Satellite autonomy requires on-board image registration that is both accurate and efficient under strict power and thermal constraints. We introduce a characterization methodology and a benchmark of registration tasks spanning terrain types and difficulty levels, and study the accuracy/energy trade-offs of feature- and intensity-based pipelines to derive guidelines for bringing registration on board.
@article{disalvo2026sigmetrics,title={Are We Ready to Enable Satellite Autonomy Through On-Board Image Registration?},author={Di Salvo, Claudio and Del Sozzo, Emanuele and Sorrentino, Giuseppe and D'Arnese, Eleonora and Panicucci, Paolo and Conficconi, Davide},journal={Proceedings of the ACM on Measurement and Analysis of Computing Systems (SIGMETRICS)},year={2026},}
Extended abstract companion to the full paper, summarizing the characterization methodology and headline findings on the accuracy/energy trade-offs of on-board satellite image registration.
@inproceedings{disalvo2026sigmetricsabstract,title={Are We Ready to Enable Satellite Autonomy Through On-Board Image Registration? (Extended Abstract)},author={Di Salvo, Claudio and Del Sozzo, Emanuele and Sorrentino, Giuseppe and D'Arnese, Eleonora and Panicucci, Paolo and Conficconi, Davide},booktitle={ACM SIGMETRICS, Extended Abstract Track},year={2026},}
We present an adaptive AIE–PL system co-designed for end-to-end pyramidal 3D image registration on AMD/Xilinx Versal platforms. The architecture balances AI Engine and programmable-logic work across pyramid levels, trading throughput for accuracy at runtime.
@inproceedings{sorrentino2026fccm,title={Adaptive {AIE--PL} Systems for Efficient End-to-End Pyramidal {3D} Image Registration},author={Sorrentino, Giuseppe and Galfano, Paolo S. and Di Salvo, Claudio and D'Arnese, Eleonora and Conficconi, Davide},booktitle={IEEE International Symposium on Field-Programmable Custom Computing Machines (FCCM)},year={2026},}
A domain-specific CGRA overlay with QAC-driven spatial reconfiguration for on-board satellite vision. PEs hold per-cell configuration registers and are orchestrated by an ARM hard-core over AXI. A planned ASIC tape-out through open shuttle programs targets radiation- and power-constrained deployment.
@inproceedings{disalvo2026uchiha,title={To Infinity and Beyond: A Reconfigurable Architecture for Space Vision Computing},author={Di Salvo, Claudio and Conficconi, Davide},booktitle={IEEE International Symposium on Field-Programmable Custom Computing Machines (FCCM), PhD Forum},year={2026},}