Abstract

Abstract. The past few years have seen a flurry of new approaches for planning under uncertainty, but their applicability to real-world problems is yet to be established since they have been tested only on toy benchmark problems. To fill this gap, the challenge of solving power supply restoration problems with existing planning tools has recently been issued. This requires the ability to deal with incompletely specified initial conditions, fault conditions, unpredictable action effects, and partial observability in real-time. This paper reports a first response to this nontrivial challenge, using the approach of planning via symbolic model-checking as implemented in the MBP planner. We show how the problem can be encoded in MBP’s input language, and report very promising experimental results on a number of significant test cases. 1

Citations