Energy-Adaptive Checkpoint-Free Intermittent Inference for Low Power Energy Harvesting Systems

2025-03-09

AI Summary

To address high latency, accuracy fluctuations, and service unreliability in DNN inference on energy-harvesting (EH) devices caused by frequent power failures, this paper proposes a checkpoint-free intermittent inference framework. Our method introduces: (1) an energy-adaptive dynamic complexity reduction mechanism, integrating elastic-width networks with energy-aware scheduling; and (2) a state-streaming cache and interruption-resilient inference engine to ensure continuous progress preservation. Evaluated on real EH hardware, the framework achieves a 98.2% inference completion rate, reduces end-to-end latency by 47%, and limits accuracy variation to under 0.3%, significantly enhancing QoS robustness. To the best of our knowledge, this is the first work to realize checkpoint-free, low-overhead, and highly reliable DNN inference on EH-constrained edge devices.

Technology Category

• Machine Learning: Hardware-aware ML
• Intelligent Robots: State Estimation
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Deep neural network (DNN) inference in energy harvesting (EH) devices poses significant challenges due to resource constraints and frequent power interruptions. These power losses not only increase end-to-end latency, but also compromise inference consistency and accuracy, as existing checkpointing and restore mechanisms are prone to errors. Consequently, the quality of service (QoS) for DNN inference on EH devices is severely impacted. In this paper, we propose an energy-adaptive DNN inference mechanism capable of dynamically transitioning the model into a low-power mode by reducing computational complexity when harvested energy is limited. This approach ensures that end-to-end latency requirements are met. Additionally, to address the limitations of error-prone checkpoint-and-restore mechanisms, we introduce a checkpoint-free intermittent inference framework that ensures consistent, progress-preserving DNN inference during power failures in energy-harvesting systems.

Problem

• Addresses DNN inference challenges in energy-harvesting devices.
• Reduces computational complexity during low energy availability.
• Ensures consistent DNN inference without checkpointing during power failures.

Innovation

• Energy-adaptive DNN inference for low-power mode
• Checkpoint-free intermittent inference framework
• Dynamic computational complexity reduction