Learning Spatial and Temporal Correlation for Navigation in a 2-Dimensional Continuous World
Abstract
A behavior-based architecture that enables a simulated agent to exist and navigate in an artificial environment without any kind of spatial representation is presented. Hebbian learning is used to combine reactive behaviors that enable the agent to exploit spatial and temporal regularities in the environment. The agent is then able to apply its innate behaviors in situations that were not initially designed to trigger these reactive behaviors. The system can also accommodate changes in the environment. Simulation results that measure the performance of the system are also presented. 1.
Cite
Text
Panangadan and Dyer. "Learning Spatial and Temporal Correlation for Navigation in a 2-Dimensional Continuous World." International Conference on Machine Learning, 2002.Markdown
[Panangadan and Dyer. "Learning Spatial and Temporal Correlation for Navigation in a 2-Dimensional Continuous World." International Conference on Machine Learning, 2002.](https://mlanthology.org/icml/2002/panangadan2002icml-learning/)BibTeX
@inproceedings{panangadan2002icml-learning,
title = {{Learning Spatial and Temporal Correlation for Navigation in a 2-Dimensional Continuous World}},
author = {Panangadan, Anand and Dyer, Michael G.},
booktitle = {International Conference on Machine Learning},
year = {2002},
pages = {474-481},
url = {https://mlanthology.org/icml/2002/panangadan2002icml-learning/}
}