Multimodal Mobility Modeling and Traffic Profiling in Cyber-Physical Systems

点击翻译按钮获取中文摘要

The main objective of this research work is to investigate the use of mobile devices along with cellular wireless networks and other sensing technology to estimate mobility and to build robust tempro-spatial mobility models. Transportation systems will be considered as a target application. Extracting a mobility model for individuals enables the profiling of traffic conditions and hence can facilitate a wide range of transportation applications: vehicle routing, optimized traffic lights signalling, road maintenance planning, and congestion management. Soft-Computing inferencing and filtering methods offer a range of capabilities for modeling mobility. However, since mobility is generally a non-linear dynamic process, a small non-linearity can lead to difficulties to represent the posterior knowledge. To investigate such challenges, Bayesian filtering and sequential Monte-Carlo methods such as particle filters will be employed. The Bayesian methods will be employed to tackle two open issues: a) wireless localization based on signal profiling (e.g, fingerprints and RSS; b) simultaneous localization and mapping. Comprehensive Sensing methods are explored to address the sparse signals issue present in such systems. A diverse range of data sources, such as cellular networks, cell-phones, road cameras, on-board GPS devices, loop detectors, and mobile augmented reality models, will be studied as complementary mobility cues. To exploit multi-modality to its utmost potential, the proposed research work will investigate several data fusion techniques for inferencing and estimating mobility and traffic trends. Here, the proposed research project will adopt a soft-computing as reasoning approach, where probabilistic techniques are used to capture trends from mobility data and fuzzy and Dempster-Shafer techniques offer linguistic paradigms for reasoning about uncertainty in the captured mobility and traffic data. Particles, Kalman, GMF, and HMM filters will be employed at various levels of the proposed modeling scheme; to address sparsity in the measurements Compressing Sensing techniques will be investigated.