Artificial neural networks (ANNs) have become essential components in various safety-critical applications, including autonomous vehicles, medical devices, and avionics, where system failures can lead to severe risks. Edge AI devices, which process data locally without relying on the cloud, are increasingly used to meet the performance and real-time demands of these applications. However, their reliability in radiation-prone environments is a significant concern. In this context, this paper evaluates the MAX78000, an ultra-low-power Edge AI microcontroller with a hardware-based convolutional neural network (CNN) accelerator, focusing on its behavior in radiation environments. To assess the reliability of the MAX78000, we performed a test campaign at the ChipIR neutron irradiation facility using two different ANNs. We implemented techniques to improve system observability during ANN inference and analyzed the radiation-induced errors observed. The results present a comparative analysis between the two ANN architectures, which shows that the complexity of the ANN directly impacts its reliability.

Tiny machine learning (TinyML) enables low-power microcontrollers to run neural network inference locally, allowing real-time classification near the sensor with reduced latency and no reliance on external infrastructure. This is especially relevant in constrained environments such as autonomous systems, satellites, and safety-critical embedded applications, where energy efficiency and reliability are essential. This work evaluates the inference performance of a quantized convolutional neural network on HARV-SoC, a fault-tolerant RISC-V system-on-chip with vector extension support. The SoC includes our port of TensorFlow Lite for Microcontrollers with accelerated vectorized convolution functions. Inference was tested using RGB and grayscale inputs to analyze how input structure impacts vectorization. Results show that, compared to the original convolution function, the RGB and grayscale models achieved 1.80 and 2.05 times speedup, respectively. The highest gains were observed when the input structure enabled better use of the available parallelism.