On minimizing expected energy usage of embedded wireless systems with probabilistic workloads

A large number of embedded wireless systems must handle complex and time-varying computational and communication workloads. Further, a significant number of these systems support real-time applications. Most of the existing energy management studies for such systems have focused on relatively simple scenarios that assume deterministic workloads, and only consider a limited range of energy management techniques, such as Dynamic Voltage Scaling (DVS). Our paper addresses these deficiencies by proposing a general purpose probabilistic workload model for computation and communication. To account for the importance of radio energy consumption, we also analyse Dynamic Modulation Scaling (DMS), an often overlooked method for energy management. We define several energy control algorithms, including an optimal combined DVS-DMS approach, and evaluate these algorithms under a wide range of workload values and hardware settings. Our results illustrate the benefits of joint power control algorithms.