arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.10207cs.GT
收藏

高级Petri博弈中策略直接生成的规范表示(完整版)

Canonical Representations for Direct Generation of Strategies in High-level Petri Games (Full Version)

Manuel Gieseking, Nick Würdemann

摘要:

Petri博弈是分布式系统中综合问题的多人博弈模型,即,本地控制器的自动生成。该模型代表的因果记忆的球员,这是一个Petri网上的令牌,分为两个团队:可控系统和不可控的环境。对于一个环境局中人和有限数目的系统局中人,求解Petri对策的问题可以归结为求解Büchi对策的问题。高级Petri博弈是普通Petri博弈的一种简洁表示。对称性,来自一个高层次的表示,可以利用显着减少相应的Büchi游戏中的状态空间。我们提出了一个新的结构,解决高层次的Petri游戏。它涉及到一个独特的,规范的表示减少Büchi游戏的定义。这使我们能够将布奇博弈中的策略直接转化为Petri博弈中的策略。应用于六个结构不同的基准家庭的实现显示在大多数情况下,更大的状态空间的性能提高。

英文摘要:

Petri games are a multi-player game model for the synthesis problem in distributed systems, i.e., the automatic generation of local controllers. The model represents causal memory of the players, which are tokens on a Petri net and divided into two teams: the controllable system and the uncontrollable environment. For one environment player and a bounded number of system players, the problem of solving Petri games can be reduced to that of solving Büchi games. High-level Petri games are a concise representation of ordinary Petri games. Symmetries, derived from a high-level representation, can be exploited to significantly reduce the state space in the corresponding Büchi game. We present a new construction for solving high-level Petri games. It involves the definition of a unique, canonical representation of the reduced Büchi game. This allows us to translate a strategy in the Büchi game directly into a strategy in the Petri game. An implementation applied on six structurally different benchmark families shows in most cases a performance increase for larger state spaces.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑