One primary goal of the project is to develop a measurement system to assess the public value of logistics activities. We have achieved a significant milestone towards this goal by completing one of the most extensive analyses of sustainability indicators in urban transportation to date, with an in-depth investigation of 346 peer-reviewed papers and identifying over 4,300 unique indicators - making it the largest compilation both by the number of papers analyzed and the indicators identified. For the first time in the literature, we conducted a comparative analysis of theoretical and applied studies on measuring the public value of transport, uncovering gaps between theoretical contributions and practical needs. We examined the differing treatment of passenger and freight transportation, highlighting the contrasting perspectives of public necessity versus business-centric priorities. We introduced a novel multi-layered hierarchical classification framework to tackle the persisting standardization issues as a major obstacle for collaborative advancement of the field. Leveraging this structure, we introduce innovative quantitative tools to enhance the design and effectiveness of sustainability assessment frameworks as a core for measuring the public value of city logistics.
In addition to our work on sustainability indicators, we have completed three pioneering studies on network design problems on emerging delivery concepts.
• In our first study, we propose a dynamic compensation scheme for crowdshipping that considers the spatial and temporal distribution of delivery demand to continuously optimize compensation rates. For the first time in the literature, we conducted experiments with real-world package delivery demand along with real-world personal trip data to assess the potential of this emerging delivery mode. Our results also show that compared to traditional dedicated delivery services, the investigated crowdshipping setting with dynamic compensations can provide more than 36% cost savings and more than 40% reductions in vehicle miles.
• In the second study, we focus on a new express shipment model that combines public transportation with Autonomous Robots. Under dynamic demand arrivals with short delivery time promises, we propose a rolling horizon framework and devise a machine learning-enhanced Column Generation (CG) methodology to solve the real-time dispatching problem. We show that the proposed system can achieve more than 85% reduction in vehicle traffic, emissions, and noise. Our results demonstrate the effectiveness of the novel learning-based CG methodology providing nearly the same quality as the classical CG method while achieving orders-of-magnitude reductions in runtime for large-size instances.
• In the third study, we introduce a collaborative pooled last-mile delivery model in which an e-grocery retailer leverages its excess capacity to simultaneously deliver e-grocery orders and e-commerce packages, aiming to enhance both profitability and sustainability. We develop a novel learning-based algorithmic framework to solve the complex sequential decision-making problem (SDM) for managing this delivery operation, contributing to the burgeoning literature on combining machine learning with optimization to address complex decision-making problems. Using the data we received from a large e-grocery chain, we show that the proposed model can decrease the marginal fulfillment cost of e-commerce packages by over 30% and reduce vehicle miles by approximately 15.7%. The learning-based approach provides solutions that are almost as good as the decisions with perfect hindsight and significantly outperform state-of-the-art methods for solving SDMs in run time and solution quality.
