Understanding and Measuring the Performance of Mixed Mode Urban Freight Delivery Systems Using Large Scale Mobility Data

This dissertation addresses the critical gap in understanding and evaluating the efficiency and sustainability of mixed-mode urban freight delivery systems. Existing performance metrics often fall short in providing a holistic view, being mostly mode-specific and unidimensional. To overcome this challenge, we propose a comprehensive metric - the Urban Freight Mobility Energy Productivity (UF-MEP), providing a multi-dimensional performance perspective. Our study uses a data-driven approach, integrating various inputs across different delivery modes. 

The first part of the research introduces a deep-learning based energy model to assess the risk associated with worst-case energy consumption in drone operations for last-mile deliveries. Utilizing a Temporal Convolutional Network (TCN), we generate a risk distribution which, together with the worst-case energy use, provides a comprehensive risk assessment framework. 

The second part focuses on inferencing truck activity from GPS data, presenting a unique method for determining truck operations, activity hotspots, and thereby enabling an understanding of the overall operational patterns. The algorithm developed in this segment was applied on a large-scale real-world dataset, providing pragmatic insights into the effect of variable data frequency and data loss. 

In the third part, we propose a novel multi-dimensional performance metric, the UF-MEP. Applying this metric to large-scale truck networks in the greater Philadelphia region, we demonstrated its applicability and relevance. UF-MEP provided a unified measure that considers cost, time, energy, and mode compatibility factors, offering insights into the quality and energy efficiency of urban freight delivery systems. 

While this work lays a solid foundation, future research could focus on extending the factors considered within the UF-MEP framework and integrating with regional freight demand models. This could provide a more nuanced tool for policymakers and urban planners, allowing for an evaluation of diverse investment strategies, policy decisions, and the influence of new technologies on urban freight transportation accessibility.