Heterogeneous Implementation of Adaptive Radar Signal Sorting on Custom RFSoC Hardware

The algorithm is then restructured under fidelity-preserving constraints, achieving timing closure within the resource budget of the platform while accelerating batch processing. The platform is engineered in-house, its custom board-level architecture, power and clock topologies, and signal-integrity design being better suited to adaptive RSS deployment than those of commercial evaluation boards. Board-level experiments confirm that the fixed-point implementation preserves the floating-point sorting quality, achieves a near order-of-magnitude speedup (about $9\times$ on average) over an x86 baseline, and maintains sorting accuracy above 95\% under composite stress and sustained continuous operation.

These results demonstrate that the implementation retains the adaptiveness and robustness of the algorithm under demanding operating conditions, advancing RSS deployment on edge hardware and enhancing the practical utility of ESM systems.