Analysis of Mobility Patterns for Urban Taxi Cabs Mohammad Asadul Hoque 1 ; Xiaoyan Hong 2 ; Brandon Dixon 3 Department of Computer Science The University of Alabama Tuscaloosa, USA {mhoque 1 ; hxy 2 ; dixon 3 }@cs.ua.edu Abstract—This paper analyzes urban taxi mobility traces obtained from San Francisco Yellow cabs. The paper presents a rigorous analysis of taxi mobility pattern with the instantaneous velocity profile, spatio-temporal distribution, connectivity of vehicle communications, clustering, hotspots and other characteristics like trip duration and empty cruise interval. The empirical data analyses presented here can be a helpful resource for wireless researchers, government organizations, taxi companies and even for the drivers or passengers. While wireless researchers can estimate the capabilities and constraints of vehicular communication from connectivity and mobility patterns, government can plan and work on issues related to implementing proper DSRC infrastructure. Finally, taxi companies and drivers can benefit from maximizing the trip revenue and minimize empty cruise time though balanced loaddistribution and awareness of the hotspots. (Abstract) Keywords-Hotspots; Cruise time; connectivity; traffic trace; Clustering, Partitioning, V2V, Hotspot. I. INTRODUCTION The GIS based computer aided taxi dispatching (CAD) systems provide an easy way to track the movement of each individual taxi and monitor the occupancy status of the vehicle. In order to distribute the load fairly among the fleet, it is nevertheless important for the taxi companies to have a prior idea about the demand and availability statistics based on historical data that can be generated from the archived GPS trace records in their systems. On the other hand, to maximize daily trip revenue and minimize empty cruise time, it is also necessary for the drivers to have a sound idea about the geographical distribution of taxi hotspots for passenger pickup and drop off which varies along time. More important, the historical archived data of mobility traces can provide significant information, such as geographical distribution and time varying density of the road traffic, for helping vehicle communications, for implementing Intelligent Transportation Systems applications, and for planning of deploying DSRC infrastructure. Recent research have shown studies on many interesting facts related to Vehicular Ad hoc Network (VANET) like urban mobility models, vehicle-to-vehicle (V2V) connectivity etc. A remarkable initiative of San Francisco Exploratorium [4] is the Cabspotting project [5], which is intended as a living framework to use the activity of commercial cabs to explore the economic, social, political and cultural issues that are revealed by the realistic GPS traces. In this paper, we present our analysis on the traces available through this project provided by San Francisco Yellow Cabs[3]. Our analysis dealt with 536 cabs generating over 10 million mobility traces over a period of one month. Our results show new interesting factors about taxi cab mobility, passenger data and communication potentials. Here our analysis of taxi mobility pattern emphasizes the following characteristics: i) Instantaneous velocity profile ii) Spatio-temporal distribution of cabs iii) Frequency distribution of pickup and drop off iv) Identification of hotspots v) Trip duration and empty cruise interval vi) V2V connectivity vii) Network partitions and Clustering The subsequent sections are organized as follows: We discuss related works in Section II, followed by our analysis model and data collection methodology in Section III. Sections IV and V presents the results for a single cab and monthly averages for whole fleet respectively. Section VI provides a detail analysis on vehicle connectivity as well as clustering of the mobile nodes. Finally, we conclude in Section VII. II. RELATED WORK Several interesting works related to taxi mobility patterns has been addressed by the researchers. Most of these works are based on analyzing GPS traces from different taxi cab companies to explore hidden characteristics of urban mobility models. Some of these researchers tend to reveal new mobility models while others focus on clustering and hot spot identification. Piorkowskiet. al[9] utilized the Cabspotting data archived over a month to propose a parsimonious mobility model called Heterogeneous Random Walk (HRW) which captures some of the important mobility characteristics observed from the macroscopic level. A key feature of the model is that nodes follow independent and statistically equivalent mobility patterns, despite the presence of long-term clusters. They also evaluate the predictive power of the HRW model in the context of epidemic dissemination, which is one of the most prominent paradigms for routing in DTNs. Their work motivates the vehicular networking community to deeply investigate the taxi mobility traces for further research. Shin et al [8] used real-life location tracking data collected from the Taxi Telematics system developed in Jeju, Korea. Their analysis aimed at obtaining meaningful moving patterns of taxi cabs .They have extracted some interesting statistical factors such as taxi’s driving type, driving time, driving area, pickup rate etc. Lee et. al [7] analyzed a pick-up pattern of taxi