Performance Optimisation of Knowledge-Based Elevator Group Supervisory Control System

A new system for elevator group-supervisory control is presented in this paper. It addresses the problem of finding optimum routes for multi-lift system with the objective of reducing the overall trip time (waiting time + boarding time) for passengers. This has been made possible firstly by executing exhaustive search for all possible moves, and secondly by incorporating a computer vision system for counting people waiting at all landings. Searching the whole space for best move is practically impossible in large buildings with four or more lifts, given the current computer technology. Therefore, pattern classification and clustering techniques, based on nearest neighbour or minimum distance, have been employed to minimise the decision-making or processing time significantly.