# Optimization Models and Numerical Algorithms for an Elevator Group Control System

Research output: Thesis › Doctoral Thesis › Collection of Articles

### Standard

**Optimization Models and Numerical Algorithms for an Elevator Group Control System.** / Sorsa, Janne.

Research output: Thesis › Doctoral Thesis › Collection of Articles

### Harvard

### APA

*Optimization Models and Numerical Algorithms for an Elevator Group Control System*. Aalto University.

### Vancouver

### Author

### Bibtex - Download

}

### RIS - Download

TY - THES

T1 - Optimization Models and Numerical Algorithms for an Elevator Group Control System

AU - Sorsa, Janne

PY - 2017

Y1 - 2017

N2 - An elevator group control system (EGCS) dispatches the elevators under its control to serve each passenger call. New passengers may appear while the elevators are serving their current assignments. This makes the control problem of the EGCS dynamic. The EGCS solves a snapshot instance of the elevator dispatching problem (EDP) anew in intervals less than a second. Its solution defines complete elevator routes by minimizing, e.g., passenger waiting or journey times, according to which the EGCS dispatches the elevators. In this dissertation, the EDP is considered as a bilevel, dynamic, stochastic, multi-objective and integer optimization problem. The bilevel approach is based on the distributed nature of a typical elevator system, in which each elevator has its own independent controller and the EGCS optimizes the service of the calls shared by the elevator group. Hence, the upper-level problem corresponding to the EGCS assigns calls to the elevators, and a lower-level problem for each elevator determines its route to serve the assigned calls. Near-future passenger arrivals are uncertain for the EDP. The number of passengers related to a call and completely new calls are modelled explicitly in the studied stochastic model. For the first time, the EDP takes into account the fact that passengers often travel in socially connected groups. Furthermore, the uncertainties are evaluated in multiple risk scenarios which converts the EDP into a robust optimization problem. Since the EDP is a difficult integer optimization problem, its solution time by an exact algorithm increases exponentially with respect to problem size. The main solution approach is based on a genetic algorithm suitable for real-time optimization in the EGCS. The EDP is also formulated as a mixed integer programming problem and solved by an exact algorithm, which are important for benchmarking and theoretical studies. The EGCS developed in this dissertation was installed as the first double-deck destination control system (DCS) in the world. However, the de facto standard DCS does not perform optimally mainly because of the immediate assignment policy, which does not allow the reassignments of the calls to other elevators. As one possible solution to this challenge, the delayed assignment policy is studied. It allows the reassignments until the last moment but requires new shared guidance devices. The first such system will be realized in the near future. According to the results, delayed assignments improve passenger service quality greatly. Therefore, the delayed double-deck DCS may enable greater core space savings than current industry practice allows.

AB - An elevator group control system (EGCS) dispatches the elevators under its control to serve each passenger call. New passengers may appear while the elevators are serving their current assignments. This makes the control problem of the EGCS dynamic. The EGCS solves a snapshot instance of the elevator dispatching problem (EDP) anew in intervals less than a second. Its solution defines complete elevator routes by minimizing, e.g., passenger waiting or journey times, according to which the EGCS dispatches the elevators. In this dissertation, the EDP is considered as a bilevel, dynamic, stochastic, multi-objective and integer optimization problem. The bilevel approach is based on the distributed nature of a typical elevator system, in which each elevator has its own independent controller and the EGCS optimizes the service of the calls shared by the elevator group. Hence, the upper-level problem corresponding to the EGCS assigns calls to the elevators, and a lower-level problem for each elevator determines its route to serve the assigned calls. Near-future passenger arrivals are uncertain for the EDP. The number of passengers related to a call and completely new calls are modelled explicitly in the studied stochastic model. For the first time, the EDP takes into account the fact that passengers often travel in socially connected groups. Furthermore, the uncertainties are evaluated in multiple risk scenarios which converts the EDP into a robust optimization problem. Since the EDP is a difficult integer optimization problem, its solution time by an exact algorithm increases exponentially with respect to problem size. The main solution approach is based on a genetic algorithm suitable for real-time optimization in the EGCS. The EDP is also formulated as a mixed integer programming problem and solved by an exact algorithm, which are important for benchmarking and theoretical studies. The EGCS developed in this dissertation was installed as the first double-deck destination control system (DCS) in the world. However, the de facto standard DCS does not perform optimally mainly because of the immediate assignment policy, which does not allow the reassignments of the calls to other elevators. As one possible solution to this challenge, the delayed assignment policy is studied. It allows the reassignments until the last moment but requires new shared guidance devices. The first such system will be realized in the near future. According to the results, delayed assignments improve passenger service quality greatly. Therefore, the delayed double-deck DCS may enable greater core space savings than current industry practice allows.

KW - bilevel optimization

KW - stochastic optimal control

KW - robust optimization

KW - genetic algorithm

KW - simulation

KW - elevator dispatching

KW - double-deck elevator

KW - destination control

KW - kaksitasoinen optimointi

KW - stokastinen optimiohjaus

KW - robusti optimointi

KW - geneettinen algoritmi

KW - simulointi

KW - hissien kutsunjakelu

KW - kaksikorinen hissi

KW - kohdekutsuohjaus

KW - bilevel optimization

KW - stochastic optimal control

KW - robust optimization

KW - genetic algorithm

KW - simulation

KW - elevator dispatching

KW - double-deck elevator

KW - destination control

M3 - Doctoral Thesis

SN - 978-952-60-7540-2

T3 - Aalto University publication series DOCTORAL DISSERTATIONS

PB - Aalto University

ER -

ID: 17674415