Abstract
Modern elevator systems are controlled by the elevator group controllers that assign moving and stopping policies to the elevator cars. Designing an adequate elevator group control (EGC) policy is challenging for a number of reasons, one of them being conflicting optimization objectives. We address this task by formulating a corresponding constrained multiobjective optimization problem, and, in contrast to most studies in this domain, approach it using true multiobjective optimization methods capable of finding approximations for Pareto-optimal solutions. Specifically, we apply five multiobjective optimization algorithms with default constraint handling techniques and demonstrate their performance in optimizing EGC for nine elevator systems of various complexity. The experimental results confirm the scalability of the proposed methodology and suggest that NSGA-II equipped with the constrained-domination principle is the best performing algorithm on the test EGC systems. The proposed problem formulation and methodology allow for better understanding of the EGC design problem and provide insightful information to the stakeholders involved in deciding on elevator system configurations and control policies.
Highlights
We proposed a bi-objective problem formulation of the elevator group control (EGC) optimization problem and used the sequential ring (S-Ring) model [3] to evaluate the solutions of the resulting optimization problem
Tuned algorithm parameter values for multiobjective evolutionary algorithms (MOEAs) aggregated over test elevator system configurations of the same size: total number of solution evaluations fe, population size np, number of generations ng, crossover probability pc, mutation probability pm, and scaling factor F
We explored the optimization of EGC, which is a task relevant in the design and operation of multi-car elevator systems
Summary
The resulting multiobjective optimization function is highly nonlinear and multimodal, highly dynamic, and stochastic, mainly because passengers do not arrive in a deterministic manner, but based on a stochastic process These problem properties render classic, gradient-based optimizers as not applicable and require advanced search strategies [1]. A true multiobjective optimization approach is needed to obtain a set of trade-off solutions if preferences between the objectives are not known in advance. We address the related issue by extending the initial bi-objective formulation with a constraint that limits the number of elevator car skips. The addition of this constraint fundamentally changes the EGC optimization problem and requires dedicated algorithms to solve the resulting constrained multiobjective optimization problem.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have