Robot insertion improved by separating tactile signals over time
PACE: Temporal-Role Tactile Conditioning for Chunk-Based Insertion Policies
Robotics
Summary
Putting objects together precisely is hard for robots because they need to feel how things fit while also seeing them. The authors developed a method called PACE that splits the robot’s sense of touch into three parts based on when the touch happens: constant touch, sudden events, and future touch predictions. This helps the robot decide better how to move and adjust during insertion tasks. Tests showed that PACE helped robots succeed much more often than older methods that mixed all touch signals at once.
What this means in practice
- •For robotic automation engineers: Improve real-world robot assembly by using tactile signals separated over time to better handle insertion tasks with delicate fits.$Commercial implications: Enables production-grade robots to insert parts with ultra-fine clearance, improving manufacturing automation quality and reducing errors.
- •For robotics software developers: Develop robot control policies that separately handle persistent touch, sudden contact events, and predictions to increase insertion success rates.
Authors
Zhiyuan Chen, Shixiong Xu, Shuai Chen, Yangwei You
Abstract
Contact-rich insertion requires robots to combine visual alignment with tactile adjustment. However, existing visuotactile chunk-based policies collapse tactile feedback into a single generation-time fusion pathway, leaving persistent contact states, abrupt contact events, and future contact consequences implicitly modeled in a single representation. We present PACE (Phase-Aware Contact-Event Conditioning), a temporal-role tactile conditioning framework for chunk-based insertion policies. PACE organizes a shared tactile stream into three temporal roles: persistent contact states condition action-chunk generation, abrupt contact events correct the generated chunk, and contact-consequence prediction shapes action representations during training. This design preserves chunk-level action continuity while assigning distinct pathways to different temporal roles of contact feedback. Across five real-robot insertion tasks, PACE achieves an average success rate of 90.7%, outperforming vision-only and generation-time tactile-fusion baselines by over 44 percentage points; on the ManipulationNet benchmark, PACE maintains 88.7% success across six geometry-clearance configurations, including an industrial-grade 0.1-mm clearance. Ablations confirm the complementary contributions of the three temporal roles, mechanistic analyses show that PACE separates normal insertion progress from contact events requiring immediate correction, and cross-signal and cross-architecture experiments indicate that this temporal-role principle extends across different physical feedback signals and policy backbones. Videos and materials are available at https://pace-insertion.github.io/.