Automatic Synthesis of Efficient Regular Strategies in Adversarial Patrolling Games
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F18%3A00100830" target="_blank" >RIV/00216224:14330/18:00100830 - isvavai.cz</a>
Výsledek na webu
<a href="http://dl.acm.org/citation.cfm?id=3237383.3237481" target="_blank" >http://dl.acm.org/citation.cfm?id=3237383.3237481</a>
DOI - Digital Object Identifier
—
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Automatic Synthesis of Efficient Regular Strategies in Adversarial Patrolling Games
Popis výsledku v původním jazyce
We give a polynomial-time algorithm for synthesizing efficient regular strategies in adversarial patrolling games with general topology. Regular strategies use finite memory to gather some relevant information about the history of Defender's moves which results in substantially better protection of the targets. So far, the scope of automatic strategy synthesis was limited to positional strategies (which ignore the history) or to regular strategies where the underlying finite-memory observer had to be supplied manually. Furthermore, the existing methods do not give any information on how far are the constructed strategies from being optimal. In this paper, we try to overcome these limitations. We develop a novel gradient-based algorithm for synthesizing regular strategies where the underlying finite-memory observers are constructed algorithmically. The running time of our algorithm is polynomial which makes the algorithm applicable to instances of realistic size. Furthermore, we develop an algorithm for computing an upper bound on the best achievable protection, and compare the quality of the constructed strategies against this bound. Thus, we can effectively measure the "distance'' of the constructed strategies from optimal strategies, and our experiments show that this distance is often quite small.
Název v anglickém jazyce
Automatic Synthesis of Efficient Regular Strategies in Adversarial Patrolling Games
Popis výsledku anglicky
We give a polynomial-time algorithm for synthesizing efficient regular strategies in adversarial patrolling games with general topology. Regular strategies use finite memory to gather some relevant information about the history of Defender's moves which results in substantially better protection of the targets. So far, the scope of automatic strategy synthesis was limited to positional strategies (which ignore the history) or to regular strategies where the underlying finite-memory observer had to be supplied manually. Furthermore, the existing methods do not give any information on how far are the constructed strategies from being optimal. In this paper, we try to overcome these limitations. We develop a novel gradient-based algorithm for synthesizing regular strategies where the underlying finite-memory observers are constructed algorithmically. The running time of our algorithm is polynomial which makes the algorithm applicable to instances of realistic size. Furthermore, we develop an algorithm for computing an upper bound on the best achievable protection, and compare the quality of the constructed strategies against this bound. Thus, we can effectively measure the "distance'' of the constructed strategies from optimal strategies, and our experiments show that this distance is often quite small.
Klasifikace
Druh
D - Stať ve sborníku
CEP obor
—
OECD FORD obor
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Návaznosti výsledku
Projekt
<a href="/cs/project/GA18-11193S" target="_blank" >GA18-11193S: Algoritmy pro diskrétní systémy a hry s nekonečně mnoha stavy</a><br>
Návaznosti
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)
Ostatní
Rok uplatnění
2018
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Údaje specifické pro druh výsledku
Název statě ve sborníku
Proceedings of the 2018 International Conference on Autonomous Agents & Multiagent Systems
ISBN
9781510868083
ISSN
1548-8403
e-ISSN
—
Počet stran výsledku
8
Strana od-do
659-666
Název nakladatele
International Foundation for Autonomous Agents and Multiagent Systems
Místo vydání
Richland, SC
Místo konání akce
Stockholm
Datum konání akce
1. 1. 2018
Typ akce podle státní příslušnosti
WRD - Celosvětová akce
Kód UT WoS článku
—