Closed-Loop Robot Autonomy for Embodied Intelligence

Robots are increasingly expected to operate beyond structured laboratories, where they must interpret incomplete sensory observations, make long horizon decisions, and execute physically grounded actions in changing environments. This shift has exposed a central limitation of many existing robotic systems: perception, decision making, learning, and control are often developed as separate components, while real world autonomy requires them to interact continuously through feedback. Progress in data driven skill acquisition, semantic world modeling, model-based decision making, and adaptive feedback control has begun to narrow this gap, but deploying such methods on physical robots remains challenging. Distribution shift, partial perception, contact uncertainty, safety constraints, and task level generalization continue to limit robustness. This Research Topic focuses on closed loop robot autonomy, with an emphasis on integrated approaches that connect what robots perceive, infer, plan, and execute in real-world embodied settings.The goal of this Research Topic is to advance the methods, theories, and systems that enable robots to operate autonomously in complex physical environments. The topic welcomes contributions across the core components of embodied autonomy, including perception, learning, planning, decision making, and control, as well as studies on how these components interact in closed loop robotic systems. A central challenge is that many robotic methods still struggle to generalize beyond controlled settings, especially under partial observability, distribution shift, contact uncertainty, dynamic environments, and long horizon task requirements. Addressing these limitations requires progress both within individual subfields and at their interfaces. Recent developments in data driven skill acquisition, semantic and object centric representations, task and motion planning, foundation models, adaptive control, and real robot benchmarking create new opportunities to improve robustness, scalability, and deployability in real world embodied settings.This Research Topic invites original contributions that advance embodied robot autonomy from both component level and system level perspectives. Relevant themes include robotic perception and environment understanding, semantic and object centric mapping, active perception, imitation learning and skill acquisition, task and motion planning, decision making under uncertainty, learning based and model based control, human robot interaction, mobile manipulation, and real world robot deployment. We also welcome work on knowledge distillation, representation transfer, foundation models, vision language action models, and efficient policy learning for resource-constrained robotic systems. Topic editors: Yansong Wu, Kuanqi Cai, Weinan Chen, Chaoqun Wang Source: https://www.frontiersin.org/research-topics/82866/closed-loop-robot-autonomy-for-embodied-intelligence