An Agent-Based Simulation Model of Automated Vehicles in San Francisco Estimated That Shared-Ride Home Pick-Up Services Would Result in a Cost Savings Benefit for 33 Percent of Trips With Human Drivers and 73 Percent of Trips With Fully Automated Vehicles.

San Francisco Study Investigates Effects of Automated Vehicles (AV) on Mode Share, Automated Taxi, and Ride-Hailing and Ridesharing with and without AV.

Date Posted
09/29/2023
Identifier
2023-B01785
TwitterLinkedInFacebook

Automated Vehicle Scenarios: Simulation of System-Level Travel Effects Using Agent-Based Demand and Supply Models in the San Francisco Bay Area

Summary Information

Given the popularity of automated vehicles (AV) and in anticipation of full automation as late as 2035, it is important to understand the system-level effects of these technologies. This study first explored the effects of AV on mode share using the San Francisco Bay Area Metropolitan Transportation Commission’s activity-based travel demand model (MTC-ABM). The individual travel activities were draw from MTC’s 2000 Bay Area Travel Behavior Survey. Second, the study simulated the effects of introducing an automated taxi (AT) service in the San Francisco Bay Area region and used the existing research on the costs of AV to represent plausible per mile AT fares. Third, the study used an integrated model for the San Francisco Bay Area that included the MTC-ABM combined with an agent-based simulation model customized for the region, to simulate different “first” mile transit access services, including ride-hailing and ridesharing with and without AV.

METHODOLOGY

Six simulation scenarios were considered, assuming 100 percent market penetration of personal AV in the same horizon year as the base case scenario for the MTC-ABM: (i) Doubled highway capacity due to smaller vehicles, shorter headways, reduced accidents, and better operations, (ii) 25 percent driving time reduction, with passengers using in-vehicle time for work and leisure, (iii) 20 percent reduced operating costs from decreased insurance and fuel usage, (iv) new drivers, assuming full automation would increase mobility for older adults, people with disabilities, young people without driver’s licenses, and people living in poverty, (v) combined effects of all the scenarios, and (vi) road pricing plus the combined effects of all scenarios assuming the per mile operating cost is doubled (to 36 cents per mile), plus the AV effects of a doubling of roadway capacity, 25 percent reduction in value of driving time, and attraction of new drivers. For the AT service simulation and analysis, this study integrated the MTC-ABM with a dynamic traffic assignment (DTA) model to simulate the choice to use a personal auto as a Single Occupancy Vehicle (SOV) or an AT based on individuals’ value of time and the per mile cost of each mode. In the MTC-ABM combined with a regional agent-based simulation, the study compared three scenarios for 'first mile' transit services. These varied by the number of customers paying for the shared ride and the pickup location, with the SOV as the base case.

FINDINGS

Results from MTC-ABM Simulation:

  • The doubling of roadway capacity increased Vehicle Miles Traveled (VMT) by six percent in the peak period, one percent in the off-peak, and four percent for an average 24-hour daily period. However, for this scenario, the Vehicle Hours of Delay (VHD) decreased by 78 percent.
  • When the operating costs for automobiles was reduced by about 20 percent (or four cents per mile), the results indicated a shift from transit trips (four percent) and walk and bike trips (four percent) to drive-alone (one percent) and shared-ride (one percent) vehicle trips, with an associated decline in VMT and vehicle volumes by about three percent and two percent, respectively. 
  • The new driver scenario results indicated a significant increase in drive-alone vehicle trips (six percent on average daily) and similarly large reductions in shared-ride (five percent on average daily), transit (12 percent on average daily), and walk and bike trips (four percent on average daily). Despite these dramatic mode shifts away from transit, walk, and bike trips, VHD declined significantly by 70 percent on average daily.
  • The pricing and combined effects scenario appeared to provide the best system-level outcome in reduced VMT, VHD, and increased transit, walk, and bike travel. Specifically, on average daily, a seven percent reduction in VMT, an 84 percent reduction in VHD, a 22 percent increase in walk and bike trips, and a six percent increase in transit trips were observed.

Results from MTC-ABM Integrated with AT DTA:

  • Modal shift to use of AT appeared to be the highest (6.2 percent) when AT was offered at a lower cost, even with increased SOV parking and operational costs compared to the base case. However, the model might underpredict SOV's shift to ATs due to congestion obscuring parking cost savings and potential flexibility in traveler schedules.
  • AT travel time was up to 43 percent less than traditional transit travel times considering different cost levels for the modes across scenarios.

Results from MTC-ABM Combined with Agent Based Simulation:

  • Compared to the base case scenario, the shared-ride home pick-up scenario indicated generalized cost savings for 33 percent of trips with drivers and 73 percent of trips with driverless vehicles.
  • The average gain was found to be $1.50 and $2.00, respectively for human-driven and driverless AV.
  • The total system benefit for the driverless AV was estimated to be about 192 percent higher than that of human driver scenario ($114,000 compared to $39,000), due to lower per mile operating costs of driverless vehicles.

Automated Vehicle Scenarios: Simulation of System-Level Travel Effects Using Agent-Based Demand and Supply Models in the San Francisco Bay Area

Automated Vehicle Scenarios: Simulation of System-Level Travel Effects Using Agent-Based Demand and Supply Models in the San Francisco Bay Area
Source Publication Date
09/01/2018
Author
Rodier, Caroline; Miguel Jaller; Elham Pourrahmani; Joschka Bischoff; Joel Freedman; and Anmol Pahwa
Publisher
Prepared by Multi-University Team for University of California-Davis
Goal Areas
Results Type