abstract

Empty vehicle movements represent a major source of inefficiency in road freight transportation because vehicles frequently travel without carrying economically useful loads after completing a delivery or when suitable return cargo is unavailable. Recent European statistics indicate that empty vehicle movements accounted for approximately 21.6% of vehicle-kilometres performed by EU-registered road freight vehicles in 2024, with the proportion varying substantially across countries and transport markets. Although empty running is partly caused by structural factors such as geographical trade imbalances, vehicle compatibility, short-haul operations and delivery-time constraints, improved information about future freight demand and available transport capacity can support more effective backhaul planning. This study proposes an integrated predictive framework for forecasting freight demand and transport capacity under uncertainty and subsequently using the forecasts to reduce empty vehicle movements. The proposed framework combines historical shipment information, vehicle availability, origin–destination patterns, temporal characteristics, freight characteristics, operational constraints, market indicators and external factors. A hybrid ensemble consisting of temporal forecasting models and machine-learning algorithms is proposed to estimate both expected freight demand and available transport capacity. Prediction intervals and probabilistic forecasts are incorporated to explicitly represent uncertainty rather than relying exclusively on point predictions. The resulting forecasts are integrated with a stochastic backhaul-matching and vehicle-repositioning optimization model that minimizes empty kilometres, operating costs, unmet freight demand and service-risk penalties. A rolling-horizon decision mechanism continuously updates predictions as new shipment and vehicle information becomes available. The proposed framework contributes to the literature by connecting freight-demand forecasting with capacity forecasting and operational backhaul optimization within a single uncertainty-aware architecture. The framework is designed to provide carriers and logistics platforms with actionable decisions regarding vehicle allocation, backhaul matching and repositioning while maintaining service reliability.