Environment-Independent Task Specifications via GLTL

TL;DR

This paper introduces GLTL, a probabilistic extension of Linear Temporal Logic for task specifications in Markov decision processes, aiming to be environment-independent and learnable in finite time.

Contribution

The paper presents a novel geometric LTL (GLTL) language that improves task specification flexibility and environment independence over traditional reward functions.

Findings

GLTL can specify reinforcement learning tasks straightforwardly.

GLTL demonstrates advantages in small environments.

GLTL permits finite-time approximations to learned specifications.

Abstract

We propose a new task-specification language for Markov decision processes that is designed to be an improvement over reward functions by being environment independent. The language is a variant of Linear Temporal Logic (LTL) that is extended to probabilistic specifications in a way that permits approximations to be learned in finite time. We provide several small environments that demonstrate the advantages of our geometric LTL (GLTL) language and illustrate how it can be used to specify standard reinforcement-learning tasks straightforwardly.