Energy-Aware Computation Offloading in Wearable Computing

Abstract

Wearable technology is an emerging computing paradigm with a broad range of attractive applications. However, wearable devices have limited compute capability, storage and battery lifetime, which hinder their use for compute-intensive applications. In this paper, we propose a lightweight computation offloading technique in which some of the workload is transferred to nearby more resourceful mobile devices in order to enhance performance and save energy of wearable devices. The computation-offloading problem is modeled as heterogeneous system scheduling problem, where an application is represented as a Directed Acyclic Graph (DAG) and devices (wearable and mobile) are considered as heterogeneous computing system. A key feature of the proposed list scheduling technique that it intelligently exploits task duplication to achieve both the objectives of saving energy consumption and enhancing performance. The effectiveness of the proposed technique is demonstrated by comparing the results with a well-known technique for randomly generated DAGs. Experimental results reveal that the proposed approach provides a significant reduction in energy and schedule length over existing approach.