The article provides a roundup of significant developments in edge AI, quantum chips, and embedded computing, highlighting the release of the ESD November 2025 eBook. It encourages readers to explore additional resources on the Embedded website for more insights.
Ambient computing enhances the Internet of Things (IoT) by enabling real-time responses to monitored parameters through a smart network. This integration allows for efficient tracking of data such as movement and temperature with minimal human intervention.
The article discusses the challenges of deploying large-scale Deep Learning models on edge devices due to computational and energy constraints. It introduces a hardware-aware Neural Architecture Search framework aimed at optimizing Early Exiting Neural Networks for edge accelerators, considering factors like quantization.
The paper introduces a method for enhancing the functional stability of neural networks by employing hardware redundancy at the neuron level. This approach allows the system to maintain operation despite individual hardware failures.
Robert shares his hands-on experience with the Zephyr RTOS and highlights his contribution of a sensor driver for the MTCH9010 liquid leak detector. The article provides insights into the practical applications of the RTOS in sensor technology.
ASRock Industrial showcased its advancements in secure device integration and FIDO Device Onboard (FDO) technology at the FIDO Taipei Seminar. This marks a significant achievement as ASRock becomes the first industrial computer company in Taiwan to receive FDO certification.
Researchers from Tokyo University of Science and National Institute of Advanced Industrial Science and Technology have introduced approximate domain unlearning, a method that enhances vision-language models (VLMs) by removing irrelevant information while maintaining performance. This approach addresses the limitations of current class unlearning methods, which are inadequate for practical applications in domain-specific recognition.
The article discusses how Edge AI technologies can enhance performance and power efficiency while keeping data local. However, it also highlights that these advancements introduce new security risks that need to be addressed.
The article introduces KVNAND, a novel architecture that enables large language model inference on edge devices without the need for DRAM by utilizing compute-enabled 3D NAND flash for both model weights and key-value cache. This innovation addresses the challenges of low arithmetic intensity and high bandwidth demands associated with traditional methods.
The article highlights key collaborations that are advancing technologies in AI processing, IoT connectivity, and energy harvesting. These innovations are pivotal for enhancing edge intelligence and supporting next-generation communication standards.
Efinix is a Silicon Valley startup that is transforming programmable hardware and edge AI with its innovative FPGA technology. The article highlights the company's journey and its potential impact on the industry.
The article discusses the goals of optimizing AI workloads at the edge, focusing on performance enhancement, cost reduction, data security, and energy efficiency. These objectives are critical for the advancement of edge computing technologies.
The article discusses a new near-memory architecture designed to enhance the performance of corner detection algorithms in event-based cameras, which are crucial for applications like surveillance and autonomous driving. This architecture aims to reduce latency and power consumption, addressing challenges faced by resource-constrained edge devices.
The article introduces SAT-MapIt, a SAT-based mapper designed for Coarse-Grain Reconfigurable Architectures (CGRAs) that enhances the mapping efficiency of compute-intensive application loops. It proposes a novel kernel mobility schedule (KMS) to optimize the mapping process compared to traditional modulo scheduling techniques.
The article discusses recent advancements in enhancing intelligence at the edge of computing. It highlights the growing importance of edge AI in various applications.
SECO has launched new Edge AI applications that improve safety monitoring, low-power machine vision, and geospatial analysis. This expansion enhances SECO's ecosystem of production-ready AI solutions for industrial and embedded environments.
The article introduces a groundbreaking technology that enables battery- and ambient-powered AI sensors to operate using Wi-Fi or Bluetooth while consuming significantly less power. This innovation could transform the landscape of edge AI applications by drastically reducing energy requirements.
The article discusses the limitations of current DNN compression methods for homogeneous edge devices, which often fail to account for performance variations among devices over time. It introduces a new framework called Homogeneous-Device Aware Pruning (HDAP) aimed at optimizing DNN performance across these devices.
The article introduces the XgenSilicon ML Compiler, which automates the transformation of machine learning models into optimized RISC-V assembly code for ASIC accelerators. It highlights five innovations that enhance Power, Performance, and Area (PPA) metrics over conventional methods.
The article introduces crossbar-aware dendritic convolution (CADC), a new method that enhances the efficiency of in-memory computing for convolutional neural networks by increasing sparsity in partial sums. This approach significantly reduces the computational overhead associated with large convolutional layers in various neural network architectures.