DCBF-GSAC algorithm

Published: 21 July 2025| Version 1 | DOI: 10.17632/w5ns39mzy8.1
Contributor:
Kunyu Wang

Description

Here, we provide the code for our proposed GSAC and DCBF-GSAC algorithms, whose core iterative learning concept is proposed for the first time. GSAC is a standard DRL algorithm, similar to SAC, TD3, and others, and can be applied to arbitrary control tasks. DCBF-GSAC builds upon GSAC by incorporating a safety controller based on DCBF and effectively integrating the safety experience provided by DCBF with GSAC. DCBF-GSAC is a safe DRL framework that can also be combined with other safety controllers. The provided code includes detailed configuration examples, making it easy to deploy in specified control tasks.

Files

Categories

Deep Reinforcement Learning

Licence