AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

CO-PILOT: COllaborative Planning and reInforcement Learning On sub-Task curriculum

Shuang Ao, Tianyi Zhou, Guodong Long, Qinghua Lu, Liming Zhu, Jing Jiang

NeurIPS 2021 · Reinforcement learning › Visual and representation RL · approach: Goal-conditioned policy learning · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use CO-PILOT's learned planner with sub-task tree on top of SAC instead of plain SACSAC · Maze environmentsuccess rateImproved
Mutual training of planner and RL versus SoRB's planner built on a pre-trained RL policySoRB · Maze/Mujoco/BipedalWalkersample efficiencyImproved
Compare CO-PILOT with SoRB on final performanceSoRB · BipedalWalkerpercentage of mastered tracksNo clear effect
Use CO-PILOT instead of hierarchical RL (HRL)HRL · Mujoco Ant-v1average returnImproved
Vary sub-task tree depth K over 5 values (ablation)CO-PILOT · Mazesuccess rateMixed
Increase planner batch size b beyond 5 (ablation)CO-PILOT · Mazesuccess rateNo clear effect
More mutual-training episodes (Episode 1 to 5)CO-PILOT · MazeRL time cost per sub-taskImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy