Improving elevator dynamic control policies based on energy and demand visibility

Abstract

Elevator management system has received impressive attention due to its significance to transportation effectiveness for mid and high building. An important thing to improve elevator management system is to collect external information. This paper presents a method to collect number of passenger by using cameras and deep learning. By considering the status inside elevators, the directions of passenger movement, and the number of waiting passengers, the system occasionally allocates multiple elevators for a single hall call, which assists in reducing passengers' waiting time. This study applies deep learning to identify number of people queuing for an elevator. Data is gathered through some cameras to be analyzed with Region Based Convolutional Neural Network (R-CNN). Finally, we optimize the elevator dispatching rule by adding queuing data to the traditional elevator control system. Our goal is to minimize the average waiting time of the passengers.