A Simulation Study Using a Network Optimization Model for Super Bowl XLIX Estimated a 39.6 Percent Reduction in Total Vehicle-Hours with Ridesharing Drop-Off and Parking Considerations.
Transportation Network Optimization Microstimulated Using Real-World Data from Superbowl Event held in Arizona Estimated Mobility Benefits.
Glendale
Planned Special Event Network Optimization Model Considering Parking and Ridesharing Drop-Off
Summary Information
A planned special event (PSE), such as a sports game or a concert, can greatly interfere with the normal operations of a transportation system. This study focused on the PSE traffic planning problem by simultaneously considering optimized parking, ridesharing, and network configuration options for relieving congestion. An optimization problem was formulated to minimize total travel time experienced by travelers, utilizing two optimization algorithms. Both optimization algorithms were tested using the real network and traffic data from Super Bowl XLIX held in Glendale, Arizona in 2015, and were evaluated against the baseline network having a default lane configuration.
METHODOLOGY
The objective of the PSE traffic planning method proposed in this study is to minimize the total travel time. The first optimization algorithm was a customized Genetic Algorithm (GA) with Link Expansion (GALE), allowing the expansion of the number of lanes on the links when feasible. The second algorithm was a GA with Congestion Relief (GA+CR) that considered ridesharing locations, recommended routes, and lane configurations. The simulation included parking search behavior, ridesharing, and congestion at drop-off locations.
FINDINGS
- Results show that GA+CR generates a good solution after the first iteration, saving 34.2 percent of total vehicle-hours compared with the baseline network (110.10 versus 167.36 vehicle-hours). This suggests that CR can effectively optimize the network with the given route recommendations and ridesharing location.
- Results also reveal further improvement as more route recommendations and ridesharing drop-off locations are explored, and reaches 101.08 after five iterations, a 39.6 percent reduction from the baseline value.
- Regarding the GALE algorithm, the results showed that GALE was not able to improve the objective function significantly; after five iterations, the total vehicle-hours was improved by 0.9 only (0.54 percent reduction comparing default benchmark solution).
