Glass Container Fill Level Measurement via Vibration on a Low-Power Embedded System

In many European countries, up to 90% of glass is recycled. The glass is often collected in large, open, publicly placed metallic or plastic containers. In order to optimize the collection process, reliable information on the fill levels of these containers is of prime importance. Ideally, containers should be emptied once they are filled up to their capacity. Demand-based route optimization for glass collection is only possible if fill levels of containers are available. This paper introduces a novel method to measure the fill level of glass containers via passive vibration analysis on a smart, solar-powered, Narrow-Band-IoT (NB-IoT) enabled sensor attached to one of the walls of the container. Deep neural networks are used in a resource-constrained environment to reliably estimate the fill level. Considerable reductions in NB-IoT network traffic volume and sensor energy consumption were achieved by employing the principles of edge computing without compromising the quality of the algorithm in comparison to a cloud deployment.