Mobile Robot Navigation Strategies Through Behavioral Cloning and Generative Adversarial Imitation Learning

Kevin Silva, Rodrigo Calvo

2023

Abstract

The conception of robots able to navigate autonomously through several environments remains one of the main challenges to be overcome in the robotics research. The wide use of machine learning techniques as imitation learning has obtained efficient performance in this research field. The autonomous navigation is essential to carry out many kinds of task, which it can reduce the time and computational cust. One of the mechanisms to a robot be autonomous is observe the behavior of the other. Therefore, it is proposed in this research the development of a strategy of navigation based on Generative Adversarial Imitation Learning (GAIL) for the learning of the navigational behaviors of distinct mobile robots with different locomotion strategies. The CoppeliaSim simulator is used to build virtual scenarios and execute experiments to gather samples for the strategy training. Another strategy will be developed based on the behavioral cloning, which will also be trained in some environments with the same samples used in GAIL. Regression error metrics in addition with the comparison of the paths generated by the strategies in each scenario will be considered as evaluation methods. The obtained results are then discussed along with the potential future works.

Paper Citation

in Harvard Style

Silva K. and Calvo R. (2023). Mobile Robot Navigation Strategies Through Behavioral Cloning and Generative Adversarial Imitation Learning. In Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS, ISBN 978-989-758-648-4, SciTePress, pages 517-524. DOI: 10.5220/0011856700003467

in Bibtex Style

@conference{iceis23,
author={Kevin Silva and Rodrigo Calvo},
title={Mobile Robot Navigation Strategies Through Behavioral Cloning and Generative Adversarial Imitation Learning},
booktitle={Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS,},
year={2023},
pages={517-524},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011856700003467},
isbn={978-989-758-648-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS,
TI - Mobile Robot Navigation Strategies Through Behavioral Cloning and Generative Adversarial Imitation Learning
SN - 978-989-758-648-4
AU - Silva K.
AU - Calvo R.
PY - 2023
SP - 517
EP - 524
DO - 10.5220/0011856700003467
PB - SciTePress