Energy-Aware AI-Driven Framework for Edge-Computing-Based IoT Applications

Abstract

The significant growth in the number of Internetof-things (IoT) devices has given impetus to the idea of edge computing for several applications. In addition, energy harvestable or wireless-powered wearable devices are envisioned to empower the edge intelligence in IoT applications. However, the intermittent energy supply and network connectivity of such devices in scenarios including remote areas and hard-to-reach regions such as in-body applications can limit the performance of edge computing-based IoT applications. Hence, deploying stateof-the-art convolutional neural networks (CNNs) on such energyconstrained devices is not feasible due to their computational cost. Existing model compression methods such as network pruning and quantization can reduce complexity, but these methods only work for fixed computational or energy requirements, which is not the case for edge devices with an intermittent energy source. In this work, we propose a pruning scheme based on deep reinforcement learning (DRL), which can compress the CNN model adaptively according to the energy dictated by the energy management policy and accuracy requirements for IoT applications. The proposed energy policy uses predictions of energy to be harvested and dictates the amount of energy that can be used by the edge device for deep learning inference. We compare the performance of our proposed approach with existing state-of-the-art CNNs and datasets using different filter-ranking criteria and pruning ratios. We observe that by using DRL-driven pruning, the convolutional layers that consume relatively higher energy are pruned more as compared to their counterparts. Thereby, our approach outperforms existing approaches by reducing energy consumption and maintaining accuracy.

Design of a spin–orbit-torque-based self-terminated nonvolatile register for a write/leakage-energy-minimized backup operation

“…Figure 10 shows a waveform of the proposed selfterminated power-gating technique where for NV-FFs are used. Just after CMP[0], CMP [1], CMP [2], and CMP [3] rise up owing to write completion, SLP becomes high level and power supply of the NV-FFs are cut off. We can also see that the levels of CMP signals gradually fall down because power supply is cut off.…”

“…image recognition using deep learning) with limited power supply from small battery or energy harvesting. 1,2) Leakage power consumption in a conventional CMOS-only volatile logic LSI is critical for such power-limited IoT devices since the power supply must be continuously applied to retain internal data. 3,4) Nonvolatile (NV) logic, where internal data are stored in NV-devices is a fundamental solution for this issue.…”

An intelligent Hybrid‐Q Learning clustering approach and resource management within heterogeneous cluster networks based on reinforcement learning

“…An RL-based power manager used a present Q-learning technique to educate node-selected duty cycles dynamically. 47 Their reward function is the distance from the energy neutrality condition to the energy storage level and the throughput requirements.…”

Review of Predator Emitted Volatile Organic Compounds and Their Potential for Predator Detection in New Zealand Forests

“…However, the subsequent identification of captured species usually requires manual assessment by experienced professionals [19,20]. Alternatively, employing an advanced AI-driven framework for image recognition could reduce labour requirements but it might come at the cost of increased power consumption [21][22][23]. There are also techniques developed for tracking and detecting predators, such as faecal sample collections followed by deoxyribonucleic acid (DNA) or messenger ribonucleic acid (mRNA) extractions and identifications [24].…”