Querying Large Vehicular Networks: How to Balance On-Board Workload and Queries Response Time?

Data analysis plays a key role in designing today's Intelligent Transportation Systems (ITS) and is expected to become even more important in the future. Connected vehicles, one of the main instantiations of ITS, produce large volumes of data that are difficult to gather by centralized analysis tools. The even larger volumes of data expected from autonomous driving will further exacerbate the bottleneck problem of data retrieval. When analysts issue queries that seek data from vehicles satisfying certain criteria (e.g. those driving above a certain speed or in a certain area), the problem can partially be overcome by pushing to vehicles themselves the job of checking and reporting the compliance of their local data (e.g. recorded GPS positions or CAN data), hence avoiding a costly data retrieval phase. The problem we tackle in this work consists in spreading a set of such queries over a vehicular fleet while balancing the time needed to resolve the queries and the computational overhead induced on the vehicular network. We present in this work efficient and configurable query-spreading algorithms tailored for vehicular networks. Our tunable algorithms, which we evaluate on two large sets of real-world vehicular data, outperform baseline solutions and are able to balance the trade-off between the overall on-board workload and the response time needed to receive all answers for a set of queries.