Optimizing Energy-Efficient Task Offloading in Edge Computing: A Hybrid AI-Based Approach

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2025, Volume and Issue: 11(2)

Published: March 23, 2025

Edge computing has emerged as a pivotal technology for managing computational workloads in latency-sensitive applications by offloading tasks from resource-constrained Internet of Things (IoT) devices to nearby edge servers. However, optimizing task while ensuring energy efficiency remains significant challenge. This paper proposes Hybrid AI-Based Task Offloading (HATO) model, integrating Reinforcement Learning (RL) with Deep Neural Networks (DNNs) dynamically allocate resources minimizing consumption. The HATO framework formulates multi-objective optimization problem, considering factors such device workload, network latency, server availability, and constraints. Experimental evaluations demonstrate that the proposed model achieves 27.3% reduction consumption, 19.6% improvement completion time, 31.2% enhancement overall utilization compared conventional heuristic-based methods. reinforcement learning module adapts strategies real-time, optimal load balancing latency. Approach outperforms baseline models diverse scenarios, making it scalable efficient solution next-generation IoT applications.

Language: Английский

Enhancing Cross Language for English-Telugu pairs through the Modified Transformer Model based Neural Machine Translation

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2025, Volume and Issue: 11(2)

Published: April 16, 2025

Cross-Language Translation (CLT) refers to conventional automated systems that generate translations between natural languages without human involvement. As the most of resources are mostly available in English, multi-lingual translation is badly required for penetration essence education deep roots society. Neural machine (NMT) one such intelligent technique which usually deployed an efficient process from source language another language. But these NMT techniques substantially requires large corpus data achieve improved process. This bottleneck makes apply mid-resource compared its dominant English counterparts. Although some benefit established systems, creating low-resource a challenge due their intricate morphology and lack non-parallel data. To overcome this aforementioned problem, research article proposes modified transformer architecture improve efficiency NMT. The proposed framework, consist Encoder-Decoder enhanced version with multiple fast feed forward networks multi-headed soft attention networks. designed extracts word patterns parallel during training, forming English–Telugu vocabulary via Kaggle, effectiveness evaluated using measures like Bilingual Evaluation Understudy (BLEU), character-level F-score (chrF) Word Error Rate (WER). prove excellence model, extensive comparison existing architectures performance metrics analysed. Outcomes depict has shown improvised by achieving BLEU as 0.89 low WER when models. These experimental results promise strong hold further experimentation based

Language: Английский