ReMEmbR
ReMEmbR: Building and Reasoning Over Long-Horizon Spatio-Temporal Memory for Robot Navigation
Abrar Anwar, John Welsh, Joydeep Biswas, Soha Pouya, Yan Chang · 2024-09-20
- Management
- T2 Dynamic Memory
- Memory changes during deployment through a stable management policy.
- Representation
- Hybrid
- Distinct operative representation paths are explicitly linked for downstream retrieval, reasoning, or control.
How experience becomes memory
Each path is one memory component, traced from what is stored to how it is found again.
long-horizon robot image history
- Represented asRetained perceptual source evidenceSource- and modality-specific
- Stored inVector-indexed collection
- Retrieved bySimilarity search, Metadata / spatiotemporal retrieval
- Source evidenceSource directly available
Continuously growing robot histories retain images for later spatial and temporal questions.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1
spatiotemporal navigation index
- Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
- Stored inGraph / relational store
- Retrieved byMetadata / spatiotemporal retrieval, Structure traversal
- Source evidenceSource linked & recoverable
Temporal and spatial metadata link retained images into a queryable navigation memory.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1
How memory changes (T2 operations)
Update · Rule-based update
New robot observations and metadata are appended throughout deployment. For ReMEmbR, this specifically supports fixed-rule insertion or replacement of long-horizon robot image history.
Reorganization · Relinking & reindexing
Spatial-temporal links are updated as the robot navigates. For ReMEmbR, this specifically supports relinking or reindexing of stored items of long-horizon robot image history.