Section: New Results

Online Energy Optimization in Embedded Systems

We have used a Markov Decision Process (MDP) approach to compute the optimal on-line speed scaling policy to minimize the energy consumption of a single processor executing a finite or infinite set of jobs with real-time constraints. We provide several qualitative properties of the optimal policy: monotonicity with respect to the jobs parameters, comparison with on-line deterministic algorithms. Numerical experiments in several scenarios show that our proposition performs well when compared with off-line optimal solutions and out-performs on-line solutions oblivious to statistical information on the jobs [33]. Several extension to online learning (Q-learning) as well as hidden Markov chain theory for offline computation of the statistical parameters of the system are currently being investigated.