A compact neuromorphic system for ultra energy-efficient, on-device robot localization

This repository contains the data for the following publication, if you use it in your work please cite appropriately;

@article{HinesLENS2025,
  author = {Adam D. Hines  and Michael Milford  and Tobias Fischer },
  title = {A compact neuromorphic system for ultra–energy-efficient, on-device robot localization},
  journal = {Science Robotics},
  volume = {10},
  number = {103},
  pages = {eads3968},
  year = {2025},
  doi = {10.1126/scirobotics.ads3968},
  URL = {https://www.science.org/doi/abs/10.1126/scirobotics.ads3968}
}

This dataset contains 3 .zip files corresponding to the data for Figure 3, 4, and 5 of our publication (named accordingly) and the code for Locational Encoding with Neuromorphic Systems (LENS, @ v0.1.3). It contains the following;

• Figure 3:
  • SoCWatch output of power traces for CPU (.csv), jetson power traces for LENS and SAD (.npy files), and the samna output of power traces (.npy)
• Figure 4:
  • Consists of 2 folders, sunset1 and sunset2, which contain 7x7 downsampled images from the Brisbane Event-VPR dataset sampled over one second timebins (https://zenodo.org/records/4302805) 
  • Two .csv files which correspond to the image names of sunset1 and sunset2, used in our model to load files
• Figure 5:
  • 2 folders, 220724-16-14-33 (indoor Hexapod data) and 240724-11-49-52 (outdoor Hexapod data)
    • Each folder consists of the LENS.log file, various plots of Recall@N and Precision-Recall, png images of the both the reference and query traversal reconstructed from the raw spikes, the raw spike data (spike_data.npy), and the similarity matrix used to calculate statistics (similarity_matrix.npy)
• LENS.zip
  • v0.1.3 of the LENS software, latest versions available at https://github.com/AdamDHines/LENS