Tech Report CS-94-35

Deliberation Scheduling for Time-Critical Scheduling in Stochastic Domains

Lloyd Greenwald, Thomas Dean

September 1994

Abstract:

In this work we explore the use of deliberation scheduling for allocating computation among competing decision procedures in solving time-critical scheduling problems in stochastic domains. We present a two-level deliberation scheduling solution. At the higher level deliberation scheduling is used to partition infinite horizon scheduling problems temporally and sequence and allot computation time among the resulting problem instances. At the lower level deliberation scheduling is used to combine phases of a decision procedure for solving each problem instance. Our approach extends the work of Dean, Kaelbling, Kirman and Nicholson on planning under time constraints in stochastic domains to handle more complicated scheduling problems. We borrow from operations research in applying bottleneck-centered scheduling heuristics to improve initial policies and make use of Monte Carlo simulation for selectively constructing partial policies in large state spaces. Additionally, we employ a variant of Drummond's situated control rules to constrain the space of possible actions.

(complete text in pdf or gzipped postscript)