Multimodal Agent MemorySubmit a paper

RoboMemory

RoboMemory: A Brain-inspired Multi-memory Agentic Framework for Interactive Environmental Learning in Physical Embodied Systems

Mingcong Lei, Honghao Cai, Yuyuan Yang, Yimou Wu, Jinke Ren, Zezhou Cui, et al. · 2026-03-24

Read the paperPDF

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. spatial knowledge graph

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store
    3. Retrieved byStructure traversal, Planner / tool-assisted retrieval
    4. Source evidenceSource linked & recoverable
    A dynamic spatial knowledge graph supports consistent environmental updates and planning.Full paper: B. Spatial Memory · PDF p. 3 · Figure 1

  2. temporal, episodic, and semantic memories

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inHierarchical multi-tier store
    3. Retrieved bySimilarity search, Metadata / spatiotemporal retrieval
    4. Source evidenceSource linked & recoverable
    Parallel memory modules retain temporal, episodic, and semantic experience for repeated tasks.Full paper: Semantic Memory. In cognitive psychology, semantic · PDF p. 5 · Figure 1

How memory changes (T2 operations)

  • Update · Prompted LLM/VLM record update

    During operation, RoboMemory applies LLM-mediated memory editing to spatial knowledge graph; the memory state changes while the governing update policy remains fixed.

  • Consolidation · Episodic / hierarchical summarization

    During operation, RoboMemory applies episodic or hierarchical summarization to spatial knowledge graph; the memory state changes while the governing update policy remains fixed.

  • Revision · Rule-based state correction

    During operation, RoboMemory applies fixed-rule correction of stored state to spatial knowledge graph; the memory state changes while the governing update policy remains fixed.

  • Reorganization · Relinking & reindexing

    Environmental changes are integrated into the dynamic knowledge graph. For RoboMemory, this specifically supports relinking or reindexing of stored items of spatial knowledge graph.

Related systems