@INPROCEEDINGS{lang-toussaint10icml,
    author = "Lang, Tobias and Toussaint, Marc",
    title = "Probabilistic Backward and Forward Reasoning in Stochastic Relational Worlds",
    booktitle = "Proc.\textasciitilde of the Int.\textasciitilde Conf.\textasciitilde on Machine Learning (ICML)",
    year = "2010",
    month = "June",
    abstract = "Inference in graphical models has emerged as a promising technique for planning.A recent approach to decision-theoretic planning in relational domainsuses forward inference in dynamic Bayesian networks compiled from learnedprobabilistic relational rules. Inspired by work in non-relational domains withsmall state spaces, we derive a backpropagation method for such nets inrelational domains starting from a goal state mixture distribution. We combinethis with forward reasoning in a bidirectional two-filter approach. Weperform experiments in a complex 3D simulated desktop environment with anarticulated manipulator and realistic physics. Empirical results show thatbidirectional probabilistic reasoning can lead to more efficient and accurateplanning in comparison to pure forward reasoning.",
    folder = "Robotics",
    pdf = "http://www.user.tu-berlin.de/lang/pub/10-lang-toussaint-ICML.pdf",
    pdfurl = "http://www.user.tu-berlin.de/lang/pub/10-lang-toussaint-ICML.pdf"
}