RoboMemory is a brain-inspired multi-memory framework for lifelong learning in physical systems. It addresses critical challenges such as continuous learning in real-world environments, multi-module memory latency, capturing task correlations, and mitigating infinite loops in closed-loop planning. Drawing on cognitive neuroscience, it integrates four core modules: an information preprocessor (thalamus-like), a lifelong embodied memory system (hippocampus-like), a closed-loop planning module (prefrontal cortex-like), and a low-level executor (cerebellum-like), enabling long-term planning and cumulative learning. The lifelong embodied memory system, at the heart of the framework, mitigates the inference speed issues of complex memory frameworks by parallelizing updates and retrievals across spatial, temporal, episodic, and semantic submodules. It integrates a dynamic knowledge graph (KG) and a consistent architectural design to enhance memory consistency and scalability. Evaluation results on EmbodiedBench demonstrate that RoboMemory achieves a new state-of-the-art (SOTA) benchmark, outperforming the open-source benchmark (Qwen2.5-VL-72B-Ins) by 25% on average and the closed-source state-of-the-art (SOTA) benchmark (Claude3.5-Sonnet) by 5%. Elimination studies validate core components (criticism, spatial memory, and long-term memory), and real-world deployments demonstrate significant improvements in success rates for repeated tasks, confirming its lifelong learning capabilities. RoboMemory mitigates high latency issues through scalability and serves as a baseline for integrating multi-mode memory systems into physical robots.
