Real-time Operation of Shared-use Autonomous Vehicle Mobility Services: Modeling, Optimization, Simulation, and Analysis

Hyland, Michael Francis

doi:https://doi.org/10.21985/n2-p50g-y585

Work

Real-time Operation of Shared-use Autonomous Vehicle Mobility Services: Modeling, Optimization, Simulation, and Analysis

Public

Download PDF

Download All Files (.zip)

Two recent developments in the transportation industry – shared-use mobility services and fully-autonomous vehicles (AVs) – have the potential to fundamentally transform urban mobility. Shared-use mobility services (e.g. Uber, Lyft, Via, Chariot, ZipCar, and Car2go) are already beginning to bridge the gap between personal vehicles and fixed-route transit service in terms of cost, convenience, comfort, and efficiency for many trip purposes. The integration of AVs within shared-use mobility services should accelerate their growth via significantly decreasing their operational costs. The two major advantages of AVs over human-driven vehicles within mobility service fleets are (1) the elimination of driver-related labor costs that currently make up a considerable portion of the total operational costs of ridesourcing and taxi services and (2) their ability to improve operational efficiency via allowing a central fleet operator to completely control all vehicle plans (i.e. routes, schedules, repositioning, and user assignments) in real-time. The first advantage is rather evident, and its potential impact on mobility services and future transportation systems is quite large. The second advantage (i.e. complete control of all AV plans) allows mobility service providers to optimize the operation of the entire mobility fleet with complete information on all vehicles and traveler requests, rather than relying on individual driver decisions with incomplete information. Given the potential paradigm shift in urban mobility due to shared-use AV mobility services (SAMSs), research in this area is particularly important. This thesis focuses on research problems associated with operating SAMSs, particularly, modeling, controlling, simulating, and analyzing the real-time operation of SAMSs. From a SAMS provider perspective, operating SAMS fleets efficiently can improve service quality, reduce operational costs, increase profitability, and increase market share. From an individual traveler perspective, assuming reasonable competition between mobility service providers, more efficient SAMS operations should lower prices and improve service quality for travelers. From a societal perspective, efficiently operating SAMS fleets can decrease (unproductive) vehicle miles thereby potentially decreasing congestion, fuel consumption, and harmful emissions. Motivated by the importance of SAMS operational efficiency in terms of capturing benefits for mobility service providers, individual travelers, and society as a whole, as well as the inclusion of AVs in shared-use mobility services of the future, the overarching goal of this thesis is to support the operation of specific SAMS offerings via defining, modeling, and presenting solution approaches for SAMS operational problems. The specific objectives of this thesis include (1) identifying relevant SAMS operational problems, (2) analyzing the efficiency of existing taxi fleets, (3) modeling and developing solution approaches for several timely SAMS operational problems, and (4) analyzing the relative operational efficiency of specific SAMSs using agent-based stochastic simulation methods. To meet the first objective, the thesis presents a taxonomy of vehicle routing problem variants relevant to SAMS operational problems. To address the second objective, this thesis presents two metrics to characterize the operational efficiency of a taxi fleet based on taxi trip data. The core of the thesis lies in meeting the third objective. One SAMS operational problem deals with dynamically assigning AVs to open user requests for an on-demand SAMS without shared rides. Another SAMS operational problem deals with simultaneously assigning AVs to open user requests and repositioning AVs throughout a service region to serve future demands, for an on-demand autonomous carsharing service. The last SAMS operational problem deals with assigning idle and en-route drop-off AVs to open user requests for an on-demand shared-ride SAMS. The SAMS operational problems presented in this thesis represent original instances of stochastic dynamic vehicle fleet operational problems. While they share many features with problems in the existing dynamic freight routing literature, taxi-dispatching literature, and ambulance-dispatching literature, the combination of the SAMS operational problems’ size, degree of dynamism, degree of urgency, spatial distribution of user requests, and short user pickup and drop-off times make the problem instances unique relative to the existing literature. Finally, to meet the fourth objective, this thesis employs agent-based stochastic simulation methods to analyze the operational efficiency of specific SAMSs under various conditions. For example, the thesis presents a methodological framework to evaluate and quantify the impact of spatio-temporal demand forecast aggregation on the performance of an on-demand SAMS without shared rides. Additionally, the thesis evaluates the operational efficiency benefits of an on-demand shared-ride SAMS ¬compared to an on-demand SAMS without shared rides from the perspective of the SAMS fleet operator. The results of the analysis comparing an on-demand shared-ride SAMS to an on-demand SAMS without shared rides are particularly valuable from a transportation planning and policy-making perspective. The analysis indicates significant SAMS fleet operator benefits associated with offering a shared-ride SAMS in addition to the individual mobility benefits (significantly lower travel costs at the expense of slightly longer in-vehicle travel times) and societal benefits (more shared-ride trips implies higher vehicle occupancy which subsequently implies lower vehicle miles, traffic congestion, fuel consumption, and vehicle emissions). From a SAMS fleet operator perspective, providing a shared-ride service requires a significantly smaller fleet size than a service without shared rides, even when the maximum number of traveler groups in an AV is two and the maximum user detour distance/time is only allowed to be 5 percent more than the user’s shortest route distance/time. In meeting the above four research objectives, this thesis makes several valuable scientific contributions and provides significant value to several entities in the transportation industry. The scientific contributions range from modeling and developing solution strategies for new stochastic dynamic vehicle routing problem instances to analyzing the operational efficiency of existing taxi fleets and future on-demand SAMSs. The solution approaches can inform existing mobility service operators and future SAMS operators. The modeling framework for SAMSs can help transportation modelers incorporate SAMSs into transportation network models. Finally, the operational efficiency analyses can inform transportation planners and policy makers as they consider plans and regulations, respectively, for AVs and SAMSs.

Creator