This paper presents a fuzzy logic-based approach for replica scaling in a Kubernetes environment, focusing on integrating Edge Computing. The proposed FHS (Fuzzy-based Horizontal Scaling) system was compared to the standard Kubernetes scaling mechanism, HPA (Horizontal Pod Autoscaler). The comparison considered resource consumption, the number of replicas used, and adherence to latency Service-Level Agreements (SLAs). The experiments were conducted in an environment simulating Edge Computing infrastructure, with virtual machines used to represent edge nodes and traffic generated via JMeter. The results demonstrate that FHS achieves a reduction in CPU consumption, uses fewer replicas under the same stress conditions, and exhibits more distributed SLA latency violation rates compared to HPA. These results indicate that FHS offers a more efficient and customizable solution for replica scaling in Kubernetes within Edge Computing environments, contributing to both operational efficiency and service quality.

Autonomous navigation in mobile robots represents a significant challenge in unknown and dynamic environments, where the need to avoid obstacles and find safe trajectories in real-time requires efficient solutions. This work presents the MetaHeuristic Real-Time Safe Navigation (MHRTSN) strategy, which combines potential fields with population-based meta-heuristics to optimize trajectory planning and navigation in such environments. The MHRTSN strategy was tested through a series of simulations using different static and dynamic scenarios, comparing the performance of MHRTSN-GA and MHRTSN-PSO versions in terms of displacement, distance traveled, and CPU and clock times. The results demonstrate that both versions provide sub-optimal solutions, with MHRTSN-PSO showing superior performance in terms of required processing time and better convergence compared to MHRTSN-GA when using a small number of individuals. Comparisons with other approaches in the literature showed that MHRTSN generated paths of equivalent length to those of the literature, but in a safer manner. Thus, the proposed approach offers an efficient and safe solution for autonomous navigation in mobile robots, contributing to the advancement of this field in real-world applications. Future work may explore the application of the strategy in physical robotic systems and more complex environments.

This work presents the development of an embedded platform using Field Programmable Gate Arrays (FPGAs) for real-time simulation of dynamic systems in industrial plants. The platform, Real-Time Simulator for Dynamic Systems in FPGA (RTSDS-FPGA), is designed for industrial and academic applications. In industrial contexts, the RTSDS-FPGA facilitates the optimization and tuning of embedded control algorithms, while in academia, it supports research on new embedded solutions in automation and control. It is also an educational tool for undergraduate and postgraduate students developing embedded control system projects. Additionally, the simulator accelerates the simulation of slow dynamic systems, significantly reducing overall simulation time. Experimental results demonstrate the platform’s capability to perform real-time simulations effectively, validating its accuracy and performance through comparative analyses with established software tools such as Matlab/Simulink.

Deep learning techniques, such as deep neural networks (DNNs), have proven highly effective in addressing various automatic modulation classification challenges. However, their computational demands pose a significant hurdle for real-time modulation detection. To tackle this issue, a novel training strategy is proposed in this study. This strategy aims to minimize both pruning and quantization losses during the training of compressed models, thereby reducing the computational complexity of DNNs. The effectiveness of this approach was demonstrated through experiments involving the classification of 24 modulations: OOK, ASK4, ASK8, BPSK, QPSK, PSK8, PSK16, PSK32, APSK16, APSK32, APSK64, APSK128, QAM16, QAM32, QAM64, QAM128, QAM256, AM SSB WC, AM SSB SC, AM DSB WC, AM DSB SC, FM, GMSK and OQPS. Remarkably, the results showed a substantial reduction in both DNN weights and operations, while maintaining a high level of classification accuracy. By streamlining the computational demands of deep learning models, this strategy opens up new possibilities for real-time modulation detection applications, particularly in scenarios where computational resources are limited. This research represents a significant advancement towards the practical deployment of deep learning in modulation classification systems, paving the way for enhanced efficiency and performance in wireless communication technologies.

Modelos de deep learning, como redes neurais convolucionais (CNNs), têm se destacado em diversas tarefas de aprendizado supervisionado, mas frequentemente exigem elevada capacidade computacional para execução eficiente. No entanto, sua elevada complexidade computacional pode dificultar a implantação eficiente em ambientes distribuídos baseados em microsserviços, como clusters Kubernetes, especialmente quando há restrições de recursos. Diante desse cenário, foi conduzido um estudo experimental utilizando uma abordagem de compressão consciente de modelos de redes neurais profundas, baseada na aplicação iterativa de poda seguida de quantização (P → Q), com o objetivo de reduzir o tamanho dos modelos, o consumo de memória e CPU, e o tempo de resposta durante a inferência. Para validar a proposta, foram utilizados modelos VGG-16 treinados com o conjunto de dados CIFAR-10, submetidos a diferentes configurações de compressão controladas por dois parâmetros: α, que define o grau de agressividade da poda, e b, que determina a quantidade de bits utilizados na quantização dos pesos. Os modelos comprimidos foram implantados como microsserviços em um cluster Kubernetes utilizando MicroK8s, e testados com geradores de carga via JMeter, sendo monitorados por Prometheus. Os resultados indicam que a técnica P → Q permite uma compressão de até 5,6 vezes no tamanho do modelo, mantendo níveis satisfatórios de acurácia, com perdas inferiores a 7%. Além disso, observou-se uma redução de até 1,8 vezes na latência de resposta, 1,71 vezes no uso de memória e 1,46 vezes no consumo de CPU. Esses achados demonstram o potencial e a viabilidade da abordagem proposta para a otimização de aplicações baseadas em DNNs em ambientes de microsserviços escaláveis.