FROM TRADITIONAL TO AUTOMATED MOBILITY ON-DEMAND:
A COMPREHENSIVE FRAMEWORK FOR MODELLING ON-DEMAND SERVICES IN SIMMOBILITY

Mobility-on-demand (MoD) systems have recently emerged as a promising paradigm for sustainable personal urban mobility in cities. In the context of multi-agent simulation technology, the state-of-the-art lacks a platform that captures the dynamics between decentralized driver’s decision-making and the centralized coordinated decision making. This work aims to fill this gap by introducing a comprehensive framework that models various facets of MoD driver’s behavior along with a decentralized fleet management system within an agent-based demand-supply simulator, SimMobility Mid-Term. SimMobility Mid-Term (MT) simulator is an agent- and activity-based demand model integrated with a dynamic multimodal network assignment model. The traffic dynamics are simulated using a multi-modal mesoscopic simulator (supply simulator). MT is part of a much larger simulation platform that also contains long term and short-term models. Simulating MoD services is extremely challenging because of complex interactions between independent drivers, the central controller, and traveler’s decision processes. To facilitate the study of such a complex and partially decentralized system, we propose an event-based modelling framework. In this framework, the drivers, the controllers, and the travelers are represented as separate decision agents making plans and event-triggered actions. Behavioral models were estimated to characterize decision making of drivers using a GPS dataset from a major MoD fleet operator in Singapore in 2013, containing position and service status data over 30 days. A unified framework was developed to model the operation of both traditional MoD fleets and emerging ride-hailing services such as Uber, and Lyft-like services. While traditional MoDs actions are made by the driver, some MoD services can coordinate some of the processes above. Ultimately, automated MoD (AMoD) services could fully control and optimize all decisions making processes. The AMoD controller would for example process traveler’s service requests and assign them to a given vehicle after considering its current occupancy (and potential route), whether the passenger is willing to share the ride or not, time to reach to the pick-up location and the travel time to final destination. Thus, an MoD controller agent in simulation should capture MoD service status, updated vehicle locations and monitor vehicle movement through the network, reacting to incoming requests and changes on network and fleet performance accordingly. A framework to handle MoD controllers in SimMobility was integrated within the calibrated SimMobility model of Singapore. We demonstrate the benefits of the proposed framework through a large-scale case study in Singapore comparing the fully decentralized traditional MoD with the future AMoD services in a realistic simulation setting. The current work has five major contributions: 1. development of a comprehensive event-based framework that addresses complex behaviors and interactions of service drivers, MoD centralized operation, and travelers; 2. incorporation of the proposed framework within a highly realistic agent-based simulation platform, SimMobility; 3. evaluation of the suggested framework against real-world data; 4. demonstration of the proposed framework through a case study of Singapore; and 5. showcase the use of the proposed framework in the evaluation of potential mechanisms and policies when deploying MoD services.