Adaptive energy budgeting for atomic operations in intermittently-powered systems

Intermittently-powered energy-harvesting systems offer the potential to form a long-lasting and environmentally friendly Internet of Things. Atomic operations in such systems refer to those that should be executed without interruption, e.g. wireless transmission; if interrupted by power loss, they must restart rather than resume from the interrupted point. State-of-the-art approaches handle atomic operations by either taking a checkpoint before the operation, or ensuring sufficient energy is buffered through design-time profiling of the required energy. Such approaches become inefficient or fail if energy consumption or storage changes, e.g. due to aging or environmental factors. This paper proposes a method for online profiling of energy budgets for atomic operations, enabling intermittent systems to adapt to changes in environment or system conditions. We show that an existing approach fails if capacitance drops by 66%, whereas our approach successfully adapts to changes and ensures atomicity as long as the required energy can be stored