Friday 14 March 2025
The quest for more efficient artificial intelligence has led researchers to explore new frontiers in edge computing. A recent study proposes a novel approach to offloading deep neural network (DNN) inference tasks, leveraging transformer networks and historical data to minimize energy consumption while adhering to latency constraints.
The authors’ algorithm, designed specifically for mobile devices, tackles the challenge of efficiently offloading DNN tasks to edge servers. By modeling the inference tasks as directed acyclic graphs, they formulate a problem that aims to reduce the devices’ energy consumption while meeting their latency requirements and accounting for servers’ capacity.
To solve this complex optimization problem, the researchers employ a transformer DNN architecture, trained on historical data. This enables the model to learn optimal offloading decisions for each task layer, taking into account factors such as computation workload, layer type, and size.
The results are promising: when edge computing resources are limited, the proposed algorithm exhibits an 18% reduction in energy consumption compared to established baseline schemes. Moreover, it significantly decreases task failure rates, making it a more reliable solution for real-world applications.
One of the key advantages of this approach lies in its adaptability across diverse task types and conditions. By training on historical data, the model can generalize well to new scenarios, including image processing tasks such as ResNet18, MobileNetV1, and VGG16.
The study highlights the importance of considering both energy efficiency and latency constraints when designing edge AI systems. As AI applications continue to proliferate in mobile devices and IoT systems, researchers are working to develop more efficient and reliable solutions that balance these competing demands.
In a world where artificial intelligence is becoming increasingly ubiquitous, the need for efficient edge computing has never been more pressing. This research offers a significant step forward in achieving this goal, paving the way for more widespread adoption of AI in mobile devices and IoT systems.
Cite this article: “Efficient Edge Computing for Mobile Devices through Transformer-Based Offloading”, The Science Archive, 2025.
Edge Computing, Artificial Intelligence, Deep Neural Networks, Transformer Networks, Energy Consumption, Latency Constraints, Mobile Devices, Iot Systems, Offloading, Optimization Problem
