research report
This final report describes an advanced traffic surveillance technique based on pattern recognition and the use of current inductive loop technology. The focus of the investigation was a study of the feasibility of using inductive loop signatures for obtaining vehicle classification information on a network-wide level.
conference paper
working paper
At present most patronage predictions of transit systems are performed using UMTA’s UTPS package or some adaptation of it. The transit assignment produced by a typical UTPS system can be classified as an All-or-Nothing limited equilibrium assignment. However, passenger loads assigned to a transit line can far exceed the line capacity. In such a case, line headway has to be reduced to provide enough capacity to accommodate transit demand. If the increase in frequency is not accounted for by iterating again through the mode choice and assignment models, the equilibrium assumptions are violated. If equilibrium between demand and supply is achieved it might occur at a point which requires transit capacity much beyond the economically feasible or engineering practical level. Thus the present transit assignment procedure suffers from two problems. First, trips are assigned to transit lines with disregard to their actual capacity. Second, while some lines are assigned passenger loads beyond capacity, there might be other lines with just slightly longer travel times which are greatly underutilized. A realistic assignment should take into account and not exceed the actual capacity of every transit line. Furthermore, it should consider lines capacities while rationally simulating people’s travel behavior. In this paper a transit assignment algorithm is presented which takes into account the actual capacity of transit lines and assigns trips to more than a single path when the shortest path reaches its capacity. This procedure produces a practical Multipath Capacity Limited Transit Assignment (McLAT). The procedure was implemented on an IBM mainframe computer using standard UMTA’s UTPS package with the addition of only one Fortran program.
published journal article
working paper
A new facility location model and a solution algorithm are proposed that feature 1) itinerary-interception instead of flow-interception; 2) stochastic demand as dynamic service requests; and 3) queueing delay. These features are essential to analyze battery-powered electric shared-ride taxis operating in a connected, centralized dispatch manner. The model and solution method are based on a bi-level, simulation-optimization framework that combines an upper level multiple-server allocation model with queueing delay and a lower level dispatch simulation based on earlier work by Jung and Jayakrishnan. The solution algorithm is tested on a fleet of 600 shared-taxis in Seoul, Korea, spanning 603 km2, a budget of 100 charging stations, and up to 22 candidate charging locations, against a benchmark “naïve” genetic algorithm that does not consider cyclic interactions between the taxi charging demand and the charger allocations with queue delay. Results show not only that the proposed model is capable of locating charging stations with stochastic dynamic itinerary-interception and queue delay, butt that the bi-level solution method improves upon the benchmark algorithm in terms of realized queue delay, total time of operation of taxi service, and service request rejections. Furthermore, we show how much additional benefit in level of service is possible in the upper-bound scenario when the number of charging stations approaches infinity.
conference paper
policy brief
COVID-19 has had lasting effects on transit ridership, with the worst declines seen in high-income, better educated, urban neighborhoods. However, declines among immigrant and/or low-income households was well documented prior to the pandemic, as more gained access to private vehicles. This has created a unique challenge for transit agencies to bring riders back to transit in cases where they may have already switched to traveling by car or consciously chose to make fewer trips. To better understand ridership during the pandemic, we documented the recovery of bus ridership in Los Angeles County and its relationship with COVID-19 vaccinations between April and December 2021, before the Omicron COVID-19 wave. We then developed a statistical model that relates LA Metro bus ridership as a percentage of October 2019 levels with the percent of adults fully vaccinated by ZIP code. We tested whether the relationship between vaccinations and bus ridership varied by two events: first, the full reopening of businesses in California and second, the wave of COVID transmission caused by the subsequentDelta variant.
working paper
The paper shows how variations in systems of property rights explain diverse experiences of urban jitneys and buses. Scheduled bus service entails route specific investments and cultivation of a market. If these investments can be expropriated by interloping jitneys, scheduled service will be dissolved. Property rights in curbspace determine whether scheduled service will be preserved, and whether jitney services will co-exist. We analyze the dynamics of thick and thin transit markets, with and without curb rights. We develop a governance system of curb rights that would let bus operators appropriate their own investments in scheduled service, yet would avoid monopoly by letting jitneys and competing scheduled services operate along the same route. A property rights system dispenses with government ownership, franchise contracting, and regulation.
research report
Numerical simulations have shown that the network fundamental diagram (NFD) of a signalized network is significantly affected by the green ratio. An analytical approximation of the NFD has been derived from the link transmission model. However, the consistency between these approaches has not been established, and the impacts of other factors are still unrevealed. This research evalutes the impacts of start-up and clearance behaviors in a signalized network from a network fundamental diagram approach. Microscopic simulations based on Newell’s car-following model are used for testing the bounded acceleration (start-up) and aggressiveness (clearance) effects on the shape of the NFD in a signalized ring road. This new approach is shown to be consistent with theoretical results from the link transmission model, when the acceleration is unbounded and vehicles have the most aggressive clearance behaviors. This consistency validates both approaches; but the link transmission model cannot be easily extended to incorporate more realistic start-up or clearance behaviors. With the new approach, this project demonstrates that both bounded acceleration and different aggressiveness lead to distinct network capacities and fundamental diagrams. In particular, they lead to start-up and clearance lost times of several seconds; and these lost times are additive. Therefore, the important role that these behaviors play in the NFD shape is studied to reach a better understanding of how the NFD responds to changes. This will help with designing better start-up and clearance behaviors for connected and autonomous vehicles.
Phd Dissertation
For decades, the leading cause of death for American youth has been the car accident, and the largest source of domestic Greenhouse Gas (GHG) and many Criteria Air Pollutants (CAPs) has been the transportation sector. The advent of the autonomous vehicle in combination with Battery-Electric Vehicles (BEVs) and Fuel-Cell Electric Vehicles (FCEVs) presents an opportunity to transcend both pernicious challenges. In particular, the evolution of safer and more efficient autonomous (i.e., robotic) driving behavior via vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication, increased use of electric vehicles, and greater access to affordable and convenient shared (i.e., pooled) rides portend societal benefits including a significant reduction in energy demand and associated pollution. This dissertation evaluates the impact of Shared Autonomous Electric Vehicles (SAEVs, “Saves”) on the California energy grid, GHG emissions, and CAPs. Vehicle-centric impacts (i.e., efficiency changes due to vehicle design and driving behavior) are measured using a vehicle design tool together with a microscopic traffic simulation model to (1) design prototype SAEVs, and (2) measure their energy efficiency for standard and eco-driving scenarios and an array of performance characteristics (e.g., different electric drivetrains, various communication protocols, etc.). Fleet-centric impacts (i.e., changes to vehicle allocation and usage) are measured using ArcGIS with a Caltrans travel demand model dataset to allocate and size SAEV stations, where SAEVs recharge/refuel and are sent to serve nearby trips in a hypothetical SAEV-deployment construct. The Holistic Energy Grid modelling tool (HiGRID) is used to measure SAEV impacts on the California electric grid and grid GHG and CAPs. The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation Model (GREET) is used to measure corresponding transportation sector GHG and CAP impacts. Vehicle-centric energy impacts from SAEV-enabled eco-driving and platooning averaged net efficiency improvements of approximately 6-18%. Fleet-centric impacts include VMT changes from -11% to +36%, largely depending on ridesharing. Depending on SAEV design and operation, over 375,000 metric tons of annual CO2-equivalent GHG emissions could be reduced by adopting the proposed SAEV-deployment construct in lieu of the projected conventionally-driven vehicle fleet. Corresponding CAP impacts include a net reduction of over 250 metric tons of annual NOx emissions.
