| 摘要 |
<p>A rule base storing a plurality of sets of control rules is prepared, the group control performance of waiting time distribution is estimated for a set of rules in the rule base applied to current traffic, and the best set of rules is selected according to the estimated performance. A weight database storing weight parameters for a neural network corresponding to the set of rules and performance learning means for correcting the weight database according to the learning results of the neural network are provided so that the group control performance can be estimated by the neural network that uses the corrected weight parameter. Therefore, the best set of rules is always applied to control a plurality of elevators as a group to provide adequate service, and improved accuracy in estimation is achieved for the actual operation of a plurality of elevators.</p> |
