Robot Policy Adaptation

Follow this topic Ask about this topic

Robot policy adaptation studies how a trained controller can adjust when tasks, surroundings, sensors, or robot dynamics differ from training. Common approaches range from fine-tuning with demonstrations or online reinforcement learning to lightweight residual modules, latent or noise-space steering, and world-model-guided correction of a frozen policy. Work spans manipulation, locomotion, and navigation, with evaluation in simulation and on real robots using benchmarks such as LIBERO and RoboTwin. A central challenge is improving performance with little feedback while preserving skills, safety, and reliable behavior under distribution shifts.

Latest papers

175 papers57 papers in the last four weeks, up 148% on the four weeks before. 0.3% of all new papers.

Cross-Controller Transfer via In-Context Behavior Adaptation
Oct 29, 2026Jonathan Yang, Chelsea Finn, Dorsa SadighRobot Policy Adaptation Robot Policy Generalization
Department of Computer Science, Stanford University, Stanford, CA, USA
Value Function-Guided MPPI for Hierarchical Planning and Control in Off-Road Autonomous Navigation
Oct 28, 2026Anoushka Alavilli, Raymond Song, Guanya Shi +1Off-Road Robot Navigation Hierarchical Robot Control
The authors are with the Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213, USA.
When, Where, and What to Correct: Human-Guided Adaptive Residual Correction for Robot Policies
Oct 27, 2026Anukriti Singh, Maxime Caniot, Pratap Tokekar +1Robot Policy Adaptation Residual Policy Learning
University of Maryland, College Park College Park, MD · Work done during internship at Enchanted Tools. · Enchanted Tools Paris, France
GRAFT: Geometry-Aware Residual Adaptation for Foundation Trackers
Oct 27, 2026Liang Li, Guoyao Zhang, Xingyu Chen +6Humanoid Robot Locomotion Terrain-Aware Robot Locomotion
Xiaomi Robotics Lab. · HKUST(Guangzhou).
Program-Level Adversarial Co-Adaptive Self-Play for Robotic Security
Oct 26, 2026Leo Benaharon, Chi Zhang, Wei Zheng +2Robot Policy Adaptation Self-Evolving Agents
The authors are with Isuzu Technical Center of America, Plymouth, MI 48170 USA.
RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies
Oct 26, 2026Heyang Yu, Haomin Li, Sihang Li +7Robot Policy Adaptation Test-Time Adaptation
Tsinghua University · New York University · Northwestern University
Benchmarking Obstacle-Conditioned Safety in Household Robot Navigation
Oct 26, 2026Hyunwoong Bae, Jisu Han, Jiyeon Choi +2Safe Robot Navigation Robot Policy Adaptation
Interdisciplinary Program in Artificial Intelligence, Seoul National University · Department of Electrical and Computer Engineering, Seoul National University
HumanoidNDM: Closing the Training-Deployment Gap of Motion Tracking via Humanoid Neural Dynamics Model
Oct 25, 2026Chao Chen, Zhenrong Wang, Jilong Wang +4Robot Policy Adaptation Humanoid Motion Tracking
CFCS, School of Computer Science, Peking University · Galbot · Institute of Automation, Chinese Academy of Sciences +2
RoboLongR: Causally Consistent Rewards Teach Robots to Progressively Master Long-Horizon Manipulation
Oct 24, 2026Zixuan Chen, Lin Cao, Fenggang Zhang +5Long-Horizon Robotic Manipulation Robot Policy Adaptation
Nanjing University · Southeast University · Fudan University
One Scan is Enough: Demonstration-Free Adaptation for Language-Guided Navigation
Oct 24, 2026Zhuoyuan Yu, Yuxing Long, Du Wang +3Vision-Language Navigation Robot Policy Adaptation
CFCS, School of Computer Science, Peking University · PrimeBot
Do Earth-Trained Humanoid Controllers Walk on the Moon? Gravity, Terrain, and Soil Transfer
Oct 21, 2026Guillem Casadesus Vila, Sri Aditya Deevi, Grace X. GaoHumanoid Robot Locomotion Legged Robotics
Department of Aeronautics and Astronautics, Stanford University, Stanford, CA 94305, USA
Flexible Language-Guided Representation Editing for Adaptive Robot Behavior
Oct 20, 2026Audrey Lee, Nathaniel Dennler, Ilia Sucholutsky +1Robot Policy Adaptation Language-Conditioned Robot Control
MIT Computer Science & Artificial Intelligence Laboratory · New York University
Self-Crystallizing Behavior Trees: Amortizing Foundation-Model Reasoning into Deterministic Control for Autonomous Drones
Oct 18, 2026Behnam MoradiBehavior Trees Robot Policy Adaptation
Software Department, KavAI. California, USA
Learning When to Adapt: Reactivity-Conditioned Residual Adaptation under Human-Robot Distribution Shift
Oct 18, 2026Immanuel R. Santjoko, Benedictus C. G. Cinun, Jing Chen +1Robot Policy Adaptation Social Robot Navigation
Department of Electrical & Computer Engineering, University of Houston · Department of Psychological Sciences, Rice University · Department of Engineering Technology, University of Houston
ChairNav: Cross-Embodiment Pre-training and Personalization for Long-Horizon Wheelchair Navigation
Oct 15, 2026Junhua Huang, Zhizheng Liu, Honglin He +2Robot Navigation Assistive Robotics
University of California, Los Angeles · Louisiana State University
Persistent Profile-Guided Reset-Free Recovery Improves Final Gait Selection after Irreversible Leg Damage
Oct 14, 2026Minjoong KimRobot Failure Recovery Robot Policy Adaptation
Minjoong Kim is an independent researcher.
Efficient Test-Time Adaptation For Robot Policies
Oct 14, 2026Motasem Alfarra, Pietro Mazzaglia, Markus Peschl +2Robot Policy Adaptation Test-Time RL
Qualcomm AI Research, Amsterdam, the Netherlands
MorphAct: Robot Policies Morphing with the Worlds They Act In
Oct 14, 2026Ziyan Liu, Jiting Liu, Weihong Zhong +7Robot Policy Adaptation VLMs for Robotics
University of Science and Technology of China · School of Artificial Intelligence, Shanghai Jiao Tong University · Tongji University +3
Back on Track: Deployment-time Policy Improvement via World Model Planning
Oct 14, 2026Colton Crosby, Hao Wang, Mac Schwager +1Robot Policy Adaptation World Model-Based Planning
The authors are associated with the Aeronautics and Astronautics Department, Stanford University. · The author is associated with the Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California.
From Failures to Learnable Worlds: Policy-Conditioned Environment Generation for Navigation Agents
Oct 14, 2026Yue Zhang, Han Lin, Zun Wang +3Robot Policy Adaptation Synthetic Data Generation for Robotics
UNC Chapel Hill
Collective Learning for Unified Robotic Manipulation via Predictive Dynamics Adaptation
Oct 13, 2026Shaocheng Liu, Yuan Meng, Zhenshan Bing +1Robotic Manipulation Cross-Embodiment Robot Learning
School of Computation, Information and Technology, Technical University of Munich, Garching, Germany. · State Key Laboratory for Novel Software Technology and the School of Science and Technology, Nanjing University (Suzhou Campus), China.
LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts
Oct 13, 2026Wenxuan Guo, Yingping Liang, Zhaoguo Wang +11Robot Policy Learning Robot Policy Adaptation
Tsinghua University · Dexmal · Loopeva
EqC: Equilibrium Energy Control for Introspective Lifelong Robot Learning
Oct 13, 2026Haodong Yang, Runhan Huang, Bowen Zhang +7Robot Policy Adaptation Continual Robot Learning
Tsinghua University · Galaxea AI · Princeton University
Steertuning Policies from a Few Human Videos
Oct 13, 2026Billy Yan, Isaac Tu, Kelly Lee +8Robot Policy Adaptation Data-Efficient Robot Learning
New York University
ReEnvision: Future-Guided Selective Action Repair for Long-Horizon Manipulation
Oct 13, 2026Xiyin Zeng, Xuewei Huang, Yi Lu +2Long-Horizon Robotic Manipulation Robot Policy Adaptation
Information Hub, The Hong Kong University of Science and Technology (Guangzhou) · Daimon Robotics · School of Intelligence Science and Technology, Nanjing University
MOSAIC: Mode- and Support-Aware Inference-Time Control of Robot Diffusion Models
Oct 12, 2026Sabrina Bodmer, Rahel Rickenbach, René Zurbrügg +2Inference-Time Guidance Diffusion Guidance
Department of Mechanical and Process Engineering, ETH Zürich, Switzerland. · Automatic Control Laboratory, EPFL, Switzerland.
SHOES: Sim-to-real Hardware-adaptation via On-policy Expansion of Skills for Legged Robots
Oct 12, 2026Kota Tsuruhara, Tsung-Yuan Chen, Kento KawaharazukaRobot Policy Adaptation Robot Locomotion
Tokyo University of Agriculture and Technology · The University of Tokyo
PathLift: Depth-Conditioned Metric Lifting for Visual-Navigation Priors
Oct 12, 2026Gwanjun Shin, Hyeongseok Suh, Hongmo Jung +1Robot Navigation Robot Policy Adaptation
Interdisciplinary Program in Artificial Intelligence, Seoul National University · Department of Mechanical Engineering, Seoul National University
RARA: Robust Policy Fine-Tuning through Retrieval-Anchored Representation Alignment for Robotic Manipulation
Oct 9, 2026Xuhui Kang, Sung-Wook Lee, Yan Zhuang +2Robot Policy Adaptation Robustness of VLA Models
University of Virginia
EndoRSI: Physical-Evidence-Driven Recursive Improvement of Endoscopic Robot Policies
Oct 9, 2026Chi Kit Ng, Yidong Zhang, Tianchun Wu +4Robotic Endoscopy Robot Policy Adaptation
Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong SAR, China · Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong SAR, China

Oct 29, 2026J/K move · Enter open · S save

Cross-Controller Transfer via In-Context Behavior Adaptation

Jonathan Yang, Chelsea Finn, Dorsa Sadigh

Department of Computer Science, Stanford University, Stanford, CA, USA

Robot Policy Adaptation Robot Policy Generalization Cross-Embodiment Robot Learning Transfer Learning in RL ICL for Robotics

Open paper AskPDF arXiv

Generalist robot policies increasingly learn from data collected across diverse tasks and embodiments, yet typically assume a single controller for each robot. This assumption limits transfer because different controllers can assign different physical meanings to the same action and may require different strategies to complete a task. We study how to transfer robot data across both embodiments and controllers without allowing conflicting action interpretations to interfere with learning. Our setting spans six controller types as well as variations in controller gains and compliance. We propose a framework that encodes behavioral interaction under a controller through examples of commanded actions and the resulting robot motions. A transformer context encoder summarizes these interactions, allowing a single policy to adapt in context. Across real-world experiments, our method improves average success by 33 percentage points over naive data pooling, achieves 50% success on unseen controller modes and 52% on unseen controller gains, and transfers to tasks with limited controller-specific data. On contact-rich tasks such as USB insertion and surface wiping, behavioral interaction context enables the policy to adapt not only its actions but also its strategy. Our results provide a path toward generalist robot policies that can leverage data collected across diverse action interpretations. For more information, visit https://controller-transfer.github.io.