Poster: Redundancy Aided Vehicular Networking

Vehicular applications are increasingly connected to cloud services. For example, route planning, gas price applications, and Siri-like personal assistants all respond to user queries based in part on cloud processing. Network communication thus is often a substantial component of user-perceived latency in vehicular applications. Current vehicular computing platforms typically connect to the cloud using a single cellular network provider. Network conditions can change rapidly as a vehicle moves due to geographical variation in coverage, radio shadows, and differing traffic density. Such variation is often exacerbated by connection re-establishment after an interface has entered a sleep state. Thus, vehicular applications can often appear unresponsive due to high wireless network latency. Even worse, the responsiveness is unpredictable; high tail latency makes some user interactions take longer, even when most interactions complete in an acceptable amount of time. This unpredictability is especially worrisome in a vehicular environment, in which occasional unexpected performance anomalies distract the driver of the vehicle.