Xiaomi's robotics research paper accepted by CoRL 2026, with an oral presentation acceptance rate of only about 1%

·by Henderson·AI
Xiaomi's robotics research paper accepted by CoRL 2026, with an oral presentation acceptance rate of only about 1%

Xiaomi Technology officially announced on November 12 that the research paper from its Robotics Division, titled "PACE: Temporal-Role Tactile Conditioning for Chunk-Based Insertion Policies," has been accepted by the prestigious international conference CoRL 2026 in the field of robotics and machine learning, and has been selected for an oral presentation. According to relevant information from Xiaomi Robotics, CoRL is a top academic conference at the intersection of robotics and machine learning.

The 2026 conference received a total of 2,094 valid submissions, of which 687 were accepted, and only 21 were selected for oral presentations, accounting for approximately 1% of all submitted papers, reflecting the conference's recognition of the research's innovation and academic value.

This research focuses on the challenges of tactile perception and fine-tuning in robotic precision insertion tasks, proposing the PACE tactile conditioning strategy framework. Its innovation lies in organizing tactile information into three temporal roles: contact state, sudden contact events, and contact consequence prediction. These three roles are used for action generation, rapid correction during execution, and optimization of action representation during the training phase.

Xiaomi Robotics Research Recognized at International Academic Conference

It is worth mentioning that in five categories of real robotic insertion tasks, including keys, dual-pin plugs, USB, and network ports, the average success rate of this method reached 90.7%, significantly outperforming traditional tactile fusion methods. In the future, this technological approach is expected to further expand into more complex contact scenarios such as robotic dexterous manipulation and precision assembly, providing new technical pathways to enhance the accuracy and robustness of robotic fine operations.

H
About the author
Henderson