What question did this study set out to answer?

The aim is to identify components of action determination in humanoid locomotion using a new framework.

April 10, 20260 citationsOpen Access

Decomposing Gait Robustness in Demonstration-Free Humanoid Locomotion: A Predictive Diagnostic for ROM-Guided Reinforcement Learning

Key Points

The aim is to identify components of action determination in humanoid locomotion using a new framework.
Introduced a diagnostic decomposition for locomotion controllers.
Partitioned action determination into External Perception, Constraint Enforcement, and Policy Residue.
Applied the decomposition to the ROM-GRL framework.
Proposed four experiments with falsification criteria for evaluation.
No experimental results are reported in the presented work.

Abstract

We propose a diagnostic decomposition for reinforcement learning locomotion controllers, partitioning action determination into External Perception (E), Constraint Enforcement (C), and Policy Residue (R), normalized to E+C+R=1. Applied to the ROM-GRL framework (Liu & Lau, 2025), the decomposition provides a vocabulary for distinguishing sensor dependence, reward overfitting, and autonomous gait mastery. Four experiments with falsification criteria are proposed. No experimental results are presented.

KI fragen

Bookmark

View Full Paper