AI Insight
Researchers developed a framework that dynamically partitions neural network computations across IoT devices, edge servers, and cloud infrastructure in real-time, adapting to changing network conditions. Testing on actual hardware using a Raspberry Pi, laptop, and desktop PC with three common neural networks (VGG16, AlexNet, MobileNetV2) showed energy reductions of 27-36% and latency reductions of 6-23% compared to static partitioning methods. The system profiles models at startup and periodically reassesses the optimal distribution of computational tasks across the edge-cloud continuum.
Why it matters
This approach addresses a critical limitation in deploying AI on resource-constrained IoT devices by making real-time decisions about where to process data. The demonstrated energy and speed improvements could enable more efficient AI applications in smart homes, autonomous vehicles, and industrial IoT systems where both computational resources and network conditions vary.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
-cross
Abstract: In recent years, the use of artificial intelligence on resource-constrained IoT devices has grown significantly. However, existing approaches to AI task partitioning and offloading across the edge-cloud continuum typically rely on static methods that ignore runtime dynamics. Furthermore, they are often evaluated in simulated environments rather than on real hardware. To address this gap, we propose a framework that dynamically splits neural network layers across the heterogeneous continuum. The framework profiles the model at startup, measures network link conditions between nodes, and periodically re-evaluates the partition to adapt to environmental changes. We created a physical testbed comprising a Raspberry Pi edge device, a laptop fog, and a high-performance desktop PC as the cloud. We evaluated the framework over three widely adopted convolutional neural networks: VGG16, AlexNet, and MobileNetV2. Our results show that the framework achieves reductions in energy and end-to-end latency of 27.09–35.82% and 6.34–22.92%, respectively, compared to a static partitioning baseline. These findings confirm the superiority of adaptive to static partitioning.
Source: Adaptive AI Task Partitioning and Safe Offloading in Heterogeneous Edge-Cloud Continuum