The problem

Litter bin collection in the UK and not only is not only very inefficient, but also involves very high levels of CO2 released into the atmosphere. Bins are routinely overflown or emptied when not needed, because local councils do not know where bins actually need collecting. Councils want to know where to best place new bins and recycling bins.

The system

Our management system shows the status of all bins in real time and provides the optimal route to collect full bins and minimise distance travelled and therefore emissions. All this is displayed on an online Azure map.

Cheap ultrasonic sensors measure the depth of the bin and buttons allow users to indicate if they threw a recyclable object. Data is fetched over WiFi at fixed intervals. Power and connectivity were an issue; in real life, these sensors would probably be running off solar power and a GSM chip.

When needed, the system takes an average fullness of bins and suggests the placement of new bins in high intensity areas. During a big event, this can help organisers optimise bin placement after a day of use. If some bins show high recycling content, the Python script will suggest adding an extra recycling bin there.

The optimal route algorithm

The data is modeled as a graph where the nodes represent bins and the edges are roads that connect them. The cities are split into regions and each day all the bins from one region need to be emptied. The algorithm we implemented analyzes the data from the hardware on the bins and selects the most critical region that needs to be cleaned. It then finds a path to that region and back to the waste center that minimizes fuel and maximizes how much trash is collected on our way, using a modified Dijkstra technique for minimal paths.


Connecting the different components was tricky because of the different languages used and of trying to continually read data from them. We learnt that the central server required to read the data required more attention! With the correct wireless infrastructure, the system is easily scalable to an entire local council or an entire convention centre. Dealing with inaccurate sensors was also difficult although there are known better sensors (e.g. VCSEL).

Accessing the Azure map data and showing it took some time and we learnt how to use their APIs.

Share this project: