Fast Gaussian process occupancy maps (GPOM) for dynamic environments using Big Data GP
Required software: python 2.7 (last tested), numpy, matplotlib, sklearn, and GPflow 0.3.5 (last tested)
Demonstration The demo input file has 173234 data points. A conventional Gaussian process is limited to a few thousand data points.
Paper:
@inproceedings{senanayake2017learning,
title={Learning highly dynamic environments with stochastic variational inference},
author={Senanayake, Ransalu and O'Callaghan, Simon and Ramos, Fabio},
booktitle={Robotics and Automation (ICRA), 2017 IEEE International Conference on},
pages={2532--2539},
year={2017},
organization={IEEE}
}
Video: https://youtu.be/RItH8HH82ss