When the Past Shapes Action: Understanding Embodied Memory
cs.RO, cs.AI, cs.LG
Submitted: 2026-10-30
Updated: 2026-10-30
License: http://creativecommons.org/licenses/by-nc-sa/4.0/
The gist: Recent advances in generative foundation models have expanded the capabilities of embodied AI, enabling increasingly general navigation, manipulation, and long-horizon task execution.
Terminology
Abstract
Recent advances in generative foundation models have expanded the capabilities of embodied AI, enabling increasingly general navigation, manipulation, and long-horizon task execution. However, effective interaction with dynamic environments requires more than only strong perception and reasoning: agents must act with partial knowledge and retain, retrieve, and update information from past observations, actions, and experiences. This survey examines embodied memory as a mechanism for preserving decision-relevant information beyond the current observation. We organize existing approaches along two complementary dimensions: six fundamental memory requirements, spanning world-state persistence, task progress, belief maintenance, and experiential learning; and eight architectural categories, including recurrent states, fast weights, visual and symbolic representations, spatial memory, and experience-based systems. We analyze their capabilities and limitations, and examine how existing benchmarks evaluate memory-dependent behavior, distinguishing genuine use of historical information from general task performance. Finally, we identify open challenges in selecting relevant experiences, maintaining consistent beliefs in changing environments, and consolidating knowledge across episodes. Our analysis provides a framework for understanding memory as an essential component of reliable, adaptive, and lifelong-horizon embodied intelligence.
Sources
- ReMem-VLA: Empowering Vision-Language-Action Model with Memory via Dual-Level Recurrent Queries
- $\mu$VLA: On Recurrent Memory for Partially Observable Manipulation in VLA Models
- KEMO: Event-Driven Keyframe Memory for Long-Horizon Robot Manipulation with VLA Policies
- MEM: Multi-Scale Embodied Memory for Vision Language Action Models
- Being-0: A Humanoid Robotic Agent with Vision-Language Models and Modular Skills
- Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments
- PRACTICE: From Experience to Expertise in Self-Evolving Embodied Agents
- RMBench: Memory-Dependent Robotic Manipulation Benchmark with Insights into Policy Design
- RT-1: Robotics Transformer for Real-World Control at Scale
- Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
- OpenVLA: An Open-Source Vision-Language-Action Model
- $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization
- ContextVLA: Vision-Language-Action Model with Amortized Multi-Frame Context
- 4D-VLA: Spatiotemporal Vision-Language-Action Pretraining with Cross-Scene Calibration
- Unsupervised Predictive Memory in a Goal-Directed Agent
- Recursive Belief Vision Language Action Models
- Action-Effect Memory Pretraining for Robot Manipulation
- Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training
- RoboTTT: Context Scaling for Robot Policies
- WAM-TTT: Steering World-Action Models by Watching Human Play at Test Time
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation