To avert catastrophic climate change, society must rapidly shift to decarbonized forms of transportation; recent reductions in the cost of batteries has made battery electric vehicles (BEVs) a promising alternative. Meanwhile, mobility on-demand vehicle (MODV) services have grown explosively in recent years, in some cases threatening targets for local air pollution and global carbon emissions. Mobility on-demand is a novel form of transportation that represents a shift from private ownership to commodification, in which users request transportation services for individual trips via smartphone apps. Despite evidence that these services are ripe for electrification, adoption of BEVs in fleet applications has been even slower than in the private market. In this dissertation, I combine agent-based simulation with empirical methods to identify pathways for rapidly electrifying MODV services, including requirements for charging infrastructure and battery range, routing strategies, and regulatory tools.
In Chapter 1, using taxi trip data from New York City, I develop an agent-based model to predict the battery range and charging infrastructure requirements of a fleet of shared automated electric vehicles (SAEVs) operating on Manhattan Island. I also develop a model to estimate the cost and environmental impact of providing service, and I perform extensive sensitivity analysis to test the robustness of my predictions. I estimate that costs will be lowest with a battery range of 50-90 miles, with either 66 chargers per square mile rated at 11 kilowatts or 44 chargers per square mile rated at 22 kilowatts. I estimate that the cost of service provided by such an SAEV fleet will be $0.29-$0.61 per revenue mile—an order of magnitude lower than the cost of service of present-day Manhattan taxis and $0.05-$0.08/mi. lower than that of an automated fleet composed of any currently available hybrid or internal combustion engine vehicles (ICEVs). I estimate that such an SAEV fleet drawing power from the current NYC power grid would reduce GHG emissions by 73% and energy consumption by 58% compared to an automated fleet of ICEVs.
In Chapter 2, I build on this analysis to study the electrification of ridesourcing services (also known as transportation network companies, or TNCs) in the U.S. in the present day. Ridesourcing/TNC fleets present an opportunity for rapid uptake of battery electric vehicles (BEVs), but adoption has largely been limited to small pilot projects. Lack of charging infrastructure presents a major barrier to scaling up, but little public information exists on the infrastructure needed to support ridesourcing electrification. With data on ridesourcing/TNC trips for New York City and San Francisco, and using agent-based simulations of BEV fleets, I show that given a sparse network of three to four 50kW chargers per square mile, BEVs can provide the same level of service as ICEVs at lower cost. This suggests that the cost of charging infrastructure is not a significant barrier to ridesourcing/TNC electrification. With coordinated use of charging infrastructure across vehicles, I also find that fleet performance becomes robust to variation in battery range and charger placement. My analysis suggests that mandates on ridesourcing/TNCs, such as the California Clean Miles Standard, could achieve electrification without significantly increasing the cost of ridesourcing services.
In Chapter 3, I shift to look at real-world barriers to MODV electrification based on empirical data. Leveraging over two weeks of high-resolution GPS and battery data from almost 20,000 EVs in the all-electric Shenzhen taxi fleet, I analyze the potential to improve fleet-wide operations by optimizing both the location and timing of vehicle charging. I construct machine learning models to predict travel time, queuing time at charging stations, and charge consumption by time of day. Contrary to the emphasis on charging station siting in the literature, I find that optimizing charging locations would have a relatively limited impact. Instead, providing drivers with better real-time information about queuing times at charging stations, and enabling flexibility in battery charge during shift changes could reduce down-time per vehicle by over 30 minutes per day, while increasing the number of economically viable charging stations by over 50%. Moreover, taking full advantage of break periods and nighttime to charge could reduce downtime per vehicle by over one hour per day, reducing revenue losses due to charging by roughly 90%. These results are verified with evidence from real-time charging station data and driver shift-change data.
Contributions:
This dissertation contributes to the knowledge of sustainable transportation engineering through advances in theory, methodology, and empirical results that can help guide MODV electrification policy and implementation world-wide. Previous literature has claimed that MODV services are hard to electrify due to challenges with battery range and charging. In particular, operations research literature has used rigid models that either do not accurately reflect real-world constraints or require BEVs to operate in the same way as ICEVs. I hypothesize that allowing vehicles to charge during short windows in between trips and relocating to new areas to anticipate demand can greatly increase BEV fleet performance. To test this hypothesis, in Chapter 1 I develop a novel agent-based modeling method in the field of operations research, including a new theoretical approach to fleet rebalancing based on the equivalence between efficient demand forecasting and retrospective assignment. I hypothesize that by having vehicles “look back in time” to relocate to areas with unmet demand and to charging stations in the present, the model can evaluate minimum requirements for fleet size, battery range, and charging infrastructure with a fraction of the computational cost of other approaches. To test this hypothesis, I also developed a fleet rebalancing framework based on demand forecasting and show that the retrospective approach returns equivalent results in much less time.
As a result, I show that BEVs can provide MODV service at scale and at lower cost than gasoline vehicles with present-day technology, while resulting in more rapid reductions in air pollution and greenhouse gas (GHG) emissions than comparable investments in private vehicle electrification. By maximizing flexibility in the design, this method also enables application to new environments, as explored through the development of a national scale model at the end of Chapter 1, and a TNC-specific model in chapter 2.
Previous studies and pilot projects have found that there is insufficient charging infrastructure to support TNC electrification, in part because this infrastructure is prohibitively expensive. I hypothesize that the cost of charging is largely driven by usage rates, such that efficiently operated TNC fleets can drive down the cost of public fast charging for all users. To test this hypothesis, in Chapter 2 I develop a simple theoretical approach to estimating requirements for TNC charging infrastructure based on driver wage rates, driving speeds, and relocation times. I then adapt the model developed in Chapter 1 to show that drivers do in fact have sufficient idle time to charge, and that charging infrastructure is relatively cheap given reasonable usage rates.
Finally, chapter 3 applies these analyses to current BEV taxi operations in Shenzhen, China, where drivers have complained of lost revenue and long queues at charging stations. Based on my previous findings, I hypothesize that simple strategies to improve operations could greatly reduce these problems. Whereas previous studies have conducted analyses based on incomplete vehicle data, I integrate qualitative methods from in-person interviews (n=30) with quantitative insights from analysis of multiple detailed datasets, web scraping and machine learning in an interdisciplinary approach. I hypothesize that such a mixed-methods approach would result in more nuanced findings than any one method could provide on its own. I show that reduced revenue from BEVs results from irrational charging behavior that can be mitigated by simple software and policy interventions. In turn, I find that real-time data collection and analysis efforts are crucial to efficient MODV electrification.
Policy recommendations from this dissertation include establishing firm electrification targets to catalyze investment in fast-charging infrastructure; establishing citywide open data platforms to integrate real-time data on vehicle trajectory, battery charge, and charger availability; and providing drivers and companies with training on best charging practices. Such capabilities may also require labor policy reform to incentivize fleet operators to manage their drivers’ charging behavior. In turn, digitization enabled by fleet electrification holds the potential to enable a host of smart urban mobility strategies, including integration of public transit with innovative transportation systems and emission-based pricing policies. As a number of cities worldwide move toward fully electrified MODV fleets, this analysis has large-scale implications for decarbonized, cleaner urban areas.
Limitations and directions for future research: This dissertation does not explore how changes in vehicle supply and cost may impact trip demand. I have conducted a variety of sensitivity analyses to ensure my results are robust to changes in trip demand, but this is a topic that deserves further examination in the future. Similarly, I have not analyzed how changes in fleet operation or trip demand might impact congestion, which was critical to enabling computational efficiency and geographic flexibility. I have made a variety of conservative assumptions to counteract this omission, such as slightly increasing travel times and giving trip requests a 10-minute buffer, but future work should address this topic more thoroughly. Future work should also expand geographic scope, especially focusing on issues specific to suburban and rural settings. I have not closely examined the impact of vehicle charging on the power grid or potential benefits coordinated charging might provide. Future work should also seek to incorporate charging activity by other types of BEVs, such as private vehicles and taxis, while further exploring behavioral factors that influence charging decisions, such as availability of rest places and food, opportunities to meet friends and other drivers, and range anxiety.
Finally, this study does not deeply analyze impacts on social equity, which deserves further analysis in future research. Academics, social justice advocates, and policymakers must work to ensure these new technologies help redress past and present injustices rather than exacerbating them or merely maintaining the status quo.
