This repository contains the code for our paper Safety-Constrained Policy Transfer with Successor Features (ICRA-23).
In this work, we focus on the problem of safe policy transfer in reinforcement learning: we seek to leverage existing policies when learning a new task with specified constraints. This problem is important for safety-critical applications where interactions are costly and unconstrained policies can lead to undesirable or dangerous outcomes, e.g., with physical robots that interact with humans. We propose a Constrained Markov Decision Process (CMDP) formulation that simultaneously enables the transfer of policies and adherence to safety constraints. Our formulation cleanly separates task goals from safety considerations and permits the specification of a wide variety of constraints. Our approach relies on a novel extension of generalized policy improvement to constrained settings via a Lagrangian formulation. We devise a dual optimization algorithm that estimates the optimal dual variable of a target task, thus enabling safe transfer of policies derived from successor features learned on source tasks. Our experiments in simulated domains show that our approach is effective; it visits unsafe states less frequently and outperforms alternative state-of-the-art methods when taking safety constraints into account.
This repo contains code that's based on the following code: RaSF.
If you find this repository or the ideas presented in our paper useful for your research, please consider citing our paper.
@INPROCEEDINGS{10161256,
author={Feng, Zeyu and Zhang, Bowen and Bi, Jianxin and Soh, Harold},
booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
title={Safety-Constrained Policy Transfer with Successor Features},
year={2023},
volume={},
number={},
pages={7219-7225},
keywords={Automation;Reinforcement learning;Markov processes;Safety;Task analysis;Robots;Optimization},
doi={10.1109/ICRA48891.2023.10161256}
}
Feel free to contact Zeyu Feng, Bowen Zhang, Jianxin Bi or Harold Soh for any questions regarding the code or the paper. Please visit our website for more information: CLeAR.