Energy-optimal partial computation offloading using dynamic voltage scaling

Abstract

The incorporation of dynamic voltage scaling (DVS) technology into computation offloading could further reduce energy consumption of smart mobile device (SMD), which has not been fully studied. In this paper, this issue is formulated with jointly considering computational speed of SMD, transmit power of SMD, and the partial offloading, resulting in a nonconvex problem. To circumvent such a problem, we firstly study the thresholding when the offloading decision could always obtain positive gain. Then to yield more gains from offloading, we propose an alternating algorithm, which converges to a locally optimal solution fast. Extensive simulations show that the proposed algorithm not only significantly reduces energy consumption but also ensures more stringent latency against the existing schemes.