No abstract

Two major factors affecting mobile network performance are mobility and traffic patterns. Simulations and analytical-based performance evaluations rely on models to approximate factors affecting the network. Hence, the understanding of mobility and traffic is imperative to the effective evaluation and efficient design of future mobile networks. Current models target either mobility or traffic, but do not capture their interplay. Many trace-based mobility models have largely used pre-smartphone datasets (e.g., AP-logs), or much coarser granularity (e.g., cell-towers) traces. This raises questions regarding the relevance of existing models, and motivates our study to revisit this area. In this study, we conduct a multidimensional analysis, to quantitatively characterize mobility and traffic spatio-temporal patterns, for laptops and smartphones, leading to a detailed integrated mobility-traffic analysis. Our study is data-driven, as we collect and mine capacious datasets (with 30TB, 300k devices) that capture all of these dimensions. The investigation is performed using our systematic (FLAMeS) framework. Overall, dozens of mobility and traffic features have been analyzed. The insights and lessons learnt serve as guidelines and a first step towards future integrated mobility-traffic models. In addition, our work acts as a stepping-stone towards a richer, more-realistic suite of mobile test scenarios and benchmarks.1 Throughout, we use flutes for smartphones, and cellos for laptops.

No abstract

Forecasting vehicular mobility and density is essential to a wide array of mobile applications, including VANETs, crowd-sourcing, participatory sensing, network provisioning, and shared transportation. Forecasting is intrinsically complex and scarcity and lack-ofscale of vehicular mobility data is adding to the challenge. In this paper, relying on traffic cameras as the main data acquisition tool and the traffic densities extracted from the images, we explore trends pertaining to density data for the purposes of temporal and spatial forecasting. We investigate the promise of deep learning by conducting a comparative analysis of conventional (seasonal) models, and multiple variants of recurrent neural models, based on 40 day-long traffic density data from 58 cameras in London. Our findings show a dramatic reduction in forecast error using deep learning, where the best seasonal model gets 0.0176 mean squared error, and our proposed neural model achieves 0.0067 (62% less error). This is 10.5% in percentage error, down from 19.3%. We also design an end-to-end multivariate architecture that forecasts all the cameras which achieves 0.0125 error (14.5% in percentage error), but is trained in half the time needed to train 58 cameras individually. Finally, to forecast locations without explicit monitoring, we build on these insights and investigate spatial relationships between cameras. We introduce a spatial forecast model similar to the multivariate model. This results in an average reconstruction error of 0.0169 when every camera is reconstructed based on only one camera, and goes down to 0.0125 when 8 cameras are used to predict others (on par with that of the multivariate model with all 58 cameras as input). Moreover, a set of 23 cameras is found that can forecast the other cameras with an error of 0.0086. These results provide great promise for prediction in future vehicular-based networks and services.

No abstract