Study shows that utilizing distributed parking facilities can improve the overall performance of an Automated Mobililty-on-Demand system

To better capture the behavior and dynamics of travel patterns, a multi-agent modeling approach in a microscopic simulation framework is used to research how different fleet sizes and facility locations influence the performance of an AMOD system.

Date Posted
01/06/2017
Identifier
2016-B01124
TwitterLinkedInFacebook

Autonomous Mobility on Demand in SimMobility: Case Study of the Central Business District in Singapore

Summary Information

This paper introduces an extension for SimMobility—a high-fidelity agent-based simulation platform— for simulating and evaluating models for Autonomous mobility on demand (AMOD) systems. As a demonstration case study, preliminary simulations were designed to evaluate the effect of a new policy restricting private vehicle usage within in the high-traffic Central Business District (CBD) in Singapore, with AMOD being introduced as an alternative mode of transportation.

Methodology

This extension was used to explore the effects of different fleet sizes on customer waiting times for two models:

  • 1. A station-based model where cars self-drive back to stations
    2. A free-floating model where cars self-park at drop-off locations. The simulations were run for the period of 2 hours during evening peak (5:00PM to 7:00PM).

Demand Generation: For this study the SimMobility model assumes all private vehicle trips as a combined modal trip (i.e., Private vehicle + AMOD) if part of the trip is inside the CBD. The mode choice model is modified by making it sensitive to AMOD waiting times and additional cost terms. Further parking prices for private vehicles is reduced as now they have been parked outside the CBD region. For the base case, the total number of AMOD trips for the simulated period was 28,525 trips.

Facility Location and Fleet Sizes: In the station-based model, sets of 10, 20, 30 and 40 high-demand facility locations were analyzed. There was no capacity constraint on the facilities, i.e., the facility could hold as many cars as required. In the free-floating model, initial stations were assumed in the same manner as for the station-based model; however, in the free-floating model, cars were not required to return to these stations. Twelve different fleet sizes were simulated, i.e., from 2000 to 7500 AMOD vehicles in the system. At the beginning of the simulation, vehicles were uniformly distributed over the facilities.

Results

The free- floating model was compared against the station-based model with a varying number of facilities and the effect of different fleet sizes on the performance of AMOD system was assessed.

Number of Customers Served

  • In both models, increasing the vehicle fleet size resulted in a linear increase in the number of passengers served. In the free floating scheme, every additional 100 cars provisioned increased the average demand served by 3.7 percent (1055 people-trips). For the station-based model, this increase was smaller at 2.2 percent (627.55 people-trips).
  • The free-floating model was able to serve 90% of the demand, significantly more than the station-based model (68% of the requested trips). The low service rate in station-based model was likely caused by heavier traffic due to empty vehicle rides.
  • The above is consistent with the average travel time (which can be seen as a proxy metric for road congestion) of both models. The average travel time in the station-based model was higher on average, e.g., with 40 stations and 7500 vehicles the average travel time for the station-based model was 14.17 minutes, · 30% higher than in the free-floating model (10.59 minutes).

Customer Waiting Time Analysis

  • As expected, increasing the AMOD fleet size resulted in a fall in waiting times, since more vehicles were available to service the requested trips. For example, with 20 initial stations, the median waiting time decreased from 20.74 to 1.80 minutes as the fleet size grew from 2000 to 7500 (similarly, the variance in the waiting times decreased from 31.38 to 6.09).
  • Unlike the effect on total demand served, this waiting time change is non-linear and shows diminishing returns—the rate of improvement decreases with increasing fleet size and appears minimal beyond 6000 vehicles.
  • The initial distribution of vehicles (i.e. at the beginning of the day) also influenced the performance of the system; increasing the number of initial stations decreased passenger waiting times. The biggest difference is between 10 and 20 stations, where we observed an average improvement of approximately 4 minutes across fleet sizes. However, further increases in the number of stations resulted in only minimal decreases in waiting times (< 1.5 minutes).

Autonomous Mobility on Demand in SimMobility: Case Study of the Central Business District in Singapore

Autonomous Mobility on Demand in SimMobility: Case Study of the Central Business District in Singapore
Source Publication Date
06/28/2015
Author
Katarzyna Anna Marczuk, Harold Soh Soon Hong, Carlos Miguel Lima Azevedo, Muhammad Adnan, Scott Drew Pendleton, Emilio Frazzoli and Der Horng LeeNational University of SingaporeSingapore-MIT Alliance for Research and TechnologyMassachusetts Institute of Technology
Publisher
IEEE
Goal Areas
Results Type
Deployment Locations