Autors: Budakova, D. V. Title: Reinforcement learning using for planning with compromises of a rescue operation Keywords: Reinforcement learning, planning, compromises, rescue Abstract: This article explores the creation of a library of plans, including the possibilities offered by the use of available specialized means and equipment in rescue operations. As a hazardous environment, a 3D model of an electrical substation is considered, and intelligent agent behavior is modeled with application in a serious game. The proposed modification of the Reinforcement Q-learning algorithm includes the introduction of matrices for the intensity of the characteristics of the considered disaster; safety thresholds; and new rules for choosing the next possible position in the evacuation path. References Issue
|
Вид: публикация в международен форум, публикация в реферирано издание, индексирана в Scopus