Embodied intelligence system improves by learning from real interaction
ME-Brain-1.0: Memory, Cognition and Action for Evolving Embodied Intelligence
Robotics
Summary
Many robots and embodied systems today cannot learn or improve after they are built and deployed because their skills are fixed. The authors introduce ME-Brain, a system that can evolve by itself, using its own experiences to get better over time without needing new training from humans. This happens through three parts: memory that organizes experiences, a thinking core that turns experience into skills, and an action model that focuses on important moments when deciding what to do. ME-Brain showed better success on many robot tasks compared to previous systems.
What this means in practice
- •For robot developers: Create robots that improve their skills by learning autonomously while interacting with the real world after deployment.$Commercial implications: Enables selling adaptive robots that get better over time for industries needing continuously improving performance.
- •For industrial automation teams: Implement adaptive control systems that evolve from physical experience without repeated retraining, enhancing automated processes robustness.
Authors
Wei He, Hengtao Li, Zhongrui Yu, Xuhan Zhu, Maokui He, Zide Liu, Xiyue Zhang, Xianwei Mao, Chunpeng Zhou, Jia Shi, Yanze Xin, Jingwen Li, Jingxie Zheng, Sijie Zeng, Chenfeng Wang, Fan Lu, Zeyu Zhang, Shuai Guo, Hengxuan Zhang, Pengfei Yu, Jia Shi, Yu Liu, Kun Zhan, Yan Xie
Abstract
Current embodied systems largely rely on pretrained capabilities that remain fixed after deployment, limiting their ability to learn from physical interaction. We introduce MachEmbodied-Brain (ME-Brain), a self-evolving embodied system organized around a closed loop of action execution, experience acquisition, experience evolution, and improved execution. Evolvable Memory consolidates multimodal trajectories into hierarchical, reusable experience; Cognitive Core transforms physical experience into transferable skills; and the Action Model combines event-driven keyframes, EventCell local-world prediction, and action-conditioned memory modulation to focus computation on decision-critical moments, regions, and historical evidence. Together, these modules shift embodied intelligence from train-and-freeze to deploy-and-evolve without model retraining. Cognitive Core outperforms the strongest comparison models by 8.2 and 9.6 points on embodied and agent benchmarks. The Action Model achieves 47.88% mean success on RoboMME, a 3.26-point improvement over the strongest baseline. On RoboDojo, it reaches a 21.51 mean Score and 16.03% success rate, exceeding $π_{0.5}$ by 10.10 and 9.12 points. On the six-task ME-RealBench, ME-Brain achieves a 69.5 mean Score and 66.7% success rate, outperforming DM0.5 by 12.8 and 11.7 points, respectively.