PYRAMID: Probabilistic Content Reconciliation and Prioritization for V2V Communications

Emerging vehicle-to-vehicle communication technologies, such as dedicated short range communications (DSRC), offer unique opportunities to realize wireless peer-to-peer systems for vehicles. A critical component in a vehicular peer-to-peer system is an efficient content reconciliation mechanism, which guides two communicating vehicles to match their interests, prioritize task execution, and ensure redundant contents not to be exchanged. In this paper, we propose PYRAMID, a multilayer probabilistic abstraction framework, to efficiently abstract and approximate contents with different granularity. Using the multilayer Pyramid data structures, a vehicle is able to quickly get an impression of the contents on the other vehicle before the costly massive content exchange process starts. Particularly, the coarse-granularity layer estimates the contribution from potential transaction partners so that tasks could be prioritized accordingly; the fine-granularity layer helps conduct membership tests, to avoid transmitting redundant contents. Using a fleet of research vehicles equipped with 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.