Luca Sterpone

Heterogeneous FPGA platforms combining RISC-V processors and deep-learning accelerators are increasingly adopted in avionics and space, where performance must be paired with fault tolerance. We present RePAIR, a reconfigurable platform integrating an open-source RISC-V core with a TPU-like systolic-array accelerator for runtime fault detection, correction, and recovery. RePAIR extends the accelerator ISA with runtime self-test, enabling detection of structural faults in the array during inference. The platform supports dual inference modes: a plain mode with no overhead and a testing mode that performs checksum validation at a fixed cost of three extra cycles per matrix multiplication, with limited accelerator area overhead. Upon fault detection, the accelerator notifies the RISC-V processor, which triggers dynamic partial reconfiguration of the faulty region while preserving execution state, allowing inference to resume from the last correct step. Compared with full-device reconfiguration, recovery time is reduced by up to 900× on AMD KCU105 and 1400× on AMD ZCU102, while inference overhead remains ≤30% in the worst case. The methodology is hardware-agnostic and portable across FPGA devices, as shown by multi-platform implementations. Fault-injection campaigns combined with space-environment modeling estimate mean time to failure under mission conditions, demonstrating scalable and reliable FPGA-based AI acceleration for safety-critical applications.

The increasing availability of embedded systems has opened up new opportunities for distributed machine learning, where multiple low-cost boards can collaborate to execute complex algorithms, providing high performance at a lower cost. Traditional centralized algorithms, such as k-Nearest Neighbors (k-NN), often scale poorly when datasets are large or computational resources are limited. We propose SCALES (SCAlable cLassification on Embedded Systems), a framework that enables a distributed platform for k-NN classification using clusters of embedded devices. Our approach enables parallel execution, reducing computational time while preserving accuracy. The framework is designed to be flexible and plug-and-play, supporting various numbers of nodes and data partitioning strategies. We validated SCALES on both real-world and synthetic datasets of various sizes and dimensions. We systematically compared multiple data partitioning strategies, including random split, k-means, and balanced k-means. Experimental results demonstrate that distributed k-NN on embedded clusters significantly reduces execution time while preserving classification accuracy. Among the evaluated strategies, balanced k-means achieves the most effective trade-off between execution time and classification accuracy.