Simulated Visual Navigation Dataset for "Two-Stage Visual Navigation by Deep Neural Networks and Multi-Goal Reinforcement Learning"

Published: 6 April 2020| Version 1 | DOI: 10.17632/w48zc9yd9d.1
Contributors:
,
,
,
,
,
,

Description

Condensed data set (train, validation, test) of two simulated environments in the Gazebo simulator. One small kitchen of approximately 45 square meters and a big room of roughly 140 square meters. The data can be loaded using numpy. The images are size 84x84x3. The labels for each image are (X position, Y position, sine of theta, and cosine of theta).

Files

Steps to reproduce

Please refer to https://github.com/ashantia-ai/borg.

Institutions

Rijksuniversiteit Groningen

Categories

Robotics, Computer Simulation Environment, Navigation

Licence