Intelligent escalator passenger safety management

Abstract

This article addresses an approach to intelligent safety control of passengers on escalators. The aim is to improve the accuracy of detecting threatening situations on escalators in the subway to make decisions to prevent threats and eliminate the consequences. The novelty of the approach lies in the complex processing of information from three types of sources (video, audio, sensors) using machine learning methods and recurrent neural networks with controlled elements. The conditions and indicators of safety assurance efficiency are clarified. New methods and algorithms for managing the safety of passengers on escalators are proposed. The architecture of a promising safety software system is developed, and implementation of its components for cloud and fog computing environments is provided. Modeling results confirm the capabilities and advantages of the proposed technological solutions for enhancing the safety of escalator passengers, efficiency of control decision making, and system usability. Due to the proposed solutions, it has become possible to increase the speed of identifying situations 3.5 times and increase the accuracy of their determination by 26%. The efficiency of decision making has increased by almost 30%.

История СПБ ФИЦ РАН: 45 Лет Научной Деятельности

“…Разработаны методы и система нейросетевого прогнозирования событий с непрерывным обучением, работающая с несовершенными выборками временных рядов, отличающиеся повышенной точностью, непрерывным учетом изменяющихся законов поведения наблюдаемых процессов, отсутствием прерывания обучения на время прогнозирования, правилами управления ассоциативным вызовом информации из памяти рекуррентных нейронных сетей, предназначенные для автономных интеллектуальных систем для прогнозирования потоков различных событий [87 -89]. В [90] рассмотрен подход к интеллектуальному контролю безопасности пассажиров на эскалаторах. Преследуется цель повышения точности обнаружения угрожающих ситуаций на эскалаторах в метрополитене и принятия решений по предотвращению угроз и устранению последствий.…”

Data fusion using neural networks for intelligent multilevel control of artificial and natural objects: a survey

“…The neurons of this layer serve as independent variables of any physical system. A separate class of combined RNNs should systems built on the basis of using two identical recurrent neural networks with controlled elements (Osipov and Osipova 2018) and a control unit for these networks (Osipov et al 2020(Osipov et al , 2021(Osipov et al , 2022. A generalized structure of such combined RNNs is shown in Fig.…”

“…An optimized and specialized hardware implementation can reduce system cost by optimizing the needed resources and decreasing power requirements while improving performance. Typically, AI algorithms are developed and tested using development platforms such as TensorFlow(Osipov et al 2022),…”

Fast Detection and Classification of Dangerous Urban Sounds Using Deep Learning

“…The majority of the research, previously published in the literature, is concerned with the detection and identification of guns in noise-free environments. There are some examples in recent studies [16][17][18]. Only two contributions, as far as the present authors are aware, deal with dangerous urban sound detection systems in reverberant environments.…”