TaxiSim: A Multiagent Simulation Platform for Evaluating Taxi Fleet Operations Shih-Fen Cheng School of Information Systems Singapore Management University Republic of Singapore sfcheng@smu.edu.sg Thi Duong Nguyen School of Information Systems Singapore Management University Republic of Singapore tdnguyen@smu.edu.sg Abstract—Taxi service is an important mode of public trans- portation in most metropolitan areas since it provides door-to- door convenience in the public domain. Unfortunately, despite all the convenience taxis bring, taxi fleets are also extremely inefficient to the point that over 50% of its operation time could be spent in idling state. Improving taxi fleet operation is an extremely challenging problem, not just because of its scale, but also due to fact that taxi drivers are self-interested agents that cannot be controlled centrally. To facilitate the study of such complex and decentralized system, we propose to construct a multiagent simulation platform that would allow researchers to investigate interactions among taxis and to evaluate the impact of implementing certain management policies. The major contribution of our work is the incorporation of our analysis on the real-world driver’s behaviors. Despite the fact that taxi drivers are selfish and unpredictable, by analyzing a huge GPS dataset collected from a major taxi fleet operator, we are able to clearly demonstrate that driver’s movements are closely related to the relative attractiveness of neighboring regions. By applying this insight, we are able to design a background agent movement strategy that generates aggregate performance patterns that are very similar to the real-world ones. Finally, we demonstrate the value of such system with a real-world case study. Keywords-multiagent simulation; urban transportation; driver behavior; mobility pattern; taxi fleet I. I NTRODUCTION Taxi service is an important mode of public transportation in most metropolitan areas (e.g., in Singapore, taxi rides accounted for around 17% of public transports in 2007/08), since it provides door-to-door convenience in the public domain. Unfortunately, despite all the convenience taxis bring, taxi fleets are also extremely inefficient. In an ordinary city, a taxi can easily spend 50% of its operation time idling (waiting in queues or roaming around empty). For cities that are getting increasingly crowded, inefficient taxi fleet not only offers lower quality of service than its potential would grant, it also creates negative impacts on environment and road congestion. As such, improving the efficiency of the taxi fleet operation is an important issue for government agencies and taxi fleet operators alike. Many past research efforts have been devoted to the modeling of the taxi fleet operations and also approaches that would improve the efficiency of taxi fleets. For example, advances in technologies like Global Positioning System (GPS) and communication networks enable advanced dis- patch system to be deployed [1], [2]. On the other hand, a series of work conducted by Yang et al. [3], [4] provides a good framework for understanding the equilibrium proper- ties of taxis in a network at the macro level. However, by reviewing these past works (which are mostly published in the transportation literature), we notice that there are very few attention paid to the decentralized nature of the taxi system. One exception is the design of taxi dispatch systems, where we do see the application of multi-agent technologies [5], [6]; nonetheless, taxi dispatch is only one possible mode of operations (the other more dominant modes being street pick-ups and queueing), and a comprehensive study that covers all modes of operations from a decentralized perspective is still not seen. Such decentralized perspective is critical in modeling taxi fleets because taxis can only be incentivized or coordinated and not centrally controlled. With proper models in place, not only can we improve the efficiency of current taxi fleets, a range of new services could be designed and evaluated as well (e.g., efficient cab- sharing service for serving last-mile travels between desired destinations and the closet public transport hubs). In this paper, we propose to build a multi-agent-based simulation platform, TaxiSim, to simulate the operation of taxi fleets. TaxiSim is designed to be capable of modeling individual taxi driver’s strategies at micro level, and it’s also designed to be scalable so that it can simulate thousands of taxis simultaneously. Real-world operational data, if avail- able, can also be imported to TaxiSim, and this allows us to construct a highly realistic simulation environment. This would allow researchers and policy makers to study and evaluate potential mechanisms, policies, and new services for improving taxi services. II. BACKGROUND Since the early days of digital computers, simulations have played an important role in transportation research. In all major areas of transportation studies, be it traffic signal control, traffic assignment (routing), or even regional planning, simulations are all involved deeply. With rapid development of computing technology, simulations have