Adaptive sampling with strict energy constraints

With more and more battery-operated devices equiped with sensors, it is becoming increasingly relevant to focus not only on energy efficiency, but also on ensuring that the available energy is not spent before the battery will be recharged. In this thesis a strategy for adaptive sampling is proposed that adapts the sample rate while respecting a predefined limit on the number of samples per discharge period. The proposed sample strategy aims to sample with uniform prediction errors, as opposed to uniform intervals. This signalagnostic strategy is tested upon data from traffic flow, a wireless sensor node, and a position trace obtained with a mobile phone. A reduction in the root mean squared error ranging from 0 to 50 percent compared to uniform sampling was achieved.