AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

A Primer on SO(3) Action Representations in Deep Reinforcement Learning

Martin Schuck, Sherif Samy, Angela P. Schoellig

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Learned state and action representations

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 delta tangent-space actions in the local frame instead of global matrix actionsPPO/SAC/TD3 · Idealized SO(3) rotation environmentepisode return: SAC dense -4.7 -> -2.9; SAC sparse -29.4 -> -7.9Improved
scale tangent increments to the permissible rotation range (ablation)delta tangent actions · Idealized SO(3) rotation environment (dense)episode return: about -1.5 difference without scalingImproved
add unit-rotation centering to the policy mean for delta quaternion/matrix actions (ablation)PPO · Idealized SO(3) rotation environmentlearning curveImproved
add unit-rotation centering to the policy mean for delta quaternion/matrix actions (ablation)SAC and TD3 · Idealized SO(3) rotation environmentlearning curveMixed
project sampled actions onto the manifold before computing probabilities (ablation)PPO with quaternion/matrix actions · Idealized SO(3) rotation environment (dense)learning curveHurt
project sampled actions onto the manifold (ablation)SAC with quaternion/matrix actions · Idealized SO(3) rotation environment (dense)learning curveNo clear effect
use quaternion actions (double cover) instead of rotation matricesglobal actions · Idealized SO(3) rotation environmentepisode return: PPO -5.4 -> -11.5Hurt
increase entropy coefficient (ablation)PPO/SAC with tangent, quaternion, matrix actions · Idealized SO(3) rotation environmentfinal policyNo clear effect
use SAC with entropy maximization and off-manifold actions under sparse rewardsglobal matrix actions · Idealized SO(3) rotation environment (sparse)episode return: TD3 -6.4 vs SAC -29.4Hurt
use delta tangent actions on drone control with PPOPPO · Drone trajectory tracking and racingrewardImproved
use delta tangent actions instead of other representationsTD3+HER · PickAndPlaceOrientsuccess rate: 54.1% -> 69.8%Improved
use delta tangent actions instead of quaternionsSAC · RoboSuite (9 tasks)rewardNo clear effect

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