IndustrialRemembr extends NVIDIA ReMEmbR to robot fleets
Original title: Industrial ReMEmbR: Scaling Spatiotemporal Memory to Robot Fleets
Robots encode video on Jetson devices and generate scene descriptions only on demand, without uploading raw frames.
IndustrialRemembr is a shared spatiotemporal memory system for warehouse robot fleets that extends NVIDIA ReMEmbR and reduces server compute by generating scene descriptions only when queried. Each robot encodes clips with Cosmos-Embed1, Cosmos3-Edge, and the Wan2.2 VAE on an NVIDIA Jetson, and zeros detected people for GDPR-oriented PII abstraction. A Nemotron agent retrieves memories by content, position, and time; cached-token decoding matched the unsplit model on all 450 tested decodes.
Unconfirmed: GDPR compliance is described as a design goal; no deployment audit is stated.
- Cosmos3-Edge is split between a robot vision encoder and an on-premises language decoder.
- Caption generation at write time is eliminated.
- Cached-token decoding matched the unsplit model in all 450 tested decodes.
- Detected people in image patches are zeroed out for PII abstraction.
- Vendors
- NVIDIA
- Products
- JetsonReMEmbRIndustrialRemembr
From the original article
1 min read at arxiv.orgWarehouse robot fleets continuously observe operational conditions, but preserving these observations as reusable semantic memory incurs substantial computation, communication, and storage costs. For example, explaining recurring aisle blockages can require reviewing video from multiple robots across several shifts.
By Jad Yammine · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.