AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

End-to-End Training of Deep Visuomotor Policies

Sergey Levine, Chelsea Finn, Trevor Darrell, Pieter Abbeel

JMLR 2016 · Reinforcement learning › Visual and representation RL · approach: RL for combinatorial routing · 2,644 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
replace softmax + fully connected layer with spatial softmax followed by expectation (feature points) (ablation)Spatial softmax CNN · PR2 target pose estimationtest error (cm): 2.59 -> 1.30Improved
replace plain fully connected layer with spatial softmax + feature points (ablation)Spatial softmax CNN · PR2 target pose estimationtest error (cm): 4.75 -> 1.30Improved
replace max-pooling + fully connected with spatial softmax + feature points (ablation)Spatial softmax CNN · PR2 target pose estimationtest error (cm): 3.71 -> 1.30Improved
add max-pooling to the fully connected networkFully connected CNN · PR2 target pose estimationtest error (cm): 4.75 -> 3.71Improved
increase training perturbation from 0 cm to 2 cmLinear-Gaussian controller · PR2 lego block task at 2 cm test perturbationsuccessful trials out of 5: 3/5 -> 5/5Improved
increase training perturbation from 0 cm to 1 cmLinear-Gaussian controller · PR2 ring on peg task at 2 cm test perturbationsuccessful trials out of 5: 0/5 -> 3/5Improved
increase training perturbation from 1 cm to 2 cmLinear-Gaussian controller · PR2 ring on peg task at 2 cm test perturbationsuccessful trials out of 5: 3/5 -> 3/5No clear effect
use learned linear-Gaussian controller instead of kinematic baselineLinear-Gaussian controller · PR2 lego block task at 1 cm test perturbationsuccessful trials out of 5: 0/5 -> 5/5Improved
train vision layers jointly end-to-end instead of pretraining for pose prediction (ablation)Deep visuomotor policy · PR2 coat hanger, cube, hammer and bottle taskssuccess rateImproved
feed pretrained pose-estimate features (rather than predicted pose) to the control layers (ablation)Deep visuomotor policy · PR2 visuomotor taskssuccess rateImproved
add Gaussian mixture model prior to trajectory optimization (ablation)Guided policy search with linear-Gaussian controllers · simulated peg insertion, octopus arm, swimmingsamples requiredImproved

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