Learning Transferable Policies for Monocular Reactive MAV Control

被引：13

作者：

Daftry, Shreyansh ^{[1
]}

Bagnell, J. Andrew ^{[1
]}

Hebert, Martial ^{[1
]}

机构：

[1] Carnegie Mellon Univ, Inst Robot, Pittsburgh, PA 15213 USA

来源：

2016 INTERNATIONAL SYMPOSIUM ON EXPERIMENTAL ROBOTICS | 2017年 / 1卷

关键词：

Transfer learning; Domain adaptation; Reactive control; Autonomous monocular navigation; Micro aerial vehicles;

D O I：

10.1007/978-3-319-50115-4_1

中图分类号：

TP24 [机器人技术];

学科分类号：

080202 ; 1405 ;

摘要：

The ability to transfer knowledge gained in previous tasks into new contexts is one of the most important mechanisms of human learning. Despite this, adapting autonomous behavior to be reused in partially similar settings is still an open problem in current robotics research. In this paper, we take a small step in this direction and propose a generic framework for learning transferable motion policies. Our goal is to solve a learning problem in a target domain by utilizing the training data in a different but related source domain. We present this in the context of an autonomous MAV flight using monocular reactive control, and demonstrate the efficacy of our proposed approach through extensive real-world flight experiments in outdoor cluttered environments.

引用

页码：3 / 11

页数：9