@InProceedings{plagemann08iros,
  title     = {Learning Predictive Terrain Models for Legged Robot Locomotion},
  author    = {Plagemann, C. and Mischke, S. and Prentice, S. and Kersting, K.
    and Roy, N. and Burgard, W.},
  booktitle = {Proc.~of the IEEE/RSJ International Conference on
    Intelligent Robots and Systems (IROS)},
  address   = {Nice, France},
  year      = {2008},
  abstract  = {
Legged robots require the ability to build accurate models of their
environment in order to plan and execute their actions. We present a
novel, probabilistic terrain model based on Gaussian processes that
can be learned and updated efficiently using sparse approximation
techniques. The major benefit of our model is its ability to predict
elevations at unseen locations more reliably than alternative
approaches, while it also yields estimates of the predictive
uncertainties. In particular, our Gaussian process adapts its
covariance to the situation at hand, allowing more accurate inference
of terrain height at points that have not been directly observed. We
show how a conventional motion planner can use the learned terrain
model to to plan a path to a goal location, using a terrain-specific
cost model to accept or reject candidate footholds. In experiments
with a real quadruped robot equipped with a laser range finder, we
demonstrate the usefulness of our approach and discuss its benefits
compared to simpler terrain models such as elevations grids.
},
  note      = {To appear},
  pdfurl    = {http://www.informatik.uni-freiburg.de/~plagem/bib/plagemann08iros.pdf}
}