Dynamic power management by reinforcement learning

Abstract

Optimization in design and utilization of both hardware and software is needed in order to achieve more energy efficient systems. In this paper we presented a Reinforcement learning based DPM approaches for our LAN card power management system. The presented approaches do not require priori model of the system as an Opposite to the existing DPM approaches. Thesis outcomes also show that sleeping is indeed feasible in the LAN and in some cases, with very little impact on other protocols. Moreover, reinforcement learning is a machine intelligence approach that has been applied in many different areas whereas Qlearning is one of the most popular algorithms that perform reinforcement learning. At last, with the desired outcomes of this thesis work, power management issues of LAN card system were solved effectively. In future we aim to compare DPM problem with mission learning problem. The RL based learning algorithm can then be implemented to find the right value of power constraint.