Transportation is the single largest source of carbon emissions in the US. Decarbonizing it is challenging because it depends on individual behaviors, which in turn, depend on local land use planning. The interdisciplinary field of Computational Mobility, focusing on collecting, analysing and influencing human travel behavior, can frame solutions to this challenge.

Innovation flows in interdisciplinary fields are bi-directional. The flow to the domain is focused on building a strong foundation for methodological improvements. As the improvements are deployed, they result in use-inspired computational research. This temporal dependency results in our initial focus on the modularity, accuracy and reproducibility of e-mission, an extensible platform for instrumenting human mobility. This open source platform has a modular architecture that supports power efficient duty cycling using virtual sensors, a read-only data model and a pipeline with novel algorithm adaptations for smartphone sensing.

We also perform the first empirical evaluations of smartphone-based platforms in this domain. The architectural evaluation is based on three real world deployments: a classic travel diary, a crowdsourcing initiative, and a behavioral study. The accuracy evaluation is based on an novel procedure that uses artificial trips and multiple parallel phones to mitigate concerns over privacy, context sensitive power consumption and inherent sensing error. Data collected from three artifical timelines was used to evaluate the trajectory, segmentation and classification accuracies vs. power for various configurations.

On computational side, challenges derived from the deployments can contribute to ongoing CS research in privacy, trustworthiness, incentivization and decision making. On the mobility side, this enables methodological innovations such as Agile Urban Planning for prototyping infrastructure changes.