مقالات پژوهشی رباتیک — صفحهٔ 20

HEFT Enables Full-Size Humanoid Robots to Lift 24 kg During Teleoperation

HEFT enables a full-size humanoid to carry 24 kg while being teleoperated via VR, using a motion-adaptive payload curriculum and privileged training.

Chenxin Liu, Qingzhou Lu, Guangxiao Yang, Xuanyang Shi +3

AI۱۴ تیر

ACID Framework Boosts Robotic Planning With Action Consistency Check

A new decision-time planning framework called ACID adds a simple action consistency check to world model–based planning, improving control quality across rig…

Gawon Seo, Dongwon Kim, Suha Kwak

AI۱۴ تیر

LIME Teaches Robots Where to Look Based on Language Intent

Researchers have built LIME, an AI model that predicts where a camera should aim next based on natural-language instructions. By learning human-like camera m…

Boyang Sun, Jiajie Li, Yung-Hsu Yang +6

AI۱۴ تیر

WorldSample Boosts Real Robot Learning with World Model Data Augmentation

Yuquan Xue, Le Xu, Zeyi Liu +5

AI۱۳ تیر

Task-Agnostic Pretraining: Teaching Robots to Move Before Teaching Tasks

Junhao Shi, Siyin Wang, Xiaopeng Yu +3

Robotics۱۳ تیر

Differentiable Simulation Trains Drones to Intercept Intruders Using Only Direction

Researchers at RISC Lab have developed a quadrotor interception policy that tracks and catches an intruder drone using only the direction to the target—no di…

Michael Anoruo, Xiaoyu Tian, Abhishek Rathod +5

drones۱۳ تیر

QuadRocket Testbed Uses a Drone to Control a Rocket-Like Flying Pendulum

QuadRocket uses a quadrotor attached via universal joint to control a rocket-like body, enabling low-cost indoor thrust-vector control experiments.

Pedro Santos, Joel Reis, Paulo Oliveira +1

AI۱۲ تیر

Controllable Traffic Agents Learn Steering Behaviors From Latent Variables

Juanwu Lu, Junyu Zhu, Ziran Wang

Robotics۱۲ تیر

Embodied.cpp Enables Efficient C++ Inference for Embodied AI on Heterogeneous Robots

Ling Xu, Chuyu Han, Borui Li +6

Robotics۱۲ تیر

Visual-Tactile World Action Model for Contact-Rich Manipulation

VT-WAM couples tactile deformation dynamics with action prediction using joint visual-tactile-action flow matching for contact-rich manipulation

Shuai Tian, Yupeng Zheng, Yuhang Zheng +7

Robotics۱۱ تیر

Sensorless Teleoperation Uses Inverse Dynamics for Transparent Control

A new control architecture for human-scale bilateral teleoperation eliminates force/torque sensors by using inverse dynamics modeling to estimate external fo…

Amir Noohian, Dylan Miller, Justin Valentine +2

AI۱۱ تیر

GPU-Parallel Linearization Error Bounds for Real-Time Robust Optimal Control of Nonlinear and Neural Network Dynamics

Jeffrey Fang, Keyi Shen, Anutam Srinivasan +1

AI۱۱ تیر

FurnitureVLA: Teaching Robots to Assemble IKEA Furniture Using Vision-Language-Action Models

Chenyang Ma, Yue Yang, Radu Corcodel +4

Humanoid۱۰ تیر

PrimeU: Humanoid Robot That Learns Tasks by Watching Human Videos

Human-aligned humanoid PrimeU learns manipulation tasks from ordinary human videos using PhysDex, reducing robot data needs by orders of magnitude.

Xiaopeng Lin, Ruoqi Yang, Shijie Lian +14

AI۱۰ تیر

Freeform Preferences Help Robots Learn From Human Feedback Without Rigid Ratings

Learning robot policies from human preferences traditionally forces annotators to give a single binary score comparing two trajectories, collapsing complex j…

Marcel Torne, Anubha Mahajan, Abhijnya Bhat +1

Robotics۱۰ تیر

Disentangled Video World Model Speeds Robotic Planning Without Sacrificing Quality

Ziyu Shan, Zhenyu Wu, Xiaofeng Wang +2

Robotics۹ تیر

Sequential Planning via Anchored Robotic Keypoints: Training-Free Robot Manipulation

Bryce Grant, Aryeh Rothenberg, Logan Senning +3

AI۹ تیر

GROW²: Teaching Robots to Select Tools and Find Their Functional Regions

GROW² enables robots to pick the best improvised tool from a cluttered scene and find its functional region using vision-language models and 3D reconstruction.

Yuhong Deng, Yuyao Liu, David Hsu

Humanoid۹ تیر

Vision-Language-Kinematics: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed 3D Scenes

Connecting egocentric observations to whole-body action is a fundamental challenge for humanoid robots operating in human-centered environments. A humanoid m…

Yen-Jen Wang, Jiaman Li, Sirui Chen +9

Robotics۸ تیر

CacheMPC: Certified Cached Model Predictive Control for Quadruped Locomotion

Nimesh Khandelwal, Mehul Anand, Shakti S. Gupta +1

Papers پژوهشی

HEFT Enables Full-Size Humanoid Robots to Lift 24 kg During Teleoperation

ACID Framework Boosts Robotic Planning With Action Consistency Check

LIME Teaches Robots Where to Look Based on Language Intent

WorldSample Boosts Real Robot Learning with World Model Data Augmentation

Task-Agnostic Pretraining: Teaching Robots to Move Before Teaching Tasks

Differentiable Simulation Trains Drones to Intercept Intruders Using Only Direction

QuadRocket Testbed Uses a Drone to Control a Rocket-Like Flying Pendulum

Controllable Traffic Agents Learn Steering Behaviors From Latent Variables

Embodied.cpp Enables Efficient C++ Inference for Embodied AI on Heterogeneous Robots

Visual-Tactile World Action Model for Contact-Rich Manipulation

Sensorless Teleoperation Uses Inverse Dynamics for Transparent Control

GPU-Parallel Linearization Error Bounds for Real-Time Robust Optimal Control of Nonlinear and Neural Network Dynamics

FurnitureVLA: Teaching Robots to Assemble IKEA Furniture Using Vision-Language-Action Models

PrimeU: Humanoid Robot That Learns Tasks by Watching Human Videos

Freeform Preferences Help Robots Learn From Human Feedback Without Rigid Ratings

Disentangled Video World Model Speeds Robotic Planning Without Sacrificing Quality

Sequential Planning via Anchored Robotic Keypoints: Training-Free Robot Manipulation

GROW²: Teaching Robots to Select Tools and Find Their Functional Regions

Vision-Language-Kinematics: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed 3D Scenes

CacheMPC: Certified Cached Model Predictive Control for Quadruped Locomotion