Destination Floors-Oriented Group Elevators Dispatching Algorithm Based on Q-Learning and Simulated Annealing Method

Optimal dispatching strategies wield a pivotal influence on improving the operational performance of modern hybrid group elevators, which directly contributes to the heightened operational efficiency of the elevators and an improved traveling experience for passengers. Existing elevator dispatching algorithms are struggling to cope with the intricacies and diversities present in real-world scenarios, especially in instances of peak hours in superstructures. Designing an elevator destination floor group dispatching algorithm has posed intricate challenges due to the complexity and non-linearity of elevator systems and unpredictability of the passenger flow. This paper proposes a novel algorithm for elevator destination floor group dispatching that skillfully combines the Q-Iearning algorithm and the simulated annealing algorithm in alignment with current state-of-the-art investigations within the domain. Numerical simulations are conducted to verify the effectiveness of the proposed composite algorithm in reducing the average waiting time of passengers comparing to the traditional proximity principle and conventional dispatching algorithm.