Output details
11 - Computer Science and Informatics
King's College London
A Translation-based Approach to Contingent Planning
<22>The paper presents a translation from deterministic planning problems involving uncertainty and partial observability (contingent problems) into fully-observable non-deterministic planning problems that is exponential in a width parameter. The translation that is provable complete for problems with width equal to 1 is used as the basis for a contingent planner, CLG, that is shown to scale-up much better than other planners do. The ideas underlying the translation form the basis for some of the best current planners that sense such as the K-replanner (Bonet and Geffner, IJCAI-2011), and SDR/MPSR (Brafman and Shani, IJCAI-2011, AAAI-2012).