Articles de recherche en robotique — Résumés en langage clair | RobotOverflow | Robot Overflow

Autonomous Robot Manipulation Skills Through Language-Driven Quality-Diversity

Autonomous Robot Manipulation Skills Through Language-Driven Quality-Diversity

Overall, most methods for autonomously acquiring robot skills focus on obtaining single, high-performing solutions.

Émiland Garrabé, Mahdi Khoramshahi, Stéphane Doncieux

Two Arms Build 3D Maps from Calibrated RGB-D Data for Robots

Two Arms Build 3D Maps from Calibrated RGB-D Data for Robots

DARP is a calibrated dual-arm RGB-D-IR dataset that aligns complementary robot-camera views for measurable 3D perception.

Manish Kansana, Mohammed Yusuf Mujawar, Sudip Mittal +2

Persistent Programs for Language-Grounded Robot Control, Learning, and Transfer

Persistent Programs for Language-Grounded Robot Control, Learning, and Transfer

Data generators scale robot experience by retargeting object-relative demonstration segments across scene configurations.

Weiqi Wang, Zhi Li, Yudong Lei +7

Pictura Trains Self-Driving Policies with Perspective Self-Play at Scale

Pictura Trains Self-Driving Policies with Perspective Self-Play at Scale

Yuan Yin, Elias Ramzi, Marc Lafon +8

Particle World Modeling for Text-Assisted Robot Video Prediction

Particle World Modeling for Text-Assisted Robot Video Prediction

Yafei Zhang, Nan Wu

ChainSplat Helps Robots Learn Flexible Object Motion From RGB Video

Roboticsavant-hier

ChainSplat Helps Robots Learn Flexible Object Motion From RGB Video

ChainSplat learns a compact physics-based digital twin for deformable linear objects from RGB video, enabling prediction, state estimation, and manipulation.

Seungyeon Kim, Noémie Jaquier

Transitional AV Lane-Change Dataset Reveals How Gaps Converge and Collision Risk Peaks

Roboticsil y a 3 j

Transitional AV Lane-Change Dataset Reveals How Gaps Converge and Collision Risk Peaks

New field experiments with a transitional autonomous vehicle (tAV) performing 78 mandatory lane changes on a public road show that lead and lag gaps consiste…

Abhinav Sharma, Md Abdullah Al Hasan, Danjue Chen +1

New Model Teaches Robots to Predict Visual Changes with Uncertain Latent Actions

New Model Teaches Robots to Predict Visual Changes with Uncertain Latent Actions

Zuojin Tang, Feifan Luo, Haoyun Liu +10

Riemann-1.0 Builds Embodied World Action Models for Physical AI

Roboticsil y a 3 j

Riemann-1.0 Builds Embodied World Action Models for Physical AI

Haofeng Sun, Jiangbo Pei, Fei Kang +13

Robotic Active Sensing Maps Plant Stress Across Individual Plant Leaves

Robotic Active Sensing Maps Plant Stress Across Individual Plant Leaves

Plant phenotyping based on observable traits plays a critical role in modern agriculture, both for stress detection and genotype-linked breeding.

Ayman Laaroussi, Peter Hanappe, David Colliaux

State-Aware Task Planning with Multimodal LLMs for Human-Robot Collaboration

State-Aware Task Planning with Multimodal LLMs for Human-Robot Collaboration

STEP enhances long-term action anticipation with multimodal large language models by explicitly modeling state transitions.

Maitrey Gramopadhye, Prakash Baskaran, Xiao Liu +2

Scheduling Marine Autonomous Vehicle Fleets for Greater Scientific Impact

Roboticsil y a 4 j

Scheduling Marine Autonomous Vehicle Fleets for Greater Scientific Impact

Mehdi El Krari, Jonathan Smith, Maria Fox

Task-Aware 3D Detection Improves Fast LiDAR Robot Perception

Task-Aware 3D Detection Improves Fast LiDAR Robot Perception

Su Wang, Yaochen Li, Min Yang +3

FlashVLA Streams Robot Actions for Faster Asynchronous VLA Control

Roboticsil y a 5 j

FlashVLA Streams Robot Actions for Faster Asynchronous VLA Control

Pretrained VLAs are trained to denoise each chunk from scratch in isolation and have never seen a partially populated buffer.

Zekai Li, Jiaming Tang, Zhijian Liu

Cross-Embodiment Video World Models Enable Zero-Shot Physical Simulation

Cross-Embodiment Video World Models Enable Zero-Shot Physical Simulation

The defining breakthroughs in large language models were precipitated by web-scale training on diverse text datasets.

Kechen Liu, Ola Shorinwa

ROS 2 Unity Toolkit for Autonomous Mobile Robot Fleet Management

Roboticsil y a 6 j

ROS 2 Unity Toolkit for Autonomous Mobile Robot Fleet Management

Andrea Drudi, Lorenzo Pichierri, Andrea Testa +1

Fast Generative Grasping Uses Lie Groups for Faster Real Robot Control

Roboticsil y a 6 j

Fast Generative Grasping Uses Lie Groups for Faster Real Robot Control

S. Talha Bukhari, Yi Wei, Ruiqi Ni +2

Zero-WAM: Human Video Instructions for Open-Ended Robot Task Generalization

Roboticsil y a 6 j

Zero-WAM: Human Video Instructions for Open-Ended Robot Task Generalization

In zero-shot cross-task manipulation, the robot must execute a novel task without finetuning.

Jiaming Zhou, Qihang Zhang, Gangwei Xu +9

ODEWorld Brings Continuous-Time World Modeling to Robot Manipulation

ODEWorld Brings Continuous-Time World Modeling to Robot Manipulation

Continuous-time world model ODEWorld hits 83.2% task success on LIBERO-LONG and translates language instructions into visual goals.

Dongxiu Liu, Haoyi Niu, Peng Cheng +5

4D Radar Odometry That Holds Up in Fog, Rain, and Dust

Roboticsil y a 7 j

4D Radar Odometry That Holds Up in Fog, Rain, and Dust

Sangwoo Jung, Dongjae Lee, Chiyun Noh +1