Skip to content

Event

ResMerge: Residual Policy Learning and Merging for Continual Adaptation

1 reports · 1 sources

Kefei Zhu and other authors propose the ResMerge two-stage framework, which learns mergeable residual policies on top of a frozen pretrained robot policy: in the adaptation stage, residual reinforcement learning is used to obtain task-specific corrections without the need for additional expert demonstrations; subsequently, compatible residuals are merged in parameter space to reduce stored parameters. The authors state that experiments on two backbone models of the LIBERO benchmark and on a real Franka robot show that its continual learning performance is superior to the strong PECFT baseline. The above results are as reported in the paper.

Report timeline

  1. ResMerge: Residual Policy Learning and Merging for Continual Adaptation of Pre-trained Robot Policies

    ResMerge proposes a two-stage framework that learns mergeable residual policies atop a frozen pre-trained policy: residual reinforcement learning obtains task-specific corrections during adaptation without additional expert demonstrations, while parameter-space merging consolidates compatible residuals to reduce stored parameters.

    arXiv cs.RO ·

Related topics

Grasping and manipulationRobot models and learningResearch papers

Interest trend

Current
3
Peak
4Nov 4 01:00
Past 24 hours
–

The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.