Radboud University Nijmegen, The Netherlands
Wednesday 18 February 2009
Seminar Room B10 (Basement)
Alexandra House, 17 Queen Square, London, WC1N 3AR
Control as an (approximate) inference problem
We reformulate a class of non-linear stochastic optimal control problems introduced by Todorov as a KL minimization problem. As a result, the optimal control computation reduces to an inference computation and approximate inference methods can be applied to efficiently compute approximate optimal controls. We show that the previously introduced path integral approach for continuous control problems can be obtained as a special case of the KL control problem. We provide an example of a block stacking task where we demonstrate how approximate inference can be successfully applied to instances that are too large for exact computation.