Agent-Based Modeling of Taxi Behavior Simulation with Probe Vehicle Data

• Main Authors: Saurav Ranjit, Apichon Witayangkurn, Masahiko Nagai, Ryosuke Shibasaki
• Format: Article
• Language: English
• Published: MDPI AG 2018-05-01
• Series: ISPRS International Journal of Geo-Information
• Subjects: agent-based modeling and simulation; origin destination; taxi demand; taxi free movement; index and search; big data; distributed computing
• Online Access: http://www.mdpi.com/2220-9964/7/5/177

Taxi behavior is a spatial–temporal dynamic process involving discrete time dependent events, such as customer pick-up, customer drop-off, cruising, and parking. Simulation models, which are a simplification of a real-world system, can help understand the effects of change of such dynamic behavior. In this paper, agent-based modeling and simulation is proposed, that describes the dynamic action of an agent, i.e., taxi, governed by behavior rules and properties, which emulate the taxi behavior. Taxi behavior simulations are fundamentally done for optimizing the service level for both taxi drivers as well as passengers. Moreover, simulation techniques, as such, could be applied to another field of application as well, where obtaining real raw data are somewhat difficult due to privacy issues, such as human mobility data or call detail record data. This paper describes the development of an agent-based simulation model which is based on multiple input parameters (taxi stay point cluster; trip information (origin and destination); taxi demand information; free taxi movement; and network travel time) that were derived from taxi probe GPS data. As such, agent’s parameters were mapped into grid network, and the road network, for which the grid network was used as a base for query/search/retrieval of taxi agent’s parameters, while the actual movement of taxi agents was on the road network with routing and interpolation. The results obtained from the simulated taxi agent data and real taxi data showed a significant level of similarity of different taxi behavior, such as trip generation; trip time; trip distance as well as trip occupancy, based on its distribution. As for efficient data handling, a distributed computing platform for large-scale data was used for extracting taxi agent parameter from the probe data by utilizing both spatial and non-spatial indexing technique.