This paper explores branch predictor design for the CVA6 RISC-V core using gem5 simulation and machine learning-driven parameter optimization with comprehensive validation and cross-architecture analysis. Seven predictors are systematically evaluated using a multi-dimensional framework capturing performance, power, and embedded-specific metrics. Optuna-based tuning with Tree-structured Parzen Estimator explores 2,847 configurations to optimize energy-delay and area-performance trade-offs. Results show ML-optimized tournament predictors achieve 23.7% better performance than static predictors and 8.4% improvement over default configurations, while consuming 31% more power. The optimized bi-modal predictor emerges as optimal for power-constrained IoT applications, offering 94% of tournament predictor performance at 42% lower power consumption. Cross-architecture comparison with ARM Cortex-A53 reveals 9.3% better prediction accuracy for RISC-V with equivalent complexity. The framework provides validated workload-specific recommendations for embedded RISC-V systems with statistical confidence intervals and failure case analysis.

The legged robots have a better performance than the wheeled and tracked mobile robots for some applications. The development of quadruped robots (QRs) has increased in the last years mainly with the mammalian configuration due to the velocity of displacement and for the implementation of new control strategies that improve its stability. However, it is necessary to integrate embedded systems into the QR design to increase the overall performance. The proposed research presents the development of an integrated system composed of a QR in mammalian configuration and an embedded system with a concurrent methodology that ensures harmonic integration. The system was tested under real environments, on flat, sandy, and rocky terrains. The experimental results were analyzed with Kinovea software, obtaining a maximum velocity in the flat terrain with 0.43 and 0.2 m/s in sandy terrain, and 0.345 m/s in rocky. Finally, an embedded system is a factor key to improving the QR’s performance because it allows for modification of the behavior through the implementation of unconventional control strategies, to make the robot smaller, and it is possible to integrate tools of IoT that are necessary for the new applications.

Latin American agriculture encounters significant challenges such as inefficient land use and vulnerability to climate change, with smallholder farmers being particularly affected. This paper proposes a modular Internet of Things (IoT)-based system that integrates embedded systems, environmental sensors, and Artificial Intelligence (AI) for real-time monitoring and adaptive control. The system utilizes a Raspberry Pi coupled with advanced sensors measuring temperature, humidity, potential of hydrogen (pH), liquid level, light intensity, air quality, and electrical current. Actuators regulate nutrient delivery in response to pH levels, while intelligent lighting systems enhance plant growth conditions. Remote monitoring facilitated by a vision-equipped camera system allows the detection of anomalies and the tracking of plant growth. Experimental results demonstrated temperature stability of ±1∘C, humidity variation within ±5%, 92% accuracy in anomaly detection, and response times of less than 10 seconds.

This letter presents a novel architecture for a high-throughput encoder for quasi-cyclic low-density parity-check codes. This low-complexity encoder is specifically tailored for the 5th generation (5G) new radio (NR) standard. To achieve high throughput, we employ an automated approach to design customized encoders for individual base graphs using a Verilog Code Generator (VeCoGen) that we developed in MATLAB. Our proposed architecture improves efficiency by rearranging the wire indices during the hardware design language code generation stage itself, instead of performing shifting operations on the data carried by those wires on the fly. This eliminates the need for RAM (for storing base graph coefficients) and dedicated barrel shifters. The parity bits are computed in parallel, while maintaining optimal performance and gate count through effective exploitation of the base graph’s sparsity. Consequently, the encoder can be designed as a fully combinational circuit between the input and output registers, enabling the generation of the complete codeword within a single clock cycle. This results in exceptionally high throughput, optimal hardware utilization, and minimal number of <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">xor</small> operations, making our approach highly effective for 5G NR applications.

FPGAs provide customizable, low-power, and real-time ML Models acceleration for embedded systems, making them ideal for edge applications like robotics and IoT. However, ML models are computationally intensive and rely heavily on multiplication operations, which dominate the overall resource and power consumption, especially in deep neural networks. Currently available open-source frameworks, such as hls4ml, FINN and Tensil artificial intelligence (AI), facilitate FPGA-based implementation of ML algorithms but exclusively use accurate arithmetic operators, failing to exploit the inherent error resilience of ML models. Meanwhile, a large body of research in approximate computing has produced Approximate Multipliers that offer substantial reductions in area, power, and latency by sacrificing a small amount of accuracy. However, these Approximate Multipliers are not integrated into widely used hardware generation workflows, and no automated mechanism exists for incorporating them into ML model implementations at both software and hardware levels. In this work, we extend the hls4ml framework to support the use of Approximate Multipliers. Our approach enables seamless evaluation of multiple approximate designs, allowing tradeoffs between resource usage and inference accuracy to be explored efficiently. Experimental results demonstrate up to 3.94% LUTs savings and 7.33% reduction in On-Chip Power, with accuracy degradation of 1% compared to accurate designs.

software-defined vehicles (SDVs) employ zonal architectures whose zones exchange periodic real-time traffic over a high-speed Time-Sensitive Ethernet (IEEE 802.1Q) backbone. Existing research has produced many deterministic routing schemes for such backbone Ethernet networks, yet has largely ignored the complementary problem of payload-level frame minimization: multiplexing dozens of small zonal messages—whether CAN, LIN, or any other in-zone protocol—into the fewest possible Ethernet frames so that bandwidth, switch buffers and latency budgets are not squandered. This letter concentrates on that neglected frame-minimization facet and presents an optimization flow dedicated to reducing the number of Ethernet frames. We first formulate an exact Satisfiability Modulo Theories (SMTs) model that finds the minimal set of multiplexed frames required across an entire hyper-period. Because SMT becomes intractable on large vehicle topologies, we then introduce a matrix-based heuristic aggregation (MaHA) algorithm that reproduces the SMT’s frame-minimization decisions to within a few percent while executing in milliseconds. Experiments on synthetic SDV workloads show that naive one zonal message per Ethernet frame policies waste significant available payload capacity; our SMT eliminates this waste completely, and the heuristic achieves closely similar savings with up to three orders of magnitude less run-time, making it a practical drop-in solution for next-generation SDV networks.

This study investigates vulnerabilities of future generation nonvolatile memory (NVM)-backed persistent TinyML hardware neural networks to side-channel attacks (SCAs) using electromagnetic (EM) analysis methods. We trained three different tinyML models: MobileNet, ResNet, and EfficientNet on three different standard datasets: F-MNIST, CIFAR-10, and MNIST. The trained networks were then mapped on to a custom FPGA-NVM setup for EM-SCA evaluation. We demonstrate that the information about the stored model parameters/weights can be extracted by applying statistical methods on the collected EM emanation data. Further, we demonstrate that the obtained model parametric information can be used for cloning some of the lightweight edge TinyML models with only 0.5%–10% of total training dataset.