Deliberation Scheduling for Time-Critical Sequential Decision Making

We describe a method for time-critical de­ cision making involving sequential tasks and stochastic processes. The method employs several iterative refinement routines for solv­ ing different aspects of the decision mak­ ing problem. This paper concentrates on the meta-level control problem of delibera­ tion scheduling, allocating computational re­ sources to these routines. We provide dif­ ferent models corresponding to optimization problems that capture the different circum­ stances and computational strategies for de­ cision making under time constraints. We consider precursor models in which all deci­ sion making is performed prior to execution and recurrent models in which decision mak­ ing is performed in parallel with execution, accounting for the states observed during ex­ ecution and anticipating future states. We describe algorithms for precursor and recur­ rent models and provide the results of our empirical investigations to date.