Elevator group control system using multiagent system

Abstract

AbstractIn this study, the authors proposed an elevator group control system using multiagent system, and verified the efficiency of the system in areas characterized by concentration of large crowds of passengers on the main floor, for example, during morning rush‐hour traffic. Congestion on the main floor in heavy morning rush‐hour traffic areas was previously alleviated using specific traffic control systems providing so‐called morning rush‐hour operation by forced dispatching of multiple elevators to the main floor. However, since forced dispatching on the main floor affects areas other than the main floor and also the subsequent service on the main floor, it makes overall improvement of transportation efficiency a difficult task. There are published reports relating to morning rush‐hour operation, but these deal with centralized control based on overall monitoring and sending instruction commands. Thus, the problem is that if this control system stopped working, it caused overall malfunctioning. In addition, in a system based on overall monitoring and generating optimal control plan in accordance with the received information, it is not easy to generate design updates following a change in the system's design or environment. Therefore, in the current study the authors introduced a multiagent system with distributed processing, in which individual elevators make action decisions by independent learning process. The authors proposed an elevator group control system which uses simplified reasoning to provide for flexible response even in the case of a changed environment, such as when a part of the elevator group has stopped functioning. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 34(1): 45–58, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.1183