dumpster reach a volume over 74%; and typically, the dumpster was emptied after roughly a 1/3rd fill level. Issues with the battery life also appeared as for days the sonar was not taking readings.
As shown, the sonar allowed me to evaluate the accuracy of two other systems our department relies on for operations and billing, as illustrated in FIGURE 4 below. The sonar was able to verify that our first system, a daily log maintained by our truck drivers, is surprisingly accurate over the given time period. During the test, there was only one instance at item “a” that a service was logged but the dumpster was not emptied. Further, the sonar was able to verify that our digital on-board system, an integrated scaling apparatus called Loadman, was accurate both as to relative weight and regular service recognition. Only at item “b” below did a driver and the sonar register a dumpster service but the Loadman system did not.
Currently, our department has expanded the use and review of the Smart Dumpster system with 43 operating sensors created and managed by Compology on-campus. With this new technology, our resource recovery manager has already increased the average volumetric fill at dumpster service by nearly 20%; that is from the average 1/3rd fill shown in FIGURE 4, to an average today of 55% fill upon each pickup.
I must thank both Roger Edberg, Senior Superintendent at UC Santa Cruz’s Grounds Services, and Dylan van Krieken, student worker at UC Santa Cruz’s Grounds Services, for their cooperation and talents in realizing this project. The featured Smart Dumpster was executed at the University of California, Santa Cruz in furtherance of the campus’s Zero Waste 2020 ambitions.
