Learning to Navigate in Mazes with Novel Layouts using Abstract Top-down Maps

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Linfeng, Wong, Lawson L. S.
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910747793031168
author Zhao, Linfeng
Wong, Lawson L. S.
author_facet Zhao, Linfeng
Wong, Lawson L. S.
contents Learning navigation capabilities in different environments has long been one of the major challenges in decision-making. In this work, we focus on zero-shot navigation ability using given abstract $2$-D top-down maps. Like human navigation by reading a paper map, the agent reads the map as an image when navigating in a novel layout, after learning to navigate on a set of training maps. We propose a model-based reinforcement learning approach for this multi-task learning problem, where it jointly learns a hypermodel that takes top-down maps as input and predicts the weights of the transition network. We use the DeepMind Lab environment and customize layouts using generated maps. Our method can adapt better to novel environments in zero-shot and is more robust to noise.
format Preprint
id arxiv_https___arxiv_org_abs_2412_12024
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning to Navigate in Mazes with Novel Layouts using Abstract Top-down Maps
Zhao, Linfeng
Wong, Lawson L. S.
Machine Learning
Artificial Intelligence
Robotics
Learning navigation capabilities in different environments has long been one of the major challenges in decision-making. In this work, we focus on zero-shot navigation ability using given abstract $2$-D top-down maps. Like human navigation by reading a paper map, the agent reads the map as an image when navigating in a novel layout, after learning to navigate on a set of training maps. We propose a model-based reinforcement learning approach for this multi-task learning problem, where it jointly learns a hypermodel that takes top-down maps as input and predicts the weights of the transition network. We use the DeepMind Lab environment and customize layouts using generated maps. Our method can adapt better to novel environments in zero-shot and is more robust to noise.
title Learning to Navigate in Mazes with Novel Layouts using Abstract Top-down Maps
topic Machine Learning
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2412.12024