FR-VLA research uses task context to regulate robot force feedback

Researchers report that FR-VLA, a robot control model, achieved the highest mean success rate over six evaluated stages in experiments covering five contact-rich tasks on two robotic platforms. The model uses visual and language context to decide when force feedback should influence an action and how strongly. A learned signal regulates the force information entering action generation, rather than sending it through a fixed pathway. Trajectory visualizations showed that this signal changed with the interaction phase. The findings are reported in an arXiv preprint.