Accept Cookies

Smart garbage collection in Ghent: past, present, future

Kenny Helsens
Artificial Intelligence Lead
Ward Van Driessche
Artificial Intelligence Engineer

Dear Stad Gent,

We are writing this blogpost to express our gratitude for Ghent’s participation in open data. The city of Ghent has jumped the bandwagon in 2011 with the Open Data Portaal. Since January 12, 2016 the Belgian federal government also launched the renewed Open Data Portal where Open Data Portals of several public services are collected. At the time of writing, no less than 406 data sets have been made publicly available. These data sets range from the unemployment rate per region, real-time information on the Blue Bikes all the way to every book reservation from the Ghent library. We believe that this way of crowdsourcing a city’s data is a first step towards a first generation of Smart Cities and want to applaud Ghent’s participation.

Smart cities

Over the last decades, numerous open source projects have seen a surge in popularity. Technological advances are growing at an unseen rate and there is no sign of this stopping in the near future. Such advances make our lives easier, but this nuclear explosion of data overload is also making the organization of a city challenging. Crime detection, water supply, energy consumption, waste disposal, etc. are only a handful of examples that can be vastly improved when the right data is available. When enough motivated people get sparked in their creativity with open data, open source projects can produce wonders. With some of our team time, we at In The Pocket came into contact with the city’s data on illegal dumping sites. Below, we describe some of our insights from analysing this dataset, and share ideas on possible next steps for making Ghent an even smarter and cleaner city than it already is.

The “Meldingsapp Gent” (Google Play, App Store) was launched in June 2017 and presented to the public on September 15, 2017. Users are able to report garbage dumps, specify the type of garbage and add a picture of the sighting. Not long after the first public announcement, the first hundreds of notifications came streaming in. Below are some exploratory statistics.

  • Since late October 2018 up and until the end of March 2019, the most active user has submitted no less than 833 reports (an average of 5-6 reports per day), see Figure 1. Thank you, fellow citizen!
  • The second most active user, still submitting a whopping 607 reports in 5 months, seems to be on a mission to clean up the entire city as can be seen in Figure 2.

Locations of the garbage dumps that were reported by the most active user in the period of October 2018 up and until March 2019.

  • Most users used the app to report less than 5 cases.
  • Only about 10% of the users are active on a weekly basis (195 out of 1.995 users submit, on average, around 1.2 times per week.)
  • In total, there have been 43.974 mentions of illegal garbage dumps (490/week) since the app was launched in 2017. All in all, the app can be seen as a huge success!

The data is handled by IVAGO (company responsible for garbage collection of Ghent and surrounding areas). They inspect all notified dump locations and clean them up when necessary. Because the user is also able to attach pictures of the dump sites, staff of IVAGO is able to send out people with the right material. Let’s look at the types of garbage that are reported:

  • regular garbage bags (47%)
  • dump furniture (10%)
  • paper (6.5%)
  • electrical devices (4%)
  • construction waste (2.6%)
  • shopping carts (0.9%)
  • offal (0.2%)
  • other types of garbage (28%)

Aside from the type of garbage, a lot of other data, such as the time of reporting, the location, type of location, etc. are also available. IVAGO labels the reports as well, giving us a good insight in the state of the garbage when being reported by the users. 72% of all reports are acknowledged by IVAGO, the other 28% is shown in more detail in the figure below.

The mean response time of IVAGO to act on a report seems very reasonable. Within 1 day 30% of the reports are finalized, 51% within two days and 77% within 5 days. With limited statistics one could say that Mariakerke, Sint-Amandsberg and Wondelgem are the neighbourhoods that get the fastest response time from IVAGO!

Within 5 days 77% of all reports are finalized

One of the reasons why it takes longer for a report to get cleaned is when it is located on the grounds of third parties, if a crane is required to reach the garbage or if the terrain is not suitable for a truck to drive on. Nonetheless, we are happy to see that the items that are cleaned up the fastest are shopping carts, material with asbestos and dead animals, which seem to be the stuff you wouldn’t want to find lingering around. The items that, on average, take a bit longer to collect are furniture, electrical devices and devices that were not assigned any label in the dataset. It is possible that the latter often had no images attached when the report was submitted, making it difficult for IVAGO to incorporate it into their routing plans.

With this information, we believe that if there is a significant part of the community actively using the app to report illegal garbage dumps, together with automation and garbage truck route optimization, it should be possible to clean all garbage reports within 24 hours. Other possible improvements could be that users do not have to label the data themselves, or apply computer vision to predict the garbage type and improve the user experience.

Clusters of garbage dump locations are made visible with a clustering algorithm. Problematic areas can be found more easily with this feature.

A look into the future

Because of its user-friendliness, the app has been a huge success. A large number of reports have been streaming in, giving IVAGO a better handle on tackling illegal garbage dumps. There is no need for people to call emergency services or IVAGO itself.

Open sourcing some data is the first step, but if we want to go to an efficient smart city, we’ll need to walk the extra mile to reach that goal. Real-time datasets are not widely available yet, aside from a handful of exceptions. The garbage disposal data is, for example, updated every three months.

Predictive analysis and machine learning algorithms are, as far as we know, not yet used by IVAGO. These algorithms could lift organization and efficiency to an even higher level. Also, because real-time data is not publicly available, it is not possible to set up all the features one could imagine. For example, it should be more straightforward to spot potential problems when sudden spikes in reports appear at the same location. Clustering algorithms are capable of capturing these events and possibly send warnings to IVAGO or the city when a lot of litter is being reported. Also, long-term effects can be taken into account, which could arise when infrastructure is lacking at popular locations. If garbage bins are malfunctioning or when baskets are full, spikes could be picked up and be more easily addressed. An example of clusters reports can be seen in Figure 3 for Gentbrugge. Some of these locations could be problematic due to a lack of infrastructure (although other factors, such as an active user living in the neighbourhood cannot be ruled out).

A city needs time to embrace technological and cultural changes. It is possible that certain areas suffer from frequent garbage dumping because of language and culture barriers. If such patterns can be proven through data, it will be more practical to find solutions. If language is indeed an issue, why not leverage modern language technology? We already have tools at our disposal to automatically communicate instructions in 100+ languages, so it should be possible to handle garbage related questions in any language with state-of-the-art chatbot technology. This should be more effective to inform new citizens, who are not familiar with our customs and who do not yet speak Dutch, to get rid of their garbage in the proper way.

There are also many ways to make the app more interactive. If garbage trucks would be capable of sending their location, the user could get notified when a team is on their way. Wouldn’t it be cool to submit a report and get a status update just like how Deliveroo and Uber send out information to their users? We can even go a step further and give users more ownership. We shouldn’t only have to rely on disposal companies to clean out our garbage. Why not reward users that clean up garbage? Litter is not regarded as a garbage dump, but this can be easily incorporated in the app. Smart algorithms can find litter routes for us, cleaning up while walking or running. Maybe this is all a bit too early, a lot of people won’t take the effort to take out their phone, take a picture of the dump and reporting it. But it is possible we’ll be running around in smart glasses that are capable of automatic labeling such dumps and sending an alert within seconds if you want it to. Together creating a city as clean as Tokyo.

The city of Ghent is opening up more and more of its data, giving data scientists another toy to play around in. We’ve seen that the inhabitants of Ghent are passionate about their city, some of them are assigning a lot of their free time to send out reports and help clean up the city. We want to emphasize that more steps should be taken before there can be a surge in powerful digital products that engage our communities. We also do not believe that these datasets should be unique for every city. In the case of garbage dump, you would not want to download an app for every city you are walking around in and it should not be difficult to automatically tag the right corporation if geolocation of reports is available. Why not open the platform to the whole of Belgium, or if you’re reading this, wake up your entrepreneurial spirit and build the next big startup in smart waste management?

Learn how to implement AI

Want to discover the potential of artificial intelligence for your business? Download our AI Value Canvas Toolkit and start experimenting today!

Learn how to implement AI

Related stories

Get more insights

Sign up for our newsletter and stay updated on trends, events and inspiring cases.