Elevator Traffic Pattern Recognition by Artificial Neural Network

Elevator control is ordered hierarchically: Single Car Control (SCC); Group Supervisory Control (GSC) and Building Supervisory Control (BSC). The BSC is well suited to the application of Artificial Neural Networks (ANNs). They are used to learn the traffic patterns of the building for the subsequent tuning of GSC parameters. Several parameters, such as number of car stops, number of landing calls and their ratios and changes in car weights etc., can be used as input variables to the ANN. Initially, supervised learning mode is used and later unsupervised learning is employed. The ANN provides settings for ‘weightings’ that describe the proportion of each type of traffic pattern within a mixed case which normally exists in real practice in a modern commercial building. Five types of traffic patterns, namely up-peak, down-peak, peak demand floor, four-way traffic and off-peak, can be recognised by the ANN.