Two-Stage Coarse-to-Fine AutoRL for DDPG Hyperparameter Optimization

Zenodo (CERN European Organization for Nuclear Research)

Abstract

Code release for a two-stage coarse-to-fine AutoRL workflow for DDPG hyperparameter optimization, demonstrated on a pendulum swing-up control task implemented with MATLAB, Simulink, and Optuna. The repository bundles a local pendulum Simulink model adapted from the MathWorks reinforcement learning example and provides MATLAB entrypoints, optional Python-based AutoRL scripts, and sanitized HPC templates.

Authors 1

1. Alexander Winkler corresponding Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University