A data-driven optimization approach to plan smart waste collection operations

Abstract

Waste collection is an important logistic operation that is often inefficient due to the high uncertainty associated with bin fill levels, resulting either in routes that visit empty bins or in bins overflowing due to lack of routes. To reduce such uncertainty, sensors installed in the bins can provide real-time information on waste levels. However, this is not enough, and the management of this information needs to be combined with dynamic optimization approaches to effectively design smart collection routes. This paper investigates this challenge and uses real-time information on waste levels, treated through machine learning techniques, to feed a dynamic reverse inventory routing optimization model. The resulting data-driven optimization approach allows the definition of a medium-term collection route plan that can be updated daily as new information becomes available. To demonstrate the applicability of such an approach, a real-world case is solved, and the results show significant improvements in operational efficiency and service levels.