Ant Colony Algorithm and Fuzzy Neural Network-based Intelligent Dispatching Algorithm of An Elevator Group Control System

To improve the performance of elevator group control systems (EGCS), an intelligent dispatching method based on ant colony algorithm and fuzzy neural network is presented. An elevator group control system based on fuzzy neural network adapts to various traffic flow modes. Using ant colony algorithm to optimize the weights of fuzzy neural network before training with BP algorithm can solve the problem that convergence of weights is easy to be trapped in local optimal values when trained just with BP algorithm. This intelligent dispatching algorithm makes the weights of fuzzy neural network more precise and reasonable. These weights greatly affect the performance of an EGCS. The results of simulation show that ant colony algorithm and fuzzy neural network greatly improves the performance of an EGCS. Its average waiting time is obviously shorter than that of the EGCS that is only based on fuzzy neural network.

I. Introduction

An elevator group control system (EGCS) is a kind of complex nonlinear discrete control system that systematically manages three or more elevators in a building in order to efficiently transport passengers. An EGCS consists of hall call buttons, car call buttons, elevators, and a group controller in which there is group control dispatching algorithm, the most important part of an EGCS. If a passenger wants to go to another floor, he or she presses a direction (hall call) button and waits for an elevator to arrive; in this case, the group controller selects the most appropriate elevator in the group, depending on the current positions, moving directions, current passengers' number in every elevator. Then the passenger enters the elevator and presses a floor (car call) button in the elevator. The objective of an EGCS is to provide passengers with high-quality service.