PYRAMID: Informed content reconciliation for vehicular peer-to-peer systems

In vehicular P2P systems, content reconciliation is crucial since it guides two communicating vehicles to match their interests, prioritize task execution, and ensure redundant contents not to be exchanged. We propose PYRAMID, a probabilistic abstraction framework, to efficiently abstract and approximate content with different granularity. Particularly, coarse-granularity sketches estimate the contribution from potential transaction partners so that tasks could be prioritized accordingly; fine-granularity summaries help conduct membership test, to avoid transmitting redundant contents. Using a fleet of research vehicles equipped with Dedicated Short Range Communication (DSRC) radios, we experimentally demonstrate that, across a rich variety of scenarios, PYRAMID improves the utility value of content exchanges by 20%-30% and improves effective throughput by at least 25%, while only incurring a minimal computational overhead.