An integrated approach to City Logistics: A Dynamic Taxi Crowdshipping Method 

Summary Notion of sharing economy and its’ rapid expansion towards many sectors as well as transportation and logistics is irrefutable. The fast and continuous growth of e-commerce causes a huge increase in the number of online transactions and freight transport. Thus, traditional ways of freight delivery and in general city logistics in dense urban areas are becoming more challenging than before. Providing enhanced shipping system is essential and requires a better understanding of trip patterns, integration of different modes to improve accessibility and better management of transportation system to accommodate demands. Crowdshipping or crowd logistics (CL) is an emerging solution to freight transport and especially last mile deliveries. It has a strong bond with the concept of sharing economy since it allows activities that were traditionally performed by certain companies to be outsourced to a large pool of individuals. This service is mostly provided by the ordinary people and everyday commuters who undertake the role of a carrier. According to DHL Logistics Trend Radar report in 2013, currently, almost 70% of the available transport (rail, road, private cars) capacity is underutilized. Therefore, sharing assets and capacities are very beneficial for enhancing consolidation, higher capacity utilization and it may help to decrease congestion and car emissions. The key to addressing this problem lies in the ability to utilize extra facilities such as idle taxis, buses and even passenger car trunks for moving goods around. In this dissertation, taxis are considered as a crowd to deliver packages.        Compared to other public transportation modes, taxis do not have fix infrastructure which enabling them to provide flexible services to serve requests in any part of the city. However, finding the equilibrium between the demand and taxi supply is a challenging problem. While there may be neighborhoods with unmet demand, others may have abundances of vacant taxis roaming in search of passengers. In big cities, with enormous taxi ride demand, the presence of taxi supply-demand imbalance reduces taxi utilization rates, decreases customer satisfaction and lowers taxi service reliability. This may result in a loss of profits for taxi companies. The ability to forecasting the trip demand is of a significant importance for (a) taxi companies by empowering them to relocate vacant taxis to unserved areas and decrease the fleet idle time and (b) vacant taxis can be of immense help -for carrier companies- to distribute parts of delivery tasks and earn extra profit in their idle time. This study consists of three major work streams. First: a stated preference (SP) survey is designed to collect information about people attitude towards crowdshipping, to understand their requests and accommodate them by the system. Second: taxi pickup demand is predicted by utilizing a deep learning approach that leverages Long Short-Term Memory (LSTM) neural networks. The study is based on publicly available taxi data for New York City. Taxi pickup data is binned based on geospatial and temporal informational tags, which are then clustered using a technique inspired by Principal Component Analysis. The spatiotemporal distribution of the taxi pickup demand is studied within short-term periods (next one hour) as well as long-term periods (next 48 hours) within each data cluster. Third: a taxi crowdshipping system is designed based on information and results gained in the previous sections and extra capacity in taxis is exploited to deliver parcels and serve passengers collectively without hurting the quality of taxi service. This study offers an advanced CL solution for parcel delivery tasks – a Dynamic Taxi Crowshipping Model (DTCM) - in a real-time dynamic setting. The Mixed Integer Linear Programming (MILP) formulation used within a rolling horizon scheme that periodically updates input data information and designed to handle uncertainty caused by dynamic nature of this problem. The aim of the study is to minimize the total system-wide vehicle miles incurred by system users, individual travel costs and number of delivery vehicles. The results show that the proposed dynamic model is promising and has great society advantages such as saving travel cost, reducing travel time and mitigating traffic congestion.