Adaptive vehicular computation offloading scheme for connected car applications in cloud environments

Abstract

Connected car service is one of the most popular buzzwords in these days. Through accessing computing resources from cloud computing systems, it is possible for low specification computing resource in vehicle to use various and computation-intensive applications. This technology is termed as computation offloading and many offloading methods have been researched in mobile cloud. However, most of the schemes in mobile cloud cannot consider performance degradation due to workloads in cloud resources. Also to apply mobile cloud method to vehicular environments, there is a problem, implying that high mobility causes frequent handoffs in cellular network and when vehicle offloads computation to cloud with handover, predicted execution time measured by Round Trip Time (RTT) is not static because of dynamic network throughput change by several communication factors. In this thesis, to solve the problems encountered in offloading process of vehicular computation, Adaptive Vehicular Computation Offloading Scheme (AVCOS) is proposed. Main connected car applications are introduced and overall vehicular computation offloading architecture for connected car applications is suggested. Especially, we model the offloading procedure in Long-Term Evolution (LTE) vehicular network environments. Proposed scheme is explained and evaluated. Through performance evaluation using integrated simulator for vehicular computation offloading, we profile applications of voice processing (voice command, voice query) and video processing (240p, 1080p). Also we identify that the proposed scheme, representative offloading method in mobile cloud, and simulated offloading result are compared and analyzed. As a result, the proposed adaptive vehicular offloading scheme makes decision according to application types, network configuration, and computing performance, and also the estimated execution times of the proposing method shows similar results from simulated offloading outcomes.