Meta-Level Incomplete Information
暂无分享,去创建一个
Typically, when we discuss planning with incomplete info rmation, we are referring to incomplete information about the state of the world or the effects of actions. One common approach to dealing with this type of incomplete information is to use a probabi listie model to reason about the partial information that is available. Tu make reasoning with probabili st ic models tractable, assumptions and approximations are made that necessarily make the model an incomplete representation of the partial information available on the state of the world . In this extended abstract, I will disc uss incomplete information concerning planning models of the world and the robot. Since this is incomplete information about models of incomplete information, I wiJl call it metalevel incomplete information . I will illustrate the so urces of this meta-level incomplete information and describe how a pianner can make use of this type of incomplete information. I will use a parti cuiar probabilistic model, a partially obsenable Markov deci si on process, and examples from work with the Xavier robot to make the di sc ussion more concrete.
[1] Reid G. Simmons,et al. Robot Navigation with Markov Models: A Framework for Path Planning and Learning with Limited Computational Resources , 1995, Reasoning with Uncertainty in Robotics.
[2] Leslie Pack Kaelbling,et al. Acting Optimally in Partially Observable Stochastic Domains , 1994, AAAI.
[3] Reid G. Simmons,et al. Probabilistic Robot Navigation in Partially Observable Environments , 1995, IJCAI.
[4] Mark S. Boddy,et al. An Analysis of Time-Dependent Planning , 1988, AAAI.