Multimodal Agent MemorySubmit a paper

Robo-Cortex

Robo-Cortex: A Self-Evolving Embodied Agent via Dual-Grain Cognitive Memory and Autonomous Knowledge Induction

Nga Teng Chan, Yi Zhang, Yechi Liu, Renwen Cui, Fanhu Zeng, Zeyuan Ding, et al. · 2026-05-18

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. strategy / trajectory memory

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inFlat sequential records
    3. Retrieved byPlanner / tool-assisted retrieval
    4. Source evidenceSource not retained
    State transitions are written; AKI distills experience into heuristics. planning loop uses reflective/principle memories and induced heuristics.Full paper: System Overview · PDF p. 4 · Figure 1 / method overview

How memory changes (T2 operations)

  • Consolidation · Experience abstraction

    AKI updates heuristics online across episodes. For Robo-Cortex, this specifically supports abstraction of episodes into reusable experience of strategy / trajectory memory.

  • Update · Rule-based update

    During operation, Robo-Cortex applies fixed-rule insertion or replacement to strategy / trajectory memory; the memory state changes while the governing update policy remains fixed.

Related systems