Development and Simulation of an Innovative Autonomous Knowledge-Based Smart Waste Collection System

Substantial efforts have been directed to modernize and decarbonize the waste collection industry. Smart sensor-based waste collection (SWC) systems were developed to optimize collection routes based on the actual waste levels in bins, resulting in reduced service frequency, fuel consumption, and air pollution. To date, there has not been a fully functional commercial SWC system due to the complexity and limited practicality of such a hardware-intensive design. Alternatively, this research presents an innovative cloud-based approach in which historical waste generation data, acquired from onboard-truck sensors, replaces data from bin sensors. The proposed knowledge-based waste collection (KWC) system incorporates machine learning for waste forecasting, expert bin selection, route optimization, and autonomous navigation. The system was simulated in an actual residential district to collect recyclables over three scenarios: conventional, SWC, and KWC. Historical data were used to train a generalized linear model (GLM) and a gradient-boosted tree (XGBoost) model to predict daily waste generation per bin. The heuristic bin-selection algorithm selected the bins to be served based on the actual and forecasted waste quantities in SWC and KWC, respectively. XGBoost achieved higher prediction accuracy than GLM with a 4.2% relative error. SWC and KWC significantly reduced the travel distance by 63.1% and 60.9%, respectively, whereas the number of collected bins decreased by 89% for both scenarios. The number of collection days decreased by 5 and 3 days per month in SWC and KWC, respectively. The implementation of connected and autonomous vehicles (CAVs) significantly improved the system, decreasing the total delay by 91% and 90% in SWC and KWC, respectively. Moreover, a life cycle costing analysis revealed that, compared to conventional collection, the reduced travel expenses of SWC were insufficient to offset the cost of bin sensors, whereas KWC achieved a 63% cost reduction by replacing hardware-intensive components with a cloud-based system. Overall, this research has demonstrated that KWC systems can potentially outperform hardware-intensive SWC systems, given the substantial economic and operational benefits associated with such a cloud-based approach.