A rollout algorithm framework for heuristic solutions to finite-horizon stochastic dynamic programs

详细信息    查看全文

• 作者：Justin C. Goodson ; ^a ; ^{goodson@slu.edu} ; Barrett W. Thomas^b ; ^{barrett-thomas@uiowa.edu} ; Jeffrey W. Ohlmann^b ; ^{jeffrey-ohlmann@uiowa.edu}
• 关键词：Dynamic programming ; Rollout algorithm ; Stochastic dynamic programming ; Approximate dynamic programming
• 刊名：European Journal of Operational Research
• 出版年：2017
• 出版时间：1 April 2017
• 年：2017
• 卷：258
• 期：1
• 页码：216-229
• 全文大小：881 K
• 卷排序：258

文摘

We make recent advances in rollout algorithms more accessible. We formalize rollout variants exploiting the pre- and post-decision states. We present new analytical results relating the performance of rollout variants. Our policy-based proofs make a closer connection dynamic programming. We apply our framework to a dynamic and stochastic knapsack problem.