SEASONS: Signal and Energy Aware Sensing on iNtermittent Systems

TL;DR

The paper tackles the problem of reconciling signal-aware adaptive sampling with energy-aware batteryless intermittent sensing. It proposes SEASONS, a framework that uses time buffering and sample queue dynamics to run adaptive sampling without requiring a large energy reservoir, thereby maintaining constant energy consumption while improving data accuracy. Experiments show SEASONS achieves about 31% accuracy improvement over traditional energy-aware intermittent systems and approaches battery-backed ASA performance under latency guarantees, with negligible overhead. This work enables practical, high-accuracy sensing on ultra-low-power intermittent devices and offers a first signal- and energy-aware interface for such platforms.

Abstract

Both energy-aware, batteryless intermittent systems and signal-aware adaptive sampling algorithms (ASA) aim to maximize sensor data accuracy under energy constraints in edge devices. Intuitively, combining both into a signal- & energy-aware solution would yield even better accuracy. Unfortunately, ASAs and intermittent systems rely on conflicting energy availability assumptions. So, a straightforward combination cannot achieve their combined benefits. Therefore, we propose SEASONS, the first framework for signal- and energy-aware intermittent systems. SEASONS buffers signal data in time, monitoring queue dynamics to ensure the data is reported within a user-specified latency constraint. SEASONS improves sensor data accuracy by 31% without increasing energy.

Figures (5)

• Figure 1: (a): Energy-Aware intermittent uniform sampling (purple) vs battery-backed ASA (orange) of accelerometer data over 6 seconds w/ 60% collection rate. The blue line (labeled ground truth) represents the real accelerometer signal sampled at 50Hz. (b): Behavior of sampling policies over a highly volatile phase. Adaptive sampling collects 10 more samples and achieves one order of magnitude lower error than uniform sampling. Since the adaptive method takes more samples, it consumes more energy than the uniform over this period. (c): Behavior of subsampling policies over a low volatility phase. The uniform strategy collects 10 more samples and achieves 2.2x lower error. Since the adaptive method takes less samples, it consumes less energy than uniform.
• Figure 2: Adaptive subsampling with time buffering and freshness enforcement by dropping samples.
• Figure 3: SEASONS's overview
• Figure 4: Accuracy improvement of SEASONS over energy-aware intermittent w/ uniform sampling. SEASONS improves the accuracy of intermittent systems and approaches the feasible accuracy of a battery-backed ASA.
• Figure 5: SEASONS Comparison between EI, SEB, & EI w/ SEASONS as latency constraint changes for a collection rate of 60%.