Content uploaded by Joseph A.E. Shaheen
Author content
All content in this area was uploaded by Joseph A.E. Shaheen on Jun 25, 2020
Content may be subject to copyright.
A preview of the PDF is not available
Simulating the Ridesharing Economy: The Individual Agent Metro-Washington Area Ridesharing Model (IAMWARM)
Abstract and Figures
The ridesharing economy is experiencing rapid growth and innovation. Companies such as Uber and Lyft are continuing to grow at a considerable pace while providing their platform as an organizing medium for ridesharing services, increasing consumer utility as well as employing thousands in part-time positions. However, many challenges remain in the modeling of ridesharing services, many of which are not currently under wide consideration. In this paper, an agent-based model is developed to simulate a ridesharing service in the Washington, D.C. metropolitan region. The model is used to examine levels of utility gained for both riders (customers) and drivers (service providers) of a generic ridesharing service. A description of the Individual Agent Metro-Washington Area Ridesharing Model (IAMWARM) is provided, as well as a description of a typical simulation run. We investigate the financial gains of drivers for a 24 hour period under two scenarios and two spatial movement behaviors. The two spatial behaviors were random movement and Voronoi movement, which we describe. Both movement behaviors were tested under a stationary run conditions scenario and a variable run conditions scenario. We find that Voronoi movement increased drivers’ utility gained but that emergence of this system property was only viable under variable scenario conditions. This result provides two important insights: The first is that driver movement decisions prior to passenger pickup can impact financial gain for the service and drivers, and consequently, rate of successful pickup for riders. The second is that this phenomenon is only evident under experimentation conditions where variability in passenger and driver arrival rates are administered.
Figures - uploaded by Joseph A.E. Shaheen
Author content
All figure content in this area was uploaded by Joseph A.E. Shaheen
Content may be subject to copyright.
ResearchGate has not been able to resolve any citations for this publication.
Casual carpooling is an informal form of commuter ridesharing operating in Washington, D.C.; Houston, Texas; and San Francisco, California. In contrast to new forms of shared-use mobility, casual carpooling has been in existence for over 30 years and uses no information communication technology, and is entirely run informally by its users. Researchers have been fascinated by this phenomenon and have conducted studies in the past, but there remains a lack of up-to-date quantitative data. This study examines the motivations and behaviors of casual carpoolers in the San Francisco Bay Area to understand user characteristics and motivations. In Winter 2014, the authors observed and counted participants and vehicles at four casual carpooling locations, interviewed participants riding in carpooling vehicles (N=16), and conducted intercept surveys (N=503) at 10 East Bay pickup locations. The results indicate that the motivations for casual carpooling participation include convenience, time savings, and monetary savings, while environmental and community-based motivations ranked low. Casual carpooling is an efficient transportation option for these commuters, while environmental sustainability benefits are a positive byproduct. Seventy-five percent of casual carpool users were previously public transit riders, and over 10% formerly drove alone. Logit modeling found that casual carpool role (i.e., always a rider or sometimes a driver), age, and employment status were key drivers in modal choice. Further research on a larger scale is needed to identify the elements needed for system replication in different areas.
In order to commute by carpooling, individuals need to communicate, negotiate and coordinate, and in most cases adapt their daily schedule to enable cooperation. Through negotiation, agents (individuals) can reach complex agreements in an iterative
way, which meets the criteria for the successful negotiation. The procedure of negotiation and trip execution in the long-term
carpooling consists of a number of steps namely; (i) decision to carpool, (ii) exploration and communication, (iii) negotiation,
(iv) coordination and schedule adaptation, (v) long term trip execution (carpooling), (vi) negotiation during carpooling and (vii) carpool termination and exploration for new carpool. This paper presents a conceptual design of an agent-based model (ABM) of a set of candidate carpoolers. A proof of concept implementation is presented. The proposed model is used for simulating the interactions between autonomous agents. The model enables communication to trigger the negotiation process; it measures the effect of pick-drop and shopping activities on the carpooling trips. Carpooling for commuting is simulated: we consider a set of two intermediate trips (home-to-work and work-to-home) for the long-term carpooling. Schedule adaptation during negotiation depends on personal preferences. Trip timing and duration are crucial factors. We carried out a validation study of our results with real data (partial) collected in Flanders, Belgium. Simulation results show the effect of constraining activities on the carpooling trips. The future research will mainly focus on enhancing the mechanisms for communication and negotiation between agents.
Dynamic ridesharing involves a service provider that matches potential drivers and passengers with similar itineraries allowing them to travel together and share the costs. Centralized (binary integer programming) and decentralized (dynamic auction-based multi-agent) optimization algorithms are formulated to match passengers and drivers. Numerical experiments on the decentralized approach provides near optimal solutions for single-driver, single-passenger cases with lower computational burden. The decentralized approach is then extended to accommodate both multi-passenger and multi-driver matches. The results indicate higher user cost savings and vehicle kilometers traveled (VKT) savings when allowing multi-passenger rides. Sensitivity analysis is conducted to test the impact of the service provider commission rate on revenue and system reliability. While short term revenue can be maximized at a commission rate of roughly 50% of each trip's cost, the resulting drop in system reliability would be expected to reduce patronage and revenues in the longer term.
Ridesharing offers the opportunity to make more efficient use of vehicles while preserving the benefits of individual mobility. Presenting ridesharing as a viable option for commuters, however, requires minimizing certain inconvenience factors. One of these factors includes detours which result from picking up and dropping off additional passengers. This paper proposes a method which aims to best utilize ridesharing potential while keeping detours below a specific limit. The method specifically targets ridesharing systems on a very large scale and with a high degree of dynamics which are difficult to address using classical approaches known from operations research. For this purpose, the road network is divided into distinct partitions which define the search space for ride matches. The size and shape of the partitions depend on the topology of the road network as well as on two free parameters. This allows optimizing the partitioning with regard to sharing potential utilization and inconvenience minimization. Match making is ultimately performed using an agent-based approach. As a case study, the algorithm is applied to investigate the potential for taxi sharing in Singapore. This is done by considering about 110 000 daily trips and allowing up to two sharing partners. The outcome shows that the number of trips could be reduced by 42% resulting in a daily mileage savings of 230 000 km. It is further shown that the presented approach exceeds the mileage savings achieved by a greedy heuristic by 6% while requiring 30% lower computational efforts.