Low Power McUs

Low-power microcontrollers (MCUs) are enabling the deployment of machine learning (ML) at the edge, focusing on minimizing energy consumption and memory footprint while maintaining acceptable performance. Research emphasizes efficient neural network architectures, such as optimized Tiny Transformers and lightweight Convolutional Neural Networks (CNNs), along with techniques like model compression, quantization, and incremental online learning to adapt to changing environments. This field is crucial for developing resource-constrained applications like nano-UAV navigation and smart sensors, driving innovation in both hardware (e.g., heterogeneous RISC-V SoCs) and software (optimized ML libraries).