动作(物理)
计算机科学
国家(计算机科学)
马尔可夫决策过程
颗粒过滤器
计算
树(集合论)
算法
可见的
马尔可夫链
搜索树
离散时间和连续时间
数学优化
理论计算机科学
马尔可夫过程
数学
人工智能
搜索算法
机器学习
卡尔曼滤波器
物理
数学分析
统计
量子力学
作者
Zachary N. Sunberg,Mykel J. Kochenderfer
出处
期刊:
[Association for the Advancement of Artificial Intelligence]
日期:2018-06-15
卷期号:28: 259-263
被引量:150
标识
DOI:10.1609/icaps.v28i1.13882
摘要
Online solvers for partially observable Markov decision processes have been applied to problems with large discrete state spaces, but continuous state, action, and observation spaces remain a challenge. This paper begins by investigating double progressive widening (DPW) as a solution to this challenge. However, we prove that this modification alone is not sufficient because the belief representations in the search tree collapse to a single particle causing the algorithm to converge to a policy that is suboptimal regardless of the computation time. This paper proposes and evaluates two new algorithms, POMCPOW and PFT-DPW, that overcome this deficiency by using weighted particle filtering. Simulation results show that these modifications allow the algorithms to be successful where previous approaches fail.
科研通智能强力驱动
Strongly Powered by AbleSci AI