Multi-Objective SPIBB: Seldonian Offline Policy Improvement with Safety Constraints in Finite MDPs

TL;DR

This paper introduces a multi-objective safe policy improvement method in offline RL that guarantees performance bounds while balancing multiple reward signals, demonstrated on synthetic and real-world healthcare tasks.

Contribution

It extends SPIBB to handle multiple objectives with user preferences, providing high-probability safety guarantees in finite MDPs for offline RL.

Findings

Effective in synthetic grid-world safety task

Successful application in critical care for sepsis treatment

Provides performance guarantees with multiple objectives

Abstract

We study the problem of Safe Policy Improvement (SPI) under constraints in the offline Reinforcement Learning (RL) setting. We consider the scenario where: (i) we have a dataset collected under a known baseline policy, (ii) multiple reward signals are received from the environment inducing as many objectives to optimize. We present an SPI formulation for this RL setting that takes into account the preferences of the algorithm's user for handling the trade-offs for different reward signals while ensuring that the new policy performs at least as well as the baseline policy along each individual objective. We build on traditional SPI algorithms and propose a novel method based on Safe Policy Iteration with Baseline Bootstrapping (SPIBB, Laroche et al., 2019) that provides high probability guarantees on the performance of the agent in the true environment. We show the effectiveness of our method on a synthetic grid-world safety task as well as in a real-world critical care context to learn a policy for the administration of IV fluids and vasopressors to treat sepsis.