Optimizing shared-use curbside spaces in connected transportation networks accounting for uncertainty

Abstract

For the past two decades, shared-use mobility (SUM) has been changing the landscape of urban transportation. The efficient use of shared-use mobility services to promote the sustainability of a transportation system largely relies on the effective use of curbside spaces. As the use of shared-use mobility services keeps growing, it is challenging the traditional approaches used to manage curbside spaces. The task of managing the curbside is often challenging as it is hard to predict the demand for those spaces. Also, most cities prefer to allocate curbside spaces for long-term parking needs. However, the lack of accessibility of these spaces often leads SUM services to complete their pickup and dropoff services in the active travel lanes making the network congested and worsening the emissions. The use of time-restrictions-based flexible curb-use strategies have gained popularity as it has been able to provide a solution to address the needs from various services while reducing congestion stress from the street. However, the curbside managers often allocate the spaces based on historical demand without using any optimal strategy. The optimal allocation of these spaces can help provide better access to different modes of transportation and maximize productivity. This research aims to develop an optimization module to help the curbside managers allocate spaces rationally while maximizing the benefits. It is critical to assess the impact of the shared-use services on mobility and the environment when different variables, such as road available curbside spaces and stopping time restrictions, are in effect. This research also aims to evaluate the mobility and environmental impacts of time restriction-based curbside practices considering these dimensions. Lastly, this research explores the performance of the developed optimization model in field condition when drivers can have real-time communication with the RSUs . We used a fine-resolution simulation framework to evaluate curbside management strategies for ride-sourcing and delivery vehicles. Utilizing microsimulation tools Vissim and EPA-MOVES, we analyzed mobility and environmental impacts in both a small hypothetical network and a complex real-world urban network in Chicago, Illinois. Our results show that eliminating long-term parking can worsen congestion but reduce vehicular emissions. Post-trajectory analysis revealed that increased vehicle speed due to reduced curbside spaces impacts emissions. Key findings of this part also highlight trade-offs between congestion and emission reduction, the effect of adjusting dwelling time windows on curbside productivity, and the relationship between curbside space reduction and energy consumption. Additionally, fleet electrification scenarios indicate that increased electric vehicle adoption decreases emissions, mitigating environmental impacts while meeting mobility needs. Later we developed a linear mathematical model using a knapsack constraint to allocate spaces optimally. For this purpose, Curb Productivity Index developed by Fehr & Peers has been used as objective function. The mathematical models were developed in AMPL to solve it using the IBM CPLEX Solver. This model was tested across various scenarios to obtain solutions under different variables. The results directed the impact of different variables on curbside productivity and optimal allocation. The results indicate that the curb productivity index (CPI) is highly sensitive to the number of passengers served. When arrival demand is constant, shuttles get priority at curbside spaces if they serve more passengers. If shuttle demand increases but serves fewer passengers, their space allocation decreases. This variation in CPI is due to the number of passengers served. Dynamic optimization shows that CPI changes with vehicle arrivals at each interval, allowing for various allocation strategies. By minimizing the loss of Expected Mean CPI, we can select an optimal allocation for greater real-world benefits. In the final part of the dissertation, we developed a microsimulation model to estimate curbside productivity variations from the theoretical model. Using PTV Vissim, we assessed the impact of curbside location relative to intersections and data connectivity on productivity. The average Total CPI in the simulation model is about 32% higher than in the theoretical model, although the number of TNCs (Transportation Network Company) served is around 14% less, indicating CPI depends heavily on distribution. It also showed that, curbside productivity improves significantly when curbsides are located farther from intersections, enhancing network productivity. Information availability to drivers also boosts curbside productivity.