DEMO: Unsupervised fill-level estimation for smart trash removal systems

In this demo, we show an unsupervised, non-intrusive fill level estimation system, called Smartbin, which can be easily installed on the outside surface of waste bins to measure their occupancy levels. Smartbin uses a cheap mini-motor that exploits the physical nature of vibration resonance by learning forced vibration characteristics of the bin at different fill-levels over a small number of garbage collection cycles. This learning process occurs in a completely automated fashion ultimately enabling accurate fill level estimation that can serve as a component of smart (e.g., demand-based) trash removal services. A preliminary evaluation on six different waste bins demonstrates ability of the system to accurately measure empty, half-full, and full bin states. This physical system will be demonstrated to illustrate unsupervised learning and fill level estimation.