Михаил Леонтьевич Воскобойников
◽
Роман Константинович Федоров
◽
Геннадий Михайлович Ружников
Предложен метод автоматизации активации устройств Интернета вещей на основе классификации геопозиции мобильного устройства. В отличие от других методов пользователь обучает систему активации устройств с помощью примеров и контрпримеров, что значительно снижает требования к квалификации пользователя. Проведено тестирование метода на таких двух устройствах, как шлагбаум и электромеханический замок двери. Полученные результаты тестирования позволяют судить о работоспособности метода и возможности его использования в системах умного дома и города.
Most IoT devices provide an application programming interface such as web service that allows controlling these IoT devices over Internet using a mobile phone. Activation of IoT devices is performed according to the status of user behavior. Both user behavior and activation of IoT devices are periodical. An activation of IoT device is often related with a user geolocation which can be defined by sensors of the mobile device. A method for automated activation of IoT devices based on classification of geolocation of mobile device is proposed. The method implements a supervised learning that simplifies automate activation of IoT devices for the end users. Existing methods demand appropriate end user qualification and require long time to automate activation. For indoor geolocation of the mobile device information from Wi-Fi access points and geolocation GPS sensor is utilized. Data of Wi-Fi and GPS sensors is used to form context of a mobile device. Based on context examples of invoking/not invoking web services the spatial areas are formed. When the mobile device context is within the web service invocation area, the web service is invoked and the associated IoT device is activated. To implement the method, an Android application was developed. The method was tested on a training set that contained 100 training examples of calling two web services: opening an electromechanical door lock and opening a barrier. As a result of testing, the accuracy of classifying the context of a mobile device was 98 percent. The results obtained can be used in the development of smart home and smart city systems.
Modeling User's Non-Functional Preferences for Personalized Service Ranking
