Enabling Vehicular Applications using Cloud Services through Adaptive Computation Offloading

There is growing interest in embedding new class of applications in vehicles to improve the user driving experience. However, the limited computational and storage resources in vehicles brings about a challenge of running computation and data intensive tasks of such applications in the vehicle's on-board unit (OBU). Moreover, embedded applications may not be easily updated by replacing hardware as upgrades in the vehicle OBUs can only happen over each vehicular life-cycle, which is of the order of 10-15 years. The advent of connectivity of vehicles to the Internet offers the possibility of offloading computation and data intensive tasks from the OBU to remote cloud servers for efficient execution. In this paper, we propose a novel architecture for bringing cloud-computing to vehicles where applications embedded in the vehicle OBU can benefit from remote execution of tasks provided as services in the cloud. We design a framework to identify and adaptively manage offloading of computation and data intensive tasks from the vehicle OBU to the cloud during application run-time. Through experimental evaluation using a preliminary prototype implementation of two computer vision applications that use our framework, we show that our approach can provide at least 3x reduction in the end-to-end application response time.