Safety-Constrained Policy Transfer with Successor Features

TL;DR

This paper introduces a method for safe policy transfer in reinforcement learning using successor features and a constrained MDP framework, ensuring safety constraints are met during transfer and outperforming existing methods.

Contribution

It presents a novel extension of generalized policy improvement for constrained settings and a dual optimization algorithm for safe policy transfer using successor features.

Findings

Reduces unsafe state visits in simulations

Outperforms existing safety-aware transfer methods

Effectively separates task goals from safety constraints

Abstract

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.