Combining Visual and Tactile Feedback for Sub-Millimeter Precision Assembly
和中国科学院联手,具身智能企业怎么把亚毫米级“精密操作”做成绝活?
A research team from the Chinese Academy of Sciences and PaXini has developed a method that uses tactile-enhanced reinforcement learning to improve the accuracy and stability of sub-millimeter precision assembly tasks. The approach combines imitation learning for initial positioning with reinforcement learning for fine-tuning during insertion, using high-precision tactile sensors to provide critical feedback. The study, presented at IROS 2026, shows that tactile feedback significantly increases success rates in complex, visually occluded assembly scenarios.
Source: Robotics — Chinese web discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
touch sensor sensitivity
0.01
View original evidence
在这项研究中, 帕西尼的PX6AX GEN3 凭借其原生6D感知、0.01N灵敏度、1000Hz数据输出频率、超高测力一致性以及高集成易用设计的能力,成为该研究经评估选用的触觉传感器,被集成到七轴机械臂上,负责实时测量与反馈接触状态。
Open source S18touch sensor data output frequency
1,000
View original evidence
在这项研究中, 帕西尼的PX6AX GEN3 凭借其原生6D感知、0.01N灵敏度、1000Hz数据输出频率、超高测力一致性以及高集成易用设计的能力,成为该研究经评估选用的触觉传感器,被集成到七轴机械臂上,负责实时测量与反馈接触状态。
Open source S18touch sensor measurement accuracy
0.05
View original evidence
在0.05毫米间隙下,仅靠视觉与位姿示教的纯模仿学习方案成功率直接挂零0%,无触觉的 IL+RL 方案为52%,而加入了指尖触觉反馈后,成功率陡升至67%。
Open source S20Reported success rate
67%
View original evidence
在0.05毫米间隙下,仅靠视觉与位姿示教的纯模仿学习方案成功率直接挂零0%,无触觉的 IL+RL 方案为52%,而加入了指尖触觉反馈后,成功率陡升至67%。
Open source S20Reported success rate
0%
View original evidence
在0.05毫米间隙下,仅靠视觉与位姿示教的纯模仿学习方案成功率直接挂零0%,无触觉的 IL+RL 方案为52%,而加入了指尖触觉反馈后,成功率陡升至67%。
Open source S20gap size in assembly task
0.05
View original evidence
在0.05毫米间隙下,仅靠视觉与位姿示教的纯模仿学习方案成功率直接挂零0%,无触觉的 IL+RL 方案为52%,而加入了指尖触觉反馈后,成功率陡升至67%。
Open source S20
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
在这项研究中, 帕西尼的PX6AX GEN3 凭借其原生6D感知、0.01N灵敏度、1000Hz数据输出频率、超高测力一致性以及高集成易用设计的能力,成为该研究经评估选用的触觉传感器,被集成到七轴机械臂上,负责实时测量与反馈接触状态。
Open source S18
在0.05毫米间隙下,仅靠视觉与位姿示教的纯模仿学习方案成功率直接挂零0%,无触觉的 IL+RL 方案为52%,而加入了指尖触觉反馈后,成功率陡升至67%。
Open source S20
Source:Robotics — Chinese web discovery · 163.com