International Journal of Embedded Systems (4 papers in press)

Regular Issues

• Energy-efficient fixed priority scheduling for imprecise mixed-criticality tasks on multi-processor platforms  
  by Yi-Wen Zhang, Chen Ouyang, Rong-Kun Chen 
  Abstract: The multi-processor platform has become the mainstream in embedded systems, and energy consumption has become a challenge for these systems. In this paper, we consider the energy-efficient fixed priority partitioned scheduling problem of imprecise mixed-criticality (IMC) tasks on a multi-processor platform. We propose a novel IMC task partitioning algorithm (IMCSGA) to minimise energy consumption under the criticality rate monotonic scheduling. The experimental results show that IMCSGA can save about 8.48% energy consumption compared with the existing methods.
  Keywords: energy-aware; imprecise mixed-criticality; IMC; fixed priority; partitioned scheduling; genetic algorithm.
  DOI: 10.1504/IJES.2025.10077359
   
  
• CCMT-Net: a convolutional coupled-multiscale transformer network for water body extraction from high-resolution remote sensing images  
  by Yong Zhang, Yishu Peng, Guoyun Zhang, Tianhao Liu 
  Abstract: Accurate water body identification in remote sensing images (RSI) is crucial for water resource management, yet deep learning methods often underperform in detecting small water bodies. To address this issue, a convolutional coupled-multiscale transformer network (CCMT-Net) is proposed for enhanced water body extraction from RSI. In the encoder, a convolutional embedding block is employed to effectively preserve edge details of water bodies. Subsequently, a convolutional coupled transformer module is introduced to synergistically extract both local and global features. Furthermore, a feature transformation module is utilised to mitigate the semantic gap between the encoder and decoder. In the decoder, a multi-scale efficient transformer attention model is constructed to realise collaborative optimisation of water body features by integrating cross-layer features. Experiments on the GID and LoveDA datasets show that CCMT-Net outperforms competing methods, achieving IoU values of 80.86% and 69.72% - 1.8% and 1.68% higher than those of UCTransNet, respectively.
  Keywords: transformer; water body; remote sensing images; multi-scale.
  DOI: 10.1504/IJES.2025.10079906
   
  
• Adaptive question difficulty and news classification with scaled soft voting: beyond machine and deep learning models  
  by Aradhana Saxena, A. Santhanavijayan 
  Abstract: Accurate classification of question difficulty is vital for adaptive learning but often suffers from instability and high computational cost in existing models. This study evaluates 15 machine learning classifiers and three deep learning architectures using a dataset of 9,692 questions categorised as simple, average, and tough. To enhance performance, LinearSVC, SGD classifier, and RBF SVM are integrated into a scaled soft voting ensemble, where classifier contributions are weighted based on class-level accuracy. The proposed ensemble achieves a micro-average AUC of 0.95. Cross-domain validation on the AG News benchmark further demonstrates robustness, achieving an overall accuracy of 89%. The ensemble consistently outperforms individual models and alternative strategies while maintaining competitive performance with deep learning approaches at significantly lower training and inference cost. These results highlight that a lightweight weighted ensemble provides an efficient, interpretable, and scalable solution for adaptive testing and text classification tasks.
  Keywords: question difficulty classification; ensemble learning; soft voting; support vector machines; natural language processing; NLP; adaptive learning.
  DOI: 10.1504/IJES.2026.10078383
   
  
• A node-level distributed congestion control algorithm for wireless sensor networks  
  by Lei Niu, Xianchao Wang, Feng Wang, Bo Guo, Dongdong Liu 
  Abstract: This paper proposes a node-level distributed congestion control algorithm (NL-DCC). In this algorithm, a node detects congestion by monitoring the queue length and packet loss. When congestion occurs, the node uses the SNR in listening mode to compute the ratio of data packets sent by itself to the total packets sent by the entire network in real-time. To manage congestion, the algorithm differentiates the rate reduction based on node priority. A minimum threshold for the above ratio is set for low-priority nodes to prevent starvation while ensuring the QoS of high-priority nodes. Simulation results show that the algorithm achieves better performance in terms of delay, PDR, and differentiated services. It meets the requirements for reliable transmission and differentiated services of critical data in industrial scenarios, and demonstrates strong engineering and industrial application value.
  Keywords: wireless sensor networks; congestion control; distributed algorithms; differentiated services; cross-layer.
  DOI: 10.1504/IJES.2026.10079147