	US 12,129,150 B2
	Collaborative scheduling method for high-rise elevators based on internet of things
Fusheng Zhang, Suzhou (CN); Yang Ge, Suzhou (CN); Anbo Jiang, Suzhou (CN); Lingyun Ma, Suzhou (CN); Zhen Zhao, Suzhou (CN); Jianxin Ding, Suzhou (CN); Jiancong Qin, Suzhou (CN); Yong Ren, Suzhou (CN); Guodong Sun, Suzhou (CN); Yong Feng, Suzhou (CN); and Linzhong Tang, Suzhou (CN)
Assigned to Changshu Institute of Technology, (CN); and Dongnan Elevator Co., Ltd., (CN)
Filed by Changshu Institute of Technology, Suzhou (CN); and Dongnan Elevator Co., Ltd., Suzhou (CN)
Filed on Sep. 25, 2023, as Appl. No. 18/372,168.
Claims priority of application No. 202310000515.0 (CN), filed on Jan. 3, 2023.
Prior Publication US 2024/0217772 A1, Jul. 4, 2024
Int. Cl. B66B 1/24 (2006.01); B66B 1/28 (2006.01); B66B 1/34 (2006.01)

CPC B66B 1/28 (2013.01) [B66B 1/2408 (2013.01); B66B 1/3476 (2013.01); B66B 2201/222 (2013.01); B66B 2201/403 (2013.01)]

7 Claims

1. A collaborative scheduling method for a high-rise elevator based on Internet of Things, comprising the following steps:

obtaining a number of people carried by each elevator in an elevator group at a current moment;

obtaining a target distance corresponding to the current moment of each elevator based on a current position of each elevator in the elevator group, and a number of floors with an elevator request;

obtaining an image of the current moment of an elevator door on each floor;

obtaining a waiting number on each floor at the current moment based on the image of the current moment of the elevator door and an OpenPose model;

obtaining a number of people entering each elevator in each period of a preset historical day based on monitoring video data of the elevator door in preset historical days, and obtaining a proportion of people who took an elevator up in each period based on the number of people entering each elevator in each period and a running state of each elevator, and predicting a number of people who take the elevator up and a number of people who take the elevator down at the current moment on each floor based on a proportion of people who take the elevator up and a number of people waiting at a current moment on each floor, the running state comprises going up, going down or waiting;

constructing a feature vector corresponding to the current moment of each elevator based on a running state of each elevator in the elevator group at the current moment, a position of the current moment, the target distance and a number of people carried by the elevator;

constructing a feature vector corresponding to the current moment of a skyscraper based on the number of people who take the elevator up and the number of people who take the elevator down at the current moment on each floor, obtaining a state vector corresponding to the current moment based on the feature vector corresponding to the current moment of each elevator and the feature vector corresponding to the current moment of the skyscraper;

controlling each elevator based on the state vector and a trained learning network comprising a reinforcement learning network; and

obtaining a reward function of the reinforcement learning network according to the number of people carried by each elevator, a passenger contribution of each elevator, the number of people waiting on each floor, and a comprehensive passenger mobility, wherein the comprehensive passenger mobility is a sum of passenger movements on each floor at the current moment, wherein the reward function of the reinforcement learning network is as follows:

constructing a first reward function based on the comprehensive passenger mobility, the number of people waiting on each floor, and the number of people carried by each elevator:

wherein R_jis the first reward function, TranPeo_jis a sum of the number of people carried by all elevators at a j th moment, PeoNume_i,jis a number of people waiting on an i th floor at the j th moment, AllFlow_jis a comprehensive passenger mobility at the j th moment, μ is a first adjustment coefficient, λ is an adjustment parameter, M is a total number of floors that the elevator can reach, and Time is a time consumed before the elevator is turned; and

obtaining a slice reward function corresponding to each elevator according to a prediction accuracy corresponding to each moment, the passenger contribution of each elevator at each moment, and the first reward function wherein:

wherein Slice_k(R_j) is a slice reward corresponding to a k th elevator, Val_jis a passenger contribution of the k th elevator at the j th moment, and Acc_jis a prediction accuracy corresponding to the j th moment.