Multimodal Agent MemorySubmit a paper

CAUSALNAV

CausalNav: A Long-term Embodied Navigation System for Autonomous Mobile Robots in Dynamic Outdoor Scenarios

Hongbo Duan, Shangyi Luo, Zhiyuan Deng, Yanbo Chen, Yuanhao Chiang, Yi Liu, et al. · 2026-01-05

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. multi-level semantic scene graph

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store, Vector-indexed collection
    3. Retrieved bySimilarity search, Structure traversal, Planner / tool-assisted retrieval
    4. Source evidenceSource not retained
    Embodied Graph constructed from map data and live perception, then updated in a temporal window. RAG retrieves from the Embodied Graph for semantic navigation and long-range planning under open-vocabulary queries.Full paper: A. System Overview · PDF p. 2 · Figure 1 / method overview

How memory changes (T2 operations)

  • Update · Rule-based update

    During operation, CAUSALNAV applies fixed-rule insertion or replacement to multi-level semantic scene graph; the memory state changes while the governing update policy remains fixed.

  • Revision · Rule-based state correction

    During operation, CAUSALNAV applies fixed-rule correction of stored state to multi-level semantic scene graph; the memory state changes while the governing update policy remains fixed.

  • Forgetting · Discrete eviction

    Graph is continuously updated within a temporal window to reflect environmental changes and dynamic objects. For CAUSALNAV, this specifically supports explicit deletion or eviction of stored items of multi-level semantic scene graph.

Related systems