{
  "abstract": "Embodied intelligent systems commonly maintain local representations of their environments, including maps, object states, task histories, spatial relationships, and previously acquired experience. Although local memory supports rapid access and autonomous operation, independently maintaining comprehensive environmental knowledge across every agent can introduce redundant storage, repeated observations, duplicated updates, and inconsistent representations. Distributed robotics and cloud-based knowledge management already address parts of these problems, but the provisioning question remains important: which information must each agent maintain locally, and which information can be supplied by a collective memory system without compromising task performance?\n\nThis paper proposes ZAC-Memory, a research framework for collective provisioning of memory and world-model capabilities within the broader Zero-Agent-Compute (ZAC) framework. The central principle is that an embodied agent need not independently maintain every environmental record, knowledge item, or historical representation required by its tasks. Instead, memory requirements can be distributed across local agents, shared repositories, and cooperating providers according to task criticality, access frequency, information freshness, communication conditions, storage capacity, and reliability requirements. The framework introduces dynamic memory provisioning, evidence-aware conflict resolution, temporal validity management, dependency-aware memory placement, and disconnection-aware autonomous operation. It further distinguishes raw observations, episodic records, semantic knowledge, spatial representations, and predictive world models, recognizing that these memory types have different sharing and consistency requirements.\n\nA joint optimization formulation is proposed to minimize storage, communication, synchronization, reconstruction, and failure-related costs while satisfying task-level constraints. A corresponding operational algorithm determines which memory items should remain local, be shared, be replicated selectively, or be reconstructed when necessary. The paper also develops testable hypotheses and a reproducible evaluation methodology using multi-robot environments, dynamic object states, network interruptions, and heterogeneous memory capacities. ZAC-Memory is presented as a conceptual research framework rather than an experimentally validated system. Its contribution must be established through systematic comparison with distributed mapping, cloud robotics, shared knowledge bases, and existing memory-management approaches.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": null,
  "author_total": 1,
  "authors": [
    {
      "name": "Jincheng Zhang",
      "orcid": "0009-0005-1860-0009"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/3kge43aa7x41p.bib",
  "cited_by": 0,
  "doi": "10.5281/zenodo.23450948",
  "fulltext_url": null,
  "id": "3kge43aa7x41p",
  "issns": [],
  "landing_url": "https://zenodo.org/records/23450948",
  "language": null,
  "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/23450948/files/ZAC-Memory%20Collective%20Provisioning%20of%20Memory%20and%20World%20Models%20for%20Embodied%20Intelligence.pdf/content",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-18",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/3kge43aa7x41p.ris",
  "sources": [
    {
      "first_seen": "2026-11-01T12:21:11Z",
      "native_id": "23450948",
      "source": "zenodo",
      "url": "https://zenodo.org/records/23450948"
    }
  ],
  "subjects": [],
  "title": "ZAC-Memory: Collective Provisioning of Memory and World Models for Embodied Intelligence",
  "title_text": "ZAC-Memory: Collective Provisioning of Memory and World Models for Embodied Intelligence",
  "type": "preprint",
  "updated_at": "2026-11-01T12:21:11Z",
  "url": "https://research.draconitic.com/paper/3kge43aa7x41p",
  "venue": "Zenodo",
  "versions": []
}