With the explosion of new mobile apps, protocols and services, there is a compelling need for new tools to evaluate and test mobile networks. This project aims to establish a systematic framework to produce benchmarks (called MobiBench) in the form of evaluation scenarios and test-suites for mobile networking protocols and services for user and vehicular mobility. Two threads of research are pursued. First, a multi-dimensional metric space is introduced for characterization of user (human) mobility, to be applied to mobility measurements and models. The metrics span: i. individual, ii. pair-wise and iii. collective mobility dimensions, and facilitate comprehensive analysis and evaluation of mobility models and networking protocols. Novel, trace-driven models are developed to accurately capture user mobility metrics and mobile network performance. The COBRA model is designed to capture collective communal behavior in mobile social networks. Second, new extensive vehicular traces are collected and analyzed, including vehicular imagery data from thousands of webcams around the world. Vehicular density distributions and models are developed to aid in establishing realistic vehicular mobility models and benchmarks. Estimation of vehicular density uses adapted background subtraction algorithms for image processing. This work will result in the establishment of a library of benchmarks, traces, models and test-suites for pedestrian and vehicular mobility, to be used for the realistic evaluation of future mobile protocols, networks and services. Results of this research will impact mobile networking research and aid in traffic control and transportation congestion mitigation.
