Trajectory Simulation in Communities of Commuters Ashish Dandekar * , St´ ephane Bressan * , Talel Abdessalem † , Huayu Wu ‡ and Wee Siong Ng ‡ * School of Computing, National University of Singapore, Singapore Email: ashishdandekar@u.nus.edu, steph@u.nus.edu † T´ el´ ecom ParisTech, Paris-Saclay University, Paris, France Email: talel.abdessalem@telecom-paristech.fr ‡ Institute of Infocomm Research, A*STAR, Singapore Email: huwu@i2r.a-star.edu.sg, wsng@i2r.a-star.edu.sg Abstract—Urban planning, development and management au- thorities and stakeholders need to understand and analyse the mobility patterns of urban dwellers in order to manage sociological, economic and environmental issues. Simulation is indispensable a tool for authorities and stakeholders to better design, operate and control the mobility infrastructures of smart cities. We propose an approach for the simulation of trajectories in communities of commuters. We identify communities of public transport commuters from historical automated fare collection card data using spatial latent Dirichlet allocation. We further aggregate the historical automated fare collection card data to create statistical models of visits and movements of commuters in each community. We use statistical models to simulate trajectories of synthetic individual commuters. We empirically evaluate how the synthetically generated trajectories are typical of their community of commuters and realistic. I. I NTRODUCTION Urban planning, development and management authorities and stakeholders need to understand and analyse the mobility patterns of urban dwellers in order to manage sociological, economic and environmental issues. Urban development au- thorities need to study mobilities of commuters so as to foresee mobility infrastructure requirements. Simulation serves as a faithful means to help in the study of mobility. Human mobility in urban areas is the movement of an individual as much as the dynamics of communities moving from one place to the other. An urban area comprises of different residential areas from where people commute to work places on daily basis. A faithful simulation of human mobility in the urban setting should be able to not only capture spatiotemporal patterns of movements of individual commuters but also adhere to latent patterns of movement as a member of communities. In this work, we propose an approach to simulate trajectories of commuters which are typical of their community structure. We apply spatial adoption of Latent Dirichlet Allocation to historic automated fare collection card data of a public transportation network to find latent communities of com- muters. We use the learned community structure to synthesize the commuter transportation graph for each community. We further propose a mechanism which uses random walk on the graph along with latent topics learned from LDA to generate trajectories for a synthetic individual. Experiments show that simulated trajectories conform to the underlying hidden community structure. The rest of the paper is organized as follows. Section II delineates related work. Section III presents the proposed mechanism to generate trajectories. We present experiments and evaluation in Section IV. We conclude the paper by discussing the work underway in Section V II. RELATED WORK Related work spans three different domains of research, namely Urban Computing, Latent Dirichlet Allocation (LDA) and Human Mobility. Urban Computing [1] is a process of acquisition, integra- tion, and analysis of big and heterogeneous data generated by diverse sources in urban spaces, such as sensors, devices, vehicles, buildings, and humans, to tackle the major issues that cities face. It also helps to understand the latent trends and foresee the development of the city. Public transport data has been studied by authors for the betterment of the mobility of the citizens of the cities. In [2]–[4], London public transport has been widely studied for exploring traveling behaviors, minimizing traveling time and finding communities of citizens. Ferris et al. [5] have developed an app OneBusAway to reduce the waiting time of commuters by providing real-time bus arrival information in King county, Washington. To identify tourists from the daily commuters of the public transportation services Xue et al. [6] have devised a method and tested it on the data from Singapore. Montis et al. [7] have found communities of commuters to help in the demarcation of the sub-regions in Sardinia. In [8], authors have used to spatio- temporal data generated from social networks to find mobility patterns in an urban area. Blei et al. [9] have proposed Latent Dirichlet Allocation (LDA) - a soft clustering technique used for finding latent topics by intuitively capturing the co-occurrence of the words in the textual corpora. Till the date, it is a widely used