Green service over Internet of Things: a theoretical analysis paradigm

摘要

There are two kinds of uncertainty for providing green service over Internet of Things (IoT): network service period and user satisfaction model. In this paper, we consider a power-aware service problem over IoT where both of the uncertainties are incorporated. Specifically, we consider a generic IoT's service scenario: a server provides different kinds of services without knowledge of user satisfaction model and network service period. Our objective aims at dynamically adjusting the power allocation for each service over a uncertain period to maximize expected user satisfaction. It should be noted that practical user satisfaction rate is observed over time, but the inherent functional relationship between the power and satisfaction rate is unknown. In order to present a quantitative analysis, we consider a general user satisfaction model belonging to a class of functions that do not deploy any parametric representation. In this case, a blind dynamic powering algorithm is developed, in which one learns the satisfaction function and optimizes power-aware user satisfaction with on-line operation. More precisely, the algorithm performance is measured in terms of regret which denotes the satisfaction loss compared to the optimal satisfactions that can be obtained when the service period and satisfaction rate are known. Moreover, a tight bound on this regret is proposed for any possible powering policy, and we show that the proposed algorithm can achieve a regret that is close to this bound.