Stochastic Control of Computation Offloading to a Dynamic Helper
Yunzheng Tao, Changsheng You, Ping Zhang, Kaibin Huang · 2018
Exploiting excessive idling computation resources in massive mobile devices (e.g., IoT devices, smart phones) can effectively increase the computing capability of the edge networks. This paper designs energy-efficient control policies in a computation offloading system with a random channel and a helper with a dynamically loaded CPU (due to the primary service). Specifically, the aim of the policies is to determine the sizes of offloaded and locally computed data for a given input-data in different time slots such that the total energy consumption for transmission and local CPU is minimized under a task-deadline constraint. As the result, the polices endow an offloading user robustness against channel-and-helper randomness besides balancing offloading and local computing. By modeling the channel and helper CPU as Markov chains, the problem of offloading control is converted into a Markov decision process. We show that the optimal policies can be obtained by dynamic programming (DP). Furthermore, we consider two special cases that the user has no buffer and a large buffer. For the case of zero buffer, the optimal policy derived in closed form by backward induction satisfies proportional data partitioning in different time slots. For the case of a large buffer, a suboptimal policy is derived by applying the approximation methods, which shares the similar policy structure with the counterpart of zero buffer. Simulation results show the performance gain of the proposed policies compared with simple baseline algorithms.