{
  "abstract": "Mobile robots that navigate among moving obstacles must act on 2D LiDAR\n\nscans that are sparse, noisy, and change from one step to the next.\n\nReinforcement learning policies such as PPO are usually trained on these\n\nraw scans directly. A single scan, however, does not show how obstacles\n\nare moving, nor how much of the change between consecutive scans comes\n\nfrom the robot's own motion. The policy has to learn both perception and\n\ncontrol from a sparse reward, and in our baseline experiments this made\n\nperformance sensitive to sensor noise and highly variable across random\n\nseeds. Supervising a better representation would require labels for\n\nobstacle motion, which are costly to collect outside simulation. We\n\ntherefore ask whether a self-supervised model, trained only on unlabelled\n\nLiDAR sequences and the robot's actions, can learn a representation of\n\nobstacle dynamics that improves a downstream PPO policy's success rate\n\nunder sensor noise and reduces its variability across seeds.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": null,
  "author_total": 1,
  "authors": [
    {
      "name": "Tarun Narashiman"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/dppn879x2capg.bib",
  "cited_by": 0,
  "doi": "10.5281/zenodo.23540677",
  "fulltext_url": null,
  "id": "dppn879x2capg",
  "issns": [],
  "landing_url": "https://zenodo.org/records/23540677",
  "language": "eng",
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [],
  "openalex": null,
  "pdf_url": "https://zenodo.org/api/records/23540677/files/Mobile_Robotics_Li_JEPA%20(4).pdf/content",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-21",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/dppn879x2capg.ris",
  "sources": [
    {
      "first_seen": "2026-11-04T12:20:59Z",
      "native_id": "23540677",
      "source": "zenodo",
      "url": "https://zenodo.org/records/23540677"
    }
  ],
  "subjects": [],
  "title": "Action aided Joint Embedding Prediction on 2D LiDAR Sequences for Robust Mobile Robot Navigation in Dynamic Environments",
  "title_text": "Action aided Joint Embedding Prediction on 2D LiDAR Sequences for Robust Mobile Robot Navigation in Dynamic Environments",
  "type": "preprint",
  "updated_at": "2026-11-04T12:20:59Z",
  "url": "https://research.draconitic.com/paper/dppn879x2capg",
  "venue": "Zenodo",
  "versions": []
}