Blog
Technical posts about persistent memory for robots — experience transfer, spatial search, multi-modal perception, and benchmarks.
Your Robot Learned to Push — Can It Slide?
First Evidence of Cross-Environment Experience Transfer in Robots
Train memory in FetchPush, transfer to FetchSlide. Baseline 4% → with memory 12%. No model fine-tuning — just memory.
transferbenchmark
Find Robot Memories by Location
Spatial Search for Robot Experience Retrieval
"Show me only successful grasps near position [1.3, 0.7, 0.42]." Structured context filtering + Euclidean distance sorting.
spatialsearch
5 Senses, One Database
Unifying Multi-Modal Robot Perception in a Single Memory
Visual, tactile, auditory, proprioceptive, procedural — five perception types, one API, one table.
multi-modalperception
How to Give Your Robot Persistent Memory
Robots run 1,000 episodes but start each one from scratch. Learn how to add persistent memory — +25% success rate in MuJoCo FetchPush.
getting-startedbenchmark