Robotics Research Papers — Page 7

Counterfactual Videos Scale Humanoid Object Handling from Simulation to Reality

How can humanoids learn to interact with the diverse objects encountered in everyday life?

Zihan Wang, Zhen Wu, Pieter Abbeel, Rocky Duan +5

Robots Learn New Manipulation Skills From Videos, Not Teleoperation

A visual-prompt learning recipe trains robots to interpret human manipulation videos while ignoring irrelevant changes in objects, people, and scene conditions.

Minxing Li, Minghao Han, Weizhi Zhao +10

AISep 30

How Robots Improve Policies by Reusing Skills to Repair Their Failures

Scaling robots to a wider range of real-world tasks requires a way for their policies to keep improving after deployment, without relying on continual human guidance.

Zihang Rui, Renhao Wang, Haoxu Huang +1

AISep 29

Human Demonstrations Help Robot Hands Learn Object Manipulation

Prithwish Dan, Chenyang Ma, Wei Zhan

RoboticsSep 29

RoboCompiler: Compiling Closed-Chain Robots for Consistent Control and Simulation

Mehdi Heydari Shahna, Joongheon Kim, Jouni Mattila

AISep 29

How Robots Learn to Model Hidden Disturbances for Better Robot Control

A learned, contractive disturbance model gives adaptive controllers a way to predict hidden forces that change over time.

Min Kim, José Leonardo Brenes, Fred Hadaegh +1

dronesSep 28

SatNav Benchmarks Long-Horizon UAV Navigation from Satellite Images

SatNav defines three task families—Boundary, Landmark, and Route—to evaluate complementary aspects of memory-intensive, long-horizon navigation.

Jiajun Jiang, Chunliang Hua, Zichun Chen +4

HumanoidSep 28

Humanoid Robots Learn to Walk, Run, Jump, and Read Rough Terrain

Zachary Olkin, William D. Compton, Aaron D. Ames

RoboticsSep 28

STL-Guided Robot Policy Learning from Sparse Success Signals for Control

Hongrui Zheng, Cristian Ioan Vasile, Antonio Loquercio +1

RoboticsSep 27

Training-Free Robot Behavior Cloning Through Retrieval and Action Continuation

Large datasets and pretrained representations have driven advances in robot imitation learning.

Maximilian Adang, Timothy Chen, Lars Osterberg +2

RoboticsSep 27

Choosing Environmental Contacts for Legged Loco-Manipulation

To address these challenges, we develop Capability-Tradeoff Contact Selection (CTCS) for task-dependent environmental contact selection.

Al Jaber Mahmud, Shuai Li, Xuan Wang

RoboticsSep 27

Ego-Exo4D Human Meshes Dataset for 4D Motion Reconstruction from Multi-View Video

Abhiram Maddukuri, Georgios Pavlakos

AISep 26

Object-Centric World Models Help Underwater ROVs Plan Safer Salvage

Yuncong Yang, Jinlong Li, Yulong Xue +4

AISep 26

Dense 3D Point Tracking with Sliding-Window De-Duplication

At the end of each window, we voxelize the feature cloud based on predicted target positions.

Ayush Jain, Sreeharsha Paruchuri, Ishita Gupta +5

AISep 26

How Coding Agents Solve Generalized Robot Task and Motion Planning

Each environment is represented by an MDP and a task description.

Matteo Merler, Bowen Li, Josh Roy +4

AISep 25

Rolling World Action Models for Faster Closed-Loop Robot Manipulation

Yinghua Zhou, Junjie Ye, Yiqi Zhao +8

AISep 25

RAPID Enables Robot Manipulation Programs to Generalize Across Diverse New Scenes

Yuyao Liu, Jiayuan Mao, David Hsu +2

AISep 25

Action-Discriminative World Models Improve Counterfactual Robot Control

The training and planning pipeline of AD-WM uses residual latent prediction and action-recovery objectives to shape action-discriminative transitions.

Jiabin Qiu, Zixuan Chen, Hongye Cao +3

AISep 24

How DART Restores Lost Motion in Frozen World Models for Robots

DART diagnoses why frozen latent-flow world models lose motion and restores dynamic predictions without unfreezing their visual representation.

Xiwen Chen, Rigaudiere Z. Li, Zhiruo Zhou +2

RoboticsSep 24

Always-On Robots Watch, Recall, and Act Across Concurrent Dual-Arm Streams

Ding Yi, Peiwen Sun, Chenchu Rong +4