Learning a dynamical system model for a spatiotemporal field using a mobile sensing robot

@inproceedings{lan_learning_2017,
  address = {Seattle, WA, USA},
  title = {Learning a dynamical system model for a spatiotemporal field using a mobile sensing robot},
  isbn = {978-1-5090-5992-8},
  url = {http://ieeexplore.ieee.org/document/7962949/},
  abstract = {This paper considers the problem of learning dynamic spatiotemporal fields using measurements from a single sensing robot. We first introduce a widely used parametric dynamical model for the spatiotemporal field. We then propose a motion strategy that can be used by a single sensing robot to collect sensor measurements about the field. Our motion strategy is designed to collect sufficient information at repeated locations but different times along an arbitrarily chosen periodic trajectory. In conjunction with the measurements collected, we propose a new learning algorithm based on subspace identification to learn the parameters of the dynamical model. We prove that the parameters learned by our algorithm will converge to the true parameters as long as the periodic trajectory is uniformly observable and the number of measurements collected by the robot goes to infinity. The performance of our algorithm is demonstrated in numerical simulations.},
  language = {en},
  urldate = {2020-09-15},
  booktitle = {2017 {American} {Control} {Conference} ({ACC})},
  publisher = {IEEE},
  author = {Lan, Xiaodong and Schwager, Mac},
  month = may,
  year = {2017},
  keywords = {learning\_dynamical\_systems},
  pages = {170--175},
  month_numeric = {5}
}