Thermo-VLA: A Thermal-Aware Vision-Language-Action Model with Gated Residual Modulation for Robotic Manipulation
Thermo-VLA: A Thermal-Aware Vision-Language-Action Model with Gated Residual Modulation for Robotic Manipulation
This research presents Thermo-VLA, a new Vision-Language-Action model designed for robotic manipulation that incorporates thermal sensing.
The paper introduces Thermo-VLA, a Vision-Language-Action model that uses gated residual modulation to decompose thermal images into cold and hot attributes for temperature-aware robotic manipulation, achieving an average success rate of 0.86 on four thermal-dependent tasks with a dual-arm Unitree G1 platform.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
success_rate
0.86
View original evidence
nd hot attribute representations and injects them into the RGB token through language-conditioned residual modulation, enabling temperature-driven target selection and continuous manipulation. We further construct six real-world robotic tasks on a dual-arm Unitree G1 platform, spanning basic grounding to relational reasoning. Experiments show that Thermo-VLA achieves an average success rate of 0.86 across four therma
Open source S3tasks
6
View original evidence
nd hot attribute representations and injects them into the RGB token through language-conditioned residual modulation, enabling temperature-driven target selection and continuous manipulation. We further construct six real-world robotic tasks on a dual-arm Unitree G1 platform, spanning basic grounding to relational reasoning. Experiments show that Thermo-VLA achieves an average success rate of 0.86 across four therma
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
nd hot attribute representations and injects them into the RGB token through language-conditioned residual modulation, enabling temperature-driven target selection and continuous manipulation. We further construct six real-world robotic tasks on a dual-arm Unitree G1 platform, spanning basic grounding to relational reasoning. Experiments show that Thermo-VLA achieves an average success rate of 0.86 across four therma
Open source S3
rovements in both simulation and real-robot experiments, indicating its potential to generalise beyond thermal attributes. Project website: https://j-x-chen.github.io/Thermo-VLA-web/
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org