Robustness of predictive energy harvesting systems: Analysis and adaptive prediction scaling

Internet of Things (IoT) systems can rely on energy harvesting to extend battery lifetimes or even render batteries obsolete. Such employ an scheduler optimise their behaviour and thus performance by adapting the system's operation. Predictive models sources, which are inherently non-deterministic consequently challenging predict, often necessary for performance. Because inaccurate predictions utilised scheduler, predictive model's accuracy inevitably impacts system This fact has largely been overlooked in vast amount available results schedulers predictors harvesting-based systems. The authors systematically describe effect prediction errors have defining a novel robustness metric. To alleviate severe impact performance, propose adaptive scaling method that learns from local environment behaviour. demonstrate concept with datasets both outdoor indoor scenarios. In addition, highlight improvement overhead proposed It improves non-robust up 13.8 times real-world setting.