Kagan Tumer's Publications

Display Publications by [Year] [Type] [Topic]


Autonomous Multiagent Space Exploration with High Level Human Feedback. M. Colby, L. Yliniemi, and K. Tumer. Journal of Aerospace Information Systems, 2016. (to appear)

Abstract

Robotic space exploration missions have always pushed the limits of science and technology and will continue to do so by their very nature. Such missions are particularly challenging as they operate in environments with high uncertainty, light-time delays, and high mission costs. Artificial intelligence based multiagent systems can alleviate these concerns by creating autonomous multi-robot teams that can function in uncertain environments, navigate and operate without time-sensitive commands from Earth-bound scientists, and spread the mission cost across multiple platforms that will eliminate the danger of total mission loss in case of a malfunctioning robot.

In this work, we present a novel human in the loop cooperative coevolutionary algorithm to train a multi-robot system exploring an unknown environment. Autonomous agents learn to make low-level control decisions to maximize a science return, while human scientists on Earth learn the changing mission profiles and provide high-level objectives to the robots. Results demonstrate that our algorithm reduces the number of robots needed for a particular performance level tenfold compared to traditional cooperative coevolutionary algorithms, resulting in significantly lower mission costs. Further, the trained multi-robot system is extremely robust to noise, and 10% sensor and actuator noise does not alter system performance in a statistically significant manner. Finally, the system is extremely robust to agent failures. When 10% of the robots in the system fail, the overall system performance is reduced by less than 10%.

Download

[PDF]929.9kB  

BibTeX Entry

@article{tumer-colby_jais16,
author = {M. Colby and L. Yliniemi and K. Tumer},
title = {Autonomous Multiagent Space Exploration with High Level Human Feedback},
journal = {Journal of Aerospace Information Systems},
note={(to appear)},
abstract={
Robotic space exploration missions have always pushed the limits of science and technology and will continue to do so by their very nature. Such missions are particularly challenging as they operate in environments with high uncertainty,  light-time delays, and high mission costs.  Artificial intelligence based multiagent systems can alleviate these concerns by creating autonomous multi-robot teams that can function in uncertain environments, navigate and operate without time-sensitive commands from Earth-bound scientists, and spread the mission cost across multiple platforms that will eliminate the danger of total mission loss in case of a malfunctioning robot.
  <p> In this work, we present a novel human in the loop cooperative coevolutionary algorithm to train a multi-robot system exploring an unknown environment.  Autonomous agents learn to make low-level control decisions to maximize a science return, while human scientists on Earth learn the changing mission profiles and provide high-level objectives to the robots.  Results demonstrate that our algorithm reduces the number of robots needed for a particular performance level tenfold compared to traditional cooperative coevolutionary algorithms, resulting in significantly lower mission costs.  Further, the trained multi-robot system is extremely robust to noise, and 10% sensor and actuator noise does not alter system performance in a statistically significant manner.  Finally, the system is extremely robust to agent failures.  When 10% of the robots in the system fail, the overall system performance is reduced by less than 10%.},
	bib2html_pubtype = {Journal Articles},
	bib2html_rescat = {Evolutionary Algorithms, Multiagent Systems, Robotics},
year = {2016}
}

Generated by bib2html.pl (written by Patrick Riley ) on Tue Jun 26, 2018 19:10:42