RI: Medium: Collaborative Research: Solving Stochastic Planning Problems Through Principled Determinization

在不确定条件下运行的概率和决策理论规划在科学和工程中有重要的应用，但由于这些规划器不能大规模地处理大型、复杂的问题，这些随机方法一直没有得到充分利用。相比之下，在过去的十年里，确定性规划技术的规模得到了极大的扩大，这些技术并没有直接应对同样的不确定性挑战。该项目为利用确定性规划技术的这一进展--无论是经典的、时间的还是部分满意的规划问题--以及随机规划提供了一个系统的框架，通过对确定性、事后优化和机器学习的理论和/或方法的新颖改编。该项目正在帮助弥合确定性规划社区和随机规划社区之间的传统鸿沟，并将随机规划技术应用于现实世界。该奖项是根据2009年《美国复苏和再投资法案》(公法111-5)提供资金的。

Probabilistic and decision-theoretic planning, which operates under conditions of uncertainty, has important applications in science and engineering, but such stochastic methods have been under-utilized because these planners do not scale to large, complicated problems. In contrast, the past decade has seen tremendous scale-up in deterministic planning techniques, which do not directly address the same challenges of uncertainty head on. This project provides a systematic framework for exploiting this progress in deterministic planning technology - be it classical, temporal or partial satisfaction planning problems - in stochastic planning as well, by novel adaptation of theory and/or methods of determinization, hindsight optimization, and machine learning. This project is helping to bridge the traditional divide between deterministic and stochastic planning communities, and it is bringing stochastic planning technology to real-world applications. This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).