Online optimization of dynamic power management

This paper mainly considers the application of Semi-Markov decision process (SMDP), which is a stochastic discrete event dynamic system (DEDS), in the dynamic power management (DPM) problems. There are the DPM problems in many portable devices, which can provide satisfactory performance with minimum power consumption by selectively turning off system components, which enter the idle state, to save the energy consumption and prolong the service time. Firstly, the SMDP model of DPM is introduced and then an online optimization approach with forbidden time is presented. This approach can learn to improve the policy of DPM by analyzing the data from the device itself and make the device have an individual power management policy. The optimization can be implemented during the charging and thus does not need the cloud transportation and computation, which avoids the leak of private data. Finally, a simulation example is applied to illustrate the applicability of the approach.