Multimodal Agent MemorySubmit a paper

eMEM

eMEM: A Hybrid Spatio-Temporal Memory System For Embodied Agents

A. Haroon Rasheed, Maria Kabtoul · 2026-06-22

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Embodied & 3D
Also: text
Representation
Cross-represented explicit

How experience becomes memory

Each path is one memory component, traced from what is stored to how it is found again.

  1. multi-index spatiotemporal graph

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store, Vector-indexed collection
    3. Retrieved bySimilarity search, Metadata / spatiotemporal retrieval, Structure traversal, Planner / tool-assisted retrieval
    4. Source evidenceSource linked & recoverable
    SQLite, HNSW, and R-tree indexes expose meaning-, time-, and space-aware recall over one graph.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1
  2. consolidated perceptual summaries

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inHierarchical multi-tier store
    3. Retrieved byPlanner / tool-assisted retrieval
    4. Source evidenceSource linked & recoverable
    A tiered pipeline transforms raw observations into compressed summaries for cross-layer recall.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1

How memory changes (T2 operations)

  • Consolidation · Episodic / hierarchical summarization

    Tiered consolidation compresses raw perceptual observations into summaries. For eMEM, this specifically supports episodic or hierarchical summarization of multi-index spatiotemporal graph.

  • Reorganization · Memory-level migration (short- to long-term)

    Observations move through multiple consolidation layers under a fixed pipeline. For eMEM, this specifically supports migration between memory tiers of multi-index spatiotemporal graph.

  • Consolidation · Merging & deduplication

    Consolidation reduces redundant observations while preserving cross-layer links. For eMEM, this specifically supports merging or deduplication of overlapping entries of multi-index spatiotemporal graph.

  • Forgetting · Progressive decay

    During operation, eMEM applies progressive decay of stored importance or retention to multi-index spatiotemporal graph; the memory state changes while the governing update policy remains fixed.

Related systems