Service Area-based Elevator Group Supervisory Control System using GNP with RL

Genetic network programming (GNP) was proposed several years ago as a new evolutionary computation method. Its unique features, such as highly compact structure, potential memory function, etc, are verified by many studies mainly on virtual world problems. Recently, GNP is also applied to some complicated real world problems like elevator group supervisory control systems (EGSCS) and stock price prediction systems. As we know, EGSCS is a very large scale stochastic dynamic optimization problem. Due to its vast state space, significant uncertainty and numerous resource constraints such as finite car capacities and registered hall/car calls, it is hard to manage EGSCS using conventional control methods. In this paper, we propose an enhanced algorithm of EGSCS using GNP with reinforcement learning (RL) where an importance weight tuning method and a car assignment policy based on service area are introduced