You are on page 1of 18

A Survey of Elevator

Group Control Systems


for Vertical Transportation
A Look at recent literature

Joaquín R. Fernández
and Pablo Cortés

E
levators (also called “lifts” in some locales)
are installed in buildings to meet the verti-
cal transportation needs of the building oc-
cupants. The total vertical transport capacity
of a building is a crucial factor in its success
as a working, living, or service facility. Elevators must
be easily accessible, readily available, provide quality
service, and be reliable.
Since 1945, the number of buildings of medium to
high height has increased considerably (see Table 1),
providing a strong impetus to the elevator manufac-
turers, who have been receptive to the inclusion of
new technologies. First, automatic cabins were devel-
oped, followed by hydraulic cabins. More recently,
digital equipment has been included, and control sys-
tems and traffic management have been studied [2].
The total cost of an elevator, which includes the
direct installation costs, the maintenance service
costs, and the indirect costs resulting from the space
it occupies (in the layout of the building), is too high
to tolerate suboptimal elevator placement or ineffi-
cient operation. However, elevators must be designed
and positioned to provide the required service at a
minimum cost. These opposite viewpoints result in a
compromise between the quality of provided service
and economic restrictions. This tradeoff between

Digital Object Identifier 10.1109/MCS.2015.2427045


Date of publication: 17 July 2015

38  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015 1066-033X/15©2015ieee


The duty of the EGCS is to meet all
the demands by assigning an elevator to
each landing call (call made at a floor)
while optimizing several criteria.

economy and quality of service can be optimized in different ways; therefore,


elevator companies compete against each other by offering their own economic
and service models.
In 1979, the Otis Elevator Company developed the first elevator controller
based on electronic microprocessors, which greatly improved the flexibility of
the vertical transport systems that were already installed. By developing and
implementing new elevator group control systems (EGCSs), each with its own
dispatching algorithms, the competition between companies to offer the best
service increased. Such algorithms must address the different traffic patterns
that occur in a building, that is, up-peak traffic (the period in which people
move up from the ground floor or basements to the floors in the building),
down-peak traffic (the period in which people move down from different floors
to the ground floor and basements), midday or lunch-peak traffic (for a mix of
up peak and down peak at midday), and interfloor traffic for the remaining
periods. It is common to dispatch service calls by relying on artificial intelli-
gence algorithms, including genetic algorithms, tabu algorithms, fuzzy logic,
and neural networks, and some of the more relevant companies have developed
patents using such methodologies [3].
The high demand of building construction associated with the quantity and
increased size of buildings experienced in recent decades brought elevator
issues into the spotlight, but some argue that it could also have been viewed as
a harbinger of economic collapse [4]. The recent investment in the elevator sector
together with the lack of a state-of-the-art product was the main motivation for
writing this article. Surveys of the field of elevator control are rare, although
one exception is a study of the patent literature [3]. This article is broader in
scope, in that it reviews patents and the scientific literature together. In fact, the
main objective of this article is to serve as a reference for the sector’s framework,
summarize all the sector’s contributions and achievements, point out the main
background art—image licensed by graphic stock

deficiencies and drawbacks of previously conducted research, and suggest


some possible future research lines and developments.
The article contains several parts. After the introduction, the second part
defines an EGCS and the problems associated with allocating landing calls. The
third part describes the taxonomy of the different tasks and the functionalities of
image courtesy of pixabay.com

EGCSs. This is followed by a description of the literature related to capturing the


system data and the traffic-pattern recognition systems. The article continues
with an analysis of autocalibration, followed by a discussion of the main prob-
lems of car-landing-call allocation. The article concludes with a discussion of con-
structive solutions (such as double-decker systems) and the issue of optimizing
the quantity of the service.

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  39


must also address possible future calls. Thus, pure mathe-
Table 1 Cities with the highest numbers of skyscrapers matical approaches have a limited contribution. Moreover,
(buildings taller than 100 m) [1].
the time lapse for reaching dispatching problem solutions
must be very small. Thus, a compromise exists between the
City Country Number of Skyscrapers quality of the solution that a dispatching algorithm can
Hong Kong China 2354 offer and the time required to compute it. The combinato-
New York United States 794 rial nature of the problem, and its associated uncertainty,
Tokyo Japan 556 make it ideal for artificial intelligence algorithms, but a
final simulation analysis prior to real-life implementation
Shanghai China 430
is absolutely necessary [8].
Dubai United Arab 403
Emirates
A Classification of EGCS Tasks
Bangkok Thailand 355
and Functionalities
Chicago United States 341
The tasks that an EGCS performs can be divided into two
Guangzhou China 295 different categories: primary tasks (those that improve the
Seoul South Korea 282 quality or quantity of service) and secondary tasks (those
Kuala Lumpur Malaysia 244 that help improve the performance of the EGCS while exe-
cuting a primary task; see Figure 2). The challenges facing
an EGCS are multifaceted. The main aim of all EGCSs is to
Definition of an EGCS and solve the combinatorial problem associated with dispatch-
the Landing-Call Allocation Problem ing landing calls efficiently, which means that the quality of
An EGCS consists of hall call buttons that are situated on service provided by the vertical transport system must be
every floor, car call buttons inside of each cabin, and a
group controller. The EGCS is used to manage multiple
elevators in a building to efficiently transport passengers.
Assigning the landing calls is the primary task that an
EGCS must solve efficiently. This task is presented when a
passenger wants to travel from one floor to another and
presses the landing-call button on a certain floor. This pro-
cess is termed, in the vertical transportation domain, a
landing call. The duty of the EGCS is to meet all the demands Ascending Landing Call
by assigning an elevator to each landing call (call made at a
Car Call
floor) while optimizing several criteria. These criteria can
LC1
be service oriented, such as the average wait time (AWT) or Descending Landing Call
the percentage of calls that have a long wait time (also
LC2
called long-waiting call percentages, which are usually
defined as the percentages of calls that require more than a
given number of seconds), or oriented toward the prefer-
ences of the tenants, such as energy consumption prefer- LC3
ences [5], [6]. The EGCS performance is measured according
to the values of these parameters. A snapshot elevator dis-
patching problem is NP-hard [7]. EGCSs are systems that
0 1
can be analyzed from the perspective of several criteria, Elevators
and in general terms, they are designed to satisfy each cri-
terion to a certain level. However, the optimization criteria Figure 1  A snapshot example for the dispatching problem. A landing
are often conflicting (for example, quality of service versus call (LC) is made by a passenger outside the elevator shaft, who
energy savings) and are subject to physical limitations, presses a button to go up or down. The LCs are numbered sequentially
in time. Once inside the elevator, the passenger makes the car call by
such as the acceleration/deceleration of cars or the door pressing a button indicating to which floor they’d like to go. In this exam-
closing speed, associated with human comfort [2]. Figure 1 ple, the right-hand elevator is on the fourth floor and traveling down. It
depicts a typical vertical transportation system problem. will stop at the third floor and pick up descending LC2 before proceed-
A considerable amount of uncertainty exists in vertical ing down to the second floor to deliver at least two passengers (and
transport systems because the quantity of passengers perhaps additionally the passenger(s) currently waiting on floor three).
The left-hand elevator is currently on floor five and moving up to deliver
behind a landing call and their exact destinations are two passengers to floor eight. There are two ascending LCs (LC1 and
unknown until they press the car button inside of the cabin. LC3) that are waiting to be serviced, and the dispatcher currently
Apart from the complexity and data shortage, the system doesn’t have an available elevator to dispatch to them.

40  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


optimized (there are n k assignment possibilities in a build- algorithms can be used to perform the allocation, or the
ing with n elevators and k active landing calls). According same algorithm could employ different performance param-
to some articles (see [9] for a detailed analysis), transport eters or objective functions.
allocation can be improved by diminishing the large The optimization layer maximizes the service by opti-
amount of uncertainty that is present in the system by mizing several relevant criteria. These criteria can vary de-
ensuring optimal data capture and detecting traffic pat- pending on whether the optimization is made according to
terns. Other types of problems arise when dealing with the the service point of view or the tenant point of view. From
dispatching of double-decker elevators due to the specific- a purely service perspective, the AWT, which is the average
ity of construction (two joined cars with one of them over time a passenger waits until the car arrives, is usually opti-
the other). This topic is detailed next in the section “Archi- mized. The long waiting-time percentage of calls (LWT%)
tectural Design Variations for Cars and Shafts.” Another is another measure of pure service quality and acts as a
class of problems relates to the quantity of service instead of measure of the service dispersion. The LWT% usually rep-
the quality of service. These problems have been approached resents the percentage of calls that have a waiting time of
through the use of artificial intelligence to optimize zone more than 60 s. Other criteria related to the quality of ser-
boundaries in configurations that implement the dynamic vice include the average journey time (AJT, the total amount
sectorization of floors during up-peak traffic periods. This of time that a passenger spends inside of the cabin) and the
is discussed more in the section “Approaches for Solving average total trip time (ATT, the total amount of time be-
the Car-Landing-Call Allocation” but in general terms tween initially pressing the button and the arrival of the
relates to the definition of a sector containing several adja- car at its destination). Because psychology research shows
cent floors that are (usually) attended by a dedicated eleva- that passengers do not give much importance to the time
tor. The boundaries of each zone are dynamic and can spent inside of the cabin but become annoyed if the eleva-
change with the demand. tor takes too long to arrive [2], the AWT is the most fre-
Whereas Figure 2 represents the tasks that a global EGCS quently used objective function from a service point of
performs and the relationships between these tasks (note view. In contrast, the tenant preferences include optimiz-
that not all EGCSs perform every function), Figure 3 depicts ing power consumption and the ATT (therefore workers
the different modular blocks that can be used in an EGCS lose less time) or maximizing the handling capacity (the
and their abilities to meet the functionalities defined in maximum transport capacity of the system).
Figure 2. The data-capture layer at the bottom is essential The optimization criteria also typically depend on the traf-
because the EGCSs must be checked frequently to determine fic patterns. Usually, the waiting times are preferentially opti-
the state of the vertical transport system. This data-capture mized relative to the other criteria, such as wasting energy
layer can be divided into two different sublayers, namely, the during heavy demand patterns (up peaks or down peaks) [10].
software sublayer and the hardware sublayer. The hardware Since the systems are designed to meet up-peak demand, most
sublayer comprises every physical device that belongs to the vertical transport systems are oversized for soft demands,
system. The software sublayer
executes algorithms (from alge-
braic to artificially intelligent)
to transform the sensor infor- Quality of Service Quantity of Service
mation into manageable data.
In this sense, the most complex Landing Call Optimization
task is to infer the traffic pat- Allocation of Dynamic
tern. Traffic patterns can be Primary Single-Decker Sectorization
divided into four different Task Systems
Boundaries
principal types depending on Double-Decker Optimization
the demand. These data are Systems
interpreted by the optimiza-
tion layer in several ways. The
most relevant data depend on Calibration Uncertainty
whether the traffic pattern is up
Secondary
peak (significantly ascending Task Data Capture
Autocalibration
flow from the main floor) or
Traffic-Pattern
whether the optimization re-
Detection
volves around maximizing the
quantity or quality of services.
However, depending on the
type of traffic pattern, different Figure 2  The taxonomy of the different tasks performed by an elevator group control system.

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  41


which include interfloor traffic. Con-
sequently, time optimization is often
switched to energy optimization
during off-peak periods because the

Program-
waiting time requirements are easily

Linear

ming
Methodology achieved as a collateral effect, even
when they are not properly opti-

Decision
mized. Although power consump-

Tree
tion is considered to be crucial in a
sustainable development, few ap-

Pattern Detection

Camera Scale
proaches currently include power re-
Intelligence
Artificial

Radar
duction [11]. Technology has evolved,
and new hoist elevator models allow
a portion of the energy that the eleva-
tor employed while moving up to be
regained. However, only a few stud-

Number of Passengers
Data Capture

ies [11]–[13] have incorporated this

Calculation of

Deck Position
regain of energy in their research.

on a Deck
Preferential

Hardware

Sensor
Software
Others have attempted to reduce
Service

energy consumption by modifying


the speed of the elevators [14].
Quality of Service Criteria

Because the dispatching problem


AWT LWT% AJT Energy

is NP-hard, the main approaches for


solving it are related to decision
Estimation on Number
of Passengers Behind

trees, linear programming, and arti-


a Landing Call

ficial intelligence. The scale of the


Button Panel
Optimization

problem and the need for a near-in-


Calibration

stantaneous solution make the dis-


patching problem difficult to solve
in a real application. One approach is
to have the EGCS periodically cali-
Figure 3  The relation between the subsystems of an elevator group control systems.

brate the dispatching algorithm to


Quantity of Service Criteria

optimize its performance. This opti-


RTT

mization can be achieved in real


Only Front/Front
and Rear Doors

time, in parallel, or through succes-


sive simulations conducted during
the night by employing data from a
Configuration
Setup

database. The next sections present


HC

the functionalities of each EGCS, as


described in Figure 2.
Double/Single
Decker

Approaches for Improving


Data Capture
Data capture and management are
directly related to, and support, traf-
Mechanical Orders

Control

fic-pattern recognition and landing-


Door

call allocation. Observers collected
data from the beginning of the last
century until the 1980s in manual sur-
veys [15]. By the end of the 1980s, the
Movement
Elevator

first generation of EGCSs controlled


by microprocessors was created,
which allowed for measurements and
the calculation of statistics relating to
traffic. In addition, technology was

42  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


developed that allowed for real-time estimates of the number of
passengers inside the cabin. The number of passenger trans- POP Up Traffic
Up Peak
ferred in a stop can be calculated from the initial load, the min- 15
imum load, and the final load by standard that the weight for 10 Morning Lunch (In)
all the people and traffic is characterized by daily and weekly 5 Interfloor
tables accessible to the dispatcher [16]. The installation of detec-
0
tors in the main floors of the building is suggested in [17]. Thus, Time
the number of people queuing for the elevator is known. In [18], 0
research was conducted to determine if the use of cameras with 5 Afternoon
Interfloor
special sensitivity could recognize the number of passengers 10 Lunch (Out)
inside the cabin. However, in [19], a method was suggested for 15
estimating the number of people on every floor by employing Down Traffic Down Peak
20
scales on every elevator placed under their floors.
Modern technology allows for every type of data to be Figure 4  Traffic patterns occurring throughout the day in a typi-
obtained and employs radar, digital weights, modern cam- cal office building. The figure shows population movement (POP)
eras, or infrared beams [20]. A video-aided EGCS has been versus time.
suggested that is capable of detecting passengers, estimat-
ing their movements, and controlling the elevator doors other performance measurements depend on the identifi-
according to their paths [21]. In addition, the development of cation of the corresponding peak period (see [2] and [15,
a monitoring system that tracks the passenger demand Ch. 3, 4]). For example, if up-peak traffic situations are not
independently of the main EGCS has been undertaken [22]. identified quickly, long queues can build up on the main
This system spots data at crucial points, such as inside of the entrance floor of the building (typically the ground floor),
elevator or at the waiting areas, and transmits all the gath- and passenger waiting times will become longer. Long
ered data to a monitoring system. Furthermore, it has been waiting times cause dissatisfaction with the operation of
suggested [23] that sensors should be used rather than cam- the elevators. However, the up-peak mode should not be
eras to detect and digitally create three-dimensional objects. activated unnecessarily because a rule consisting of the
direct return of lifts to the entrance floors would then be
Approaches to Traffic Pattern Recognition activated, resulting in long waiting times for passengers
Classical theory describes four traffic patterns during a who are at other floors.
typical day in buildings where people work. These patterns Traffic-pattern detection is one of the most relevant but
depend on whether the main flow significantly ascends least-studied features of vertical transportation [25]. The
from the main floor (up peak), descends toward the main evolution of the field of vertical transportation has improved
floor (down peak), both (lunch peak), or none (interfloor). from its primitive form, which consisted exclusively of up-
Figure 4 provides a representation of the typical move- peak traffic-pattern detection (based on the percentage of
ments in office buildings (vertical axes represent the popu- cabin-load measured over specific time intervals), to the use
lation movement and the horizontal axis represents the of modern complex heuristic algorithms that can identify
hourly period). several traffic patterns [26]. In [27], a historical database was
In a typical office or work building, the up peak corre- employed to diminish uncertainty. Depending on the fre-
sponds to the beginning of the working day, the down quency and tendency of the events, they were recorded as
peak corresponds to the end of the working day, the lunch weekly, daily, or real-time events. The degree of certainty of
peak corresponds to the break in the middle of the work- the data was evaluated using two threshold considerations.
ing day, and the interfloor pattern characterizes the soft The model was slightly improved in [28] but is still similar to
and uniform demand during the remainder of the day. the original [27]. However, in [19], a floor population method
Continuous variations in the passenger flow and cabin was added to the EGCS. This method consisted of counting
load influence the path that the elevators choose [24]. Sev- the passengers entering and exiting the elevators on every
eral reasons justify performing a deep analysis of the traf- floor during the up-peak period. In this case, the total popu-
fic in the building before programming a dispatching lations were inferred for each floor.
algorithm. First, most of the controller algorithms are pre- A system for determining the beginning and end of the
configured and use performance rules that are applied up-peak and down-peak periods has been suggested [29].
depending on the type of traffic in the building at that In this system, the number of boardings are counted and
time. The EGCS must therefore include some form of traf- compared with a threshold to detect the pattern. If the
fic-type detector. Second, the correct determination of the number of boardings is greater than the threshold, a pat-
period of the traffic pattern being experienced by the tern is identified. The threshold value may be a percentage
building is a key factor because most of the calculations to of the building’s population. In addition, the system pro-
estimate the passengers’ AWT, the round-trip time, and vides a mechanism for adjusting this threshold value based

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  43


An EGCS consists of hall call buttons that are situated on every floor,
car call buttons inside of each cabin, and a group controller.

on the current traffic. Once an up-peak or down-peak ing weight between periods, and the traffic pattern detected
period is detected, the system checks the load of the decks in the previous interval. Finally, the anticipation of demand
for a predefined time interval. If the loads are greater than by incorporating historical data tables into the dispatching
a determined load value, then the up-peak or down-peak algorithm has been attempted [38].
condition was detected. Some basic approaches are used to
address this problem by using models based on basic learn- Approaches to Parameter Calibration
ing [30], which uses a historical and recent database to One of the main difficulties when designing an EGCS is the
determine if the traffic pattern that occurs is up peak, down need to a priori set the values of a wide variety of parame-
peak, or neither. ters that affect system performance [39]. These parameters
According to statistical learning theory, demand can be range from the specific settings of an optimization algo-
predicted by using a support-vector machine with a qua- rithm, to weights in heuristic dispatcher or PI controllers,
dratic loss function, which is called a least-squares support- or even to the parameters defining the linguistic regions of
vector machine [31]. In [32], a system was patented to detect fuzzy approaches. A discussion of the approaches to over-
up-peak patterns by using statistical data obtained from come this difficulty is given next.
the actual incoming demand. This problem requires expert knowledge, but if the
Fuzzy logic is especially suitable for detecting traffic system is not able to adapt correctly to the traffic pattern
patterns [33] due to its own definition and the nature of the changes, then the system will not maximize its perfor-
detection problem. Fuzzy-logic rules have been suggested mance. A control system in which the parameter values are
to classify traffic patterns, but relevant input data, such as calibrated during the night while accounting for the data
the time of day or the passenger rate of arrival or depar- gathered during the day has been suggested [40]. This
ture, are not considered [34]. Another model that is able to system conducts a calibration by using a genetic algorithm
detect traffic and allocate landing calls using a set of fuzzy that aims to provide robustness to the system in which the
rules has also been suggested. However, this suggested performance does not change between the optimization
model lacks any control strategy based on typical criteria, conditions and the actual work conditions. This change
such as the AWT, percentage of long waiting calls, or power between optimization conditions and actual work condi-
consumption [35]. The proposed control system can be tions may result from changes in the objective function
divided into three different modules, one for allocating (tenant preferences) or from different boundary conditions.
calls, one for generating the control strategy, and one for These changes are addressed by using a solution bank
traffic management, which gathers and memorizes the pas- memory or excluding some quality solutions that are found
senger flow with time [36]. This module can recognize dif- for different traffic patterns to obtain fast convergence,
ferent patterns by measuring a set of five features from respectively. However, the latest trend in calibration sys-
three different perspectives that constitute the system tems consists of solving the problem without inferring any
inputs. Based on the traffic, the features of up traffic and data from statistical memory [41]–[45].
down traffic can be distinguished. From the traffic percent- An EGCS that considers two different criteria simulta-
ages, the system can examine the centralized in-traffic per- neously along with AWT and power consumption and
centages and the distributed out-traffic percentages. optimizes them simultaneously by employing an evolu-
Finally, the time can be measured throughout the day. tionary standardized-objective weighted aggregation
From these five features, the system is capable of deducing method is suggested in [41]. Because it is mathematically
eight different traffic patterns. impossible to minimize the criteria simultaneously, the
Another possibility for detecting traffic patterns is to system balances the importance of each one by using a set
combine the fuzzy inference with statistics that were calcu- of weights that depend on certain features. This balance is
lated from a data-gathering system [37]. In this case, the achieved by using a PI controller that calibrates the set of
system can adapt smoothly to different situations. In [9], weights after each iteration and depends on the deviation
the detection of the traffic pattern using a fuzzy-logic of the anticipated AWT deviation. When the demand is
model solves the lack of memory in fuzzy-logic designs by heavy, the system attempts to minimize the AWT. How-
considering as the input values the total ascending and ever, when the demand is low, the system attempts to mini-
descending weights that are carried by the cabins during a mize power consumption. A similar EGCS has been
time interval, the variation in the ascending and descend- suggested [43], but the latter’s short evaluation function is

44  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


modeled to represent how the level of uncertainty increases estimated waiting times, namely, the waiting time esti-
further into the future. mated from the future status of the elevator and the wait-
In addition, an automatic calibration system that consid- ing time estimated from the elevator’s landing pattern. The
ers the current feedback rather than the historical or statis- allocation is made based on this adjusted waiting time.
tical data has been suggested [44]. The landing allocation is Another study [49] reduces the magnitude of the problem
conducted by employing a fuzzy engine. Every 5 min, the by defining a time window and only optimizing the prob-
calibration module compares the observed AWT with the lem inside that time window based on figures of merit,
predicted AWT and acts over the fuzzy inference module such as the passenger arrival time, the arrival floor, and the
by changing the set of logic rules and the membership departure floor. In this case, the problem is to minimize the
function that defines the fuzzy variables (whether it is con- total service time for all passengers.
sidered necessary or not). Another automatic EGCS calibra- An additional study [7] demonstrated that high-quality
tion [42] suggests that the landing allocation is based on a solutions can be obtained by employing decision trees. The
fuzzy neural network that is calibrated through a combina- decision tree is a decision-support tool that uses a tree-like
tion of an ant-colony algorithm and a back-propagation graph or model of decisions and their possible conse-
algorithm. First, the ant algorithm provides a fast global quences, including chance event outcomes and expected
solution that considers the weights of the fuzzy neural-net- waiting times. The accumulation of these alternatives
work structure as inputs and uses a back-propagation algo- results in the decision tree. The heuristic aims to limit the
rithm to perform an exhaustive local search to obtain a search in an intelligent way and controls the computation
global solution in the solutions space. time. The main problem in this type of search is related to
the limited or cycle search times. The search times must be
Approaches for Solving sufficiently large to allow for a certain depth in the search
the Car-Landing-call Allocation and must be sufficiently small to provide a definitive solu-
As previously mentioned, the allocation problem is NP- tion for the control system within a reasonable period. The
hard. In a building with n elevators and where k landing computational problem lies in the fact that the number of
calls occur, n k possible allocation solutions exist, and the nodes and edges increases exponentially as a function of
computational time that is required for solving the problem the number of pending landing calls, making a complete
increases rapidly with the problem size. Consequently, search impossible within the time required for a real-life
most of the suggested solutions rely on artificial-intelli- implementation. This problem results in two different algo-
gence heuristics. The following sections present approaches rithm compromises. In this case, an allocation algorithm is
for tackling the problem. suggested based on a search tree, where a node is gener-
ated for every possible combination of a landing call and an
Approaches Based on Heuristics elevator in each level of the tree. Thus, the landing call is
A multiobjective function that considers a few criteria, assigned to the elevator that results in the lowest estimated
namely, waiting time, the equal time-interval operation cost. These estimated costs can be obtained by looking at
preferential zone, riding time, the cabin-load factor, the cost tables, especially those made for the building.
first-arriving-cage preferential zone, and the stop call was
investigated in [46]. These criteria are combined by using Approaches Based on Fuzzy Logic
different weights to calculate the final fitness of each eleva- Fuzzy logic is one of the most common methods for
tor, and the elevator with the best fitness is selected to ser- addressing elevator allocation. The main advantage of
vice the landing call. fuzzy logic is its fast system response, which is crucial for
An EGCS that employs sectoring under down-peak traf- vertical transport. The main disadvantage of most fuzzy-
fic conditions has been suggested [19]. This EGCS defines logic implementations in the literature is that such
one sector per elevator, and all sectors are formed by using approaches follow a nondynamic dispatch (although a
the same number of floors. The EGCS prioritizes down- recent contribution considering a dynamic approach has
ward landing calls in the event of a conflict with upward just been presented in [50]). Thus, the assignments cannot
landing calls. An EGCS that combines decision rules with a be undone once they are allocated and remain fixed until
cost function that estimates the suitability of the set of rules they are completed. Consequently, every landing call is
has been suggested [47]. Each assignment option contains a only evaluated once before an elevator is assigned to it. In
landing-call item and an elevator item that is used to calcu- addition, although another landing call may appear in the
late the cost function and evaluate the allocation. Some of vertical transport system that may make that assignment
the items are changed, and new cost functions are calcu- suboptimal, it cannot be undone. The suitability of each
lated iteratively. The best possible assignments are selected cabin is evaluated in [51] by using different sets of fuzzy-
according to the cost function values. One study [48] defines logic rules. Both the waiting time and riding time are con-
a figure of merit called the adjusted waiting time for every sidered as optimization criteria [52], and a hybrid model is
elevator, which is a weighted combination of two different suggested that combines fuzzy rules with neural networks

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  45


[53]. Furthermore, fuzzy logic is combined with a “reason- The two different nonmutually exclusive preferences
ing unit” that selects the appropriate set of fuzzy rules for include the load preference and the service preference. The
each elevator depending on the miss forecast rate and the main advantage of the suggested EGCS is its ability to
estimation rate of the elevators [54]. Two studies, [34] and improve the quality of service that is offered to certain
[55], optimize the assignations according to three different floors without excessively reducing the service offered to
criteria, namely, the AWT, the long waiting times percent- the remaining floors. However, the main disadvantage of
ages, and power consumption. The objective function this system is its inability to handle a large number of vari-
defined for each cabin combines every criterion by using a ables, which can overload in the system in terms of compu-
set of weights. In this case, the assignment is made from tation. Consequently, the use of a memory bank of the best
two different system modules. The allocation module solutions has been suggested [66] to improve the efficiency
assigns the landing call using a set of fuzzy rules based on of the system by reducing the computational time and
the tenants’ preferences for every traffic pattern (each traf- securing a high-quality solution. The patent [67] claims an
fic pattern has its own membership function). In addition, algorithm capable of decreasing the expected AWT by eval-
the strategy assignment module is able to shift two stan- uating only the two cars with the highest value of a second-
dard membership functions (large and small) for each ary in-house criterion, allowing a preselection of cars for
input variable according to the relevance degrees defined faster convergence.
by the tenant for each variable and each traffic pattern. The Zoning is a specific approach to deal with special traffic
landing call can be assigned to the elevator that has the best situations. It consists of separating the building in zones,
fuzzy-logic fit, but the choice can be delayed until the associating groups of elevators to the different zones, and
degree of uncertainty of the allocation (also measured moving and stopping the cars of the associated group
through a set of fuzzy rules) is below a specified level [56]. inside areas that are formed by adjacent floors. Although
The performance criteria that measure the “utility” of this approach is explained in detail later, it is appropriate to
assigning an elevator to a landing call through a set of mention the use of genetic algorithms to solve the zoning
fuzzy rules can be defined [57]. In one study [58], a fuzzy- problem [30], [40], [68], [69]. The approach is typically the
logic EGCS is designed that iteratively applies groups to the same: the floors of an area are divided into two groups, one
elevator in a predefined sequential order according to for passenger-specific destination calls and the other for
fuzzy-logic rules, while the difference between the best floor-specific up/down calls. The genome of individual
solution and the second best is not especially relevant (that elevators is specifically encoded to reflect this division. In
is, the first and second best solutions are quite close). another line of research, a genetic algorithm that is focused
Numerical panels can be placed on every floor to minimize on optimizing the average riding time rather than the AWT
the uncertainty [59]. In addition, the landing call is assigned has been suggested in [70] and provides good performance
to the most suitable elevator according to the fuzzy mea- for the analyzed buildings.
surements of different criteria, including the car calls of the Finally, researchers from the KONE Corporation [71]
elevator, the intended destination of the passenger, and the combined a genetic algorithm with a statistical traffic esti-
source landing requested by the passenger. Fuzzy-logic- mator. The different possible routes are generated and eval-
based EGCSs have been employed that aim to keep the uated in a typical manner by using a genetic algorithm.
waiting time at a reasonable level while giving service pref- However, depending on the traffic conditions, the travel-
erences to certain floors [59]. These EGCSs use AWT, power ling time is optimized if the demand is heavy or the pas-
consumption, and floor traffic as criteria. senger waiting time is optimized if the demand is low.

Approaches Based on Genetic Algorithms Approaches Based on Neural Networks


Genetic algorithms have also been widely used to solve and Hybrid Neural Networks
allocation problems. According to several researchers [61]– Several authors [72]–[74] have suggested the allocation of
[64], considerably higher computational times are required, landing calls using common neural-network approaches.
and it is possible that the searching process becomes However, the approach addressed in [75] is of particular
blocked in a specific zone of the solution space due to a interest because it estimates the remaining response time for
local optimum. Most EGCSs do not offer a particular ser- an elevator car to be assigned to a new landing call. Further-
vice for each floor. However, a genetic algorithm is pre- more, this neural network is standardized for every use in
sented in this way [65] for allocating the landing calls while any building by using an upstream fixed-length stop
setting preferences for each floor. The EGCS uses an objec- description that codifies the building status every time a
tive function composed of two terms. The first one evalu- new landing call is detected. This work is extended in [76],
ates the suitability of a cabin for attending a specific landing where a general method is suggested for any neural network
call, expressing the preference for such floor, while the based on an EGCS that trains the neural network while the
second one is used as a flag that determines if the cabin is elevators are performing the service. Therefore, the remain-
attending a preferential landing call in at a certain time. ing response time can be estimated for the elevator to

46  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


accomplish a landing call. The main advantage of neural
networks is their ability to adapt to changes in traffic flow, (6)
but the main disadvantage is the complex design, which
implies a higher implementation cost. Moreover, their ability Forwarding State
to adapt to sudden and abrupt changes in traffic flow is (1) (2)
(5) (7)
somewhat limited. To overcome this disadvantage, an EGCS (3)
based on a fuzzy neural network was suggested by [77] and
[78], which used fuzzy-logic designs to adapt to sudden traf- Waiting State Transport State
fic flow changes and the neural networks to adapt to long-
term traffic flows. In addition, a fuzzy neural network is
(4)
suggested that uses a fuzzy evaluation of the situation when
a new landing call appears and temporally assigns the land-
Figure 5  A diagram showing the possible states of an elevator and
ing call to an elevator [79]. Next, the neural network receives the various possible transitions.
the values given by the fuzzy process as input values and
checks the suitability of the election. Genetic network pro- Through communication between cabins, the elevators
gramming is a type of evolutionary algorithm that is similar that are in a waiting state may avoid competition with other
to the genetic algorithm, in which each individual represents elevators attending to the landing call, and the elevator in a
a graph that contains entwined nodes. Then such nodes pro- waiting state always has higher priority for attending land-
cess all the information regarding the edges between the ing calls than elevators in a transport state unless a trans-
individuals. Unlike the conventional genetic algorithm, the port-state elevator has a car call for the same floor as the
global searches in genetic network programming are not landing call. In addition, the elevator that is considered as
locked to a local optimum. However, a global search has the attending to the landing has priority over all other elevators
disadvantage of slow convergence and high computational in the forwarding and waiting states, if their distances to
cost. To overcome this drawback, a solution to the conver- the landing call are greater than that of the elevator that has
gence and computational problem that is associated with the the priority landing call. The objective of the learning pro-
genetic neural network is suggested by combining them cess is to obtain a compromise between the total AWT that
with an ant-colony algorithm, which provides the genetic is estimated at the beginning of the process to the current
network programming with the necessary feedback to find moment and the total average number of passengers served
the global optimal quickly [80]. in each elevator trip. Each time an elevator accomplishes a
service, the weight of the rule that was employed in the as-
Approaches Based on Multiagent Systems signation is actualized according to a formula. A multia-
Multiagent systems appear in contrast to centralized con- gent-based EGCS has been suggested where every elevator
trol systems and constitute a decentralized form of control has its own control system and competition is avoided
based on agents that communicate or act independently. A through communication [82]. Each local elevator control
multiagent control system is suggested where the elevators system is built using field programmable gate-array logic
are the independent agents that communicate with each panels that can implement four different architectures de-
other to avoid competition to attend the landing calls [81]. pending on the traffic flow. The main EGCS is only used to
The main advantage of a decentralized elevator control decide which one of the four possible configurations must
system is its robustness in the face of
breakdowns and its ability to adapt to
Table 2 Examples of in situ rules that were built for the button panels [81].
traffic fluctuations. The landing call allo-
cation consists of two different processes,
a learning process and a set of rules de- Time Elapsed After Number of Passengers
Rule Number Hall Call Button Pressed During Last 5 Min
fined in situ.
1 UP_X1 = ON 0 # UP_T1 1 10 UP_5R1 1 50
As shown in Figure 5, each cabin has
three different states, namely, the waiting 2 UP_X1 = ON 0 # UP_T1 1 10 UP_5R1 $ 50
(empty or full but without moving), for- … … … …
warding (empty and moving to a landing 7 UP_X1 = OFF UP_T1 = 0 UP_5R1 $ 50
call), and transporting (with passengers 8 UP_X 2 = ON 0 # UP_T2 1 10 UP_5R1 1 50
inside and moving to a destination) states. … … … …
Transitions 2, 4, and 7 are managed ac-
7 ^N - 1h UP_XN-1 = OFF UP_TN-1 = 0 UP_5RN-1 $ 50
cording to the in situ rules for every but-
ton situated on every floor; see Table 2. 7 ^N - 1 h + 1 DOWN_X 2 = ON 0 # DOWN_T2 1 10 DOWN_5R2 1 50
The active landing calls are ordered ac- … … … …
cording to their weights.

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  47


be executed according to fuzzy logic. This method signifi- resolves the problem in two steps. First, it formulates the
cantly reduces the amount of traffic data. Additionally, an- problem as a continuous Markov decision problem. Next, it
other multiagent-based EGCS has been suggested that uses ascertains the discrete Markov equivalent problem that is
a distributed system and overcomes computing bottlenecks solved by dynamic programming.
by defining agents that represents the single elevator (coor- Particle swarm algorithms may produce better results than
dinated agents), the proper EGCS (high-level coordinator genetic algorithms or tabu search engines [85]. In particle
agent), and the virtual management unit (management swarm algorithms, each solution is encoded as a particle that
agent) [83]. is placed in the space of possible solutions and has its own
associated speed that defines its movement through the solu-
Other Computational Approaches tion space. The algorithm itself consists of a loop that modifies
Several other computational approaches have been tested the speed and trajectory of these particles during several itera-
for controlling EGCSs. For example, a control system was tions to adjust them according to the distance between the
developed exclusively for the up-peak period based on a best positions reached by the particles and the best positions
dynamic programming model that determined when the reached by the entire swarm. Thus, the speed changes accord-
cabins on the main floor must depart [84]. This model ing to the quality of the solutions area that the particle and the
swarm are visiting at a particular moment.
Petri nets have also been suggested to deal with the
problem (see [86] for a vertical transportation system mod-
eled as a petri net). A petri net associated with a vertical
Technical
transport system implies a delimited system that can be
Services solved using a Karp–Miller tree [87]. However, such
approaches have not gained too much enthusiasm due to
the existence of convergence problems in the required
time for real-life implementation. Tabu search has not been
extensively used in EGCS optimization, and the first EGCS
based on a tabu search is provided in [88]. In [89], it is
shown that the elevator-dispatching problem can be solved
by employing a viral algorithm and can outperform the
marks of other approaches, such as genetic algorithms or
Skylobby
tabu searches.
110 Stories = 1368 ft (North), 1362 ft (South)

Technical
Services
Architectural Design Variations
for Cars and Shafts
This section reviews several design alternatives that are
used to increase the handling capacity and improve the
efficiency in the dispatching process. For this purpose,
these designs place more than one car in the same shaft,
Local either attached to each other or free to travel separately.
Elevators
This section first presents double-decker systems and then
Skylobby analyzes other multicar designs.
Technical
Services Double-Decker Systems
A double-decker system consists of two cabins that are
joined and travel together along the same shaft. In fact, a
Express Express double-decker is an elevator with two cars attached
Elevators Elevators
together, one on top of the other. This system allows pas-
sengers on two consecutive floors to use the elevator
simultaneously, significantly increasing the passenger
capacity of the elevator shaft. The main advantage of a
double-decker system is the elevator footprint reduction
Six Stories

Technical in the core of the building (between 25 and 40% [90]).


Services Local
Elevators From a service perspective, the underuse of one deck at
Plaza Level the expense of the other must be avoided, which involves
the promotion of the elevators for attending adjacent land-
Figure 6  An example of sectorization in a building. ing calls so that the number of stops is minimized. In the

48  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


early 1960s, the control system for double-decker elevators inside a chromosome that represents a global assignation.
was basic. Otis developed the trailing-deck control system Although [99] states that the collective methods for a sin-
that consists of dividing the elevator into the leading deck gle-deck elevator present approximately the same perfor-
(first deck in the direction of movement) and trailing mance as the noncollective methods, [100] considers that
deck. The dispatching method only used the collective collective dispatching is no more suitable for this type of
control principle for dispatching the trailing deck. (The system because it involves long waiting times, especially
collective control principle is a well-known and basic ele- on the lower floors and for a significant number of eleva-
vator control principle that states that every car collects tors. In addition, this method results in bunching of the
calls in its trip direction. That is, an ascending elevator, elevators during the down-peak traffic periods. The dis-
for example, will collect passengers wanting to go up if patching algorithm evaluates the assignment of attending
they are between the elevator’s current position and it to a landing call to the arrival time of the elevator and the
ultimate destination. The collective control principle is time that a passenger employs inside the system. The prob-
addressed in most elevator handbooks, such as [2] or [15].) lem itself is set out as a theory of order problem, where the
Thus, the leading deck only attends to the landing calls route that the deck follows can be represented as a pick and
adjacent to those already assigned to the trailing deck, delivery problem graph, which is built at the same time
which leads to large differences between the deck loads that the deck creates its itinerary. Different graphs are built
and a significant reduction in the handling capacity of the depending on the travel direction; hence, two different
system (throughout the interfloor periods). graphs are required for every elevator and for each time
The authors of [91] and [92] suggest a double-decker interval equal to the time required for a round trip. Each
EGCS system using neural networks, and the authors of deck is considered as an independent elevator when com-
[93] suggest an EGCS that is especially designed for double- peting for the landing calls. The computational time in the
decker systems that can assign landing calls for both decks genetic algorithms constitutes a disadvantage, but in this
and also possesses a proper assignment control device that algorithm, the time can be drastically reduced by up to 65%
decides whether the landing calls that are already assigned when using a gene bank.
to both decks can be served simultaneously. Furthermore, In a different study, a dispatching algorithm was sug-
[94] considers the path of the elevators that are already gested for a double-decker system based on genetic net-
assigned to the landing calls (from the starting floor to the work programming and was optimized by an ant-colony
final floor) [95] and suggests an EGCS that minimizes the system that provided fast convergence [101]. Genetic net-
average ride time rather than the AWT. In [96] estimated work programming with interconnected process and deci-
waiting time and estimate ride time are minimized. Two sion nodes, according to the search needs, provides
other studies ([97] and [98]) implement a genetic algorithm excellent adaptation to the demand fluctuation. The system
for dispatching double-decker systems. In this case, each names the deck that was considered by the algorithm for
possible allocation was characterized as a typical gene optimizing the landing-call assignment as “self-cage” and

Table 3 Classification according to the objectives.

Objective
Data Traffic-Pattern Landing-Call Parameter Double-Decker Multicar Optimization
Reference Gathering Detection Allocation Calibration System System Dynamic Set
[17]–[19]
[26], [37]
[35], [36], [74]
[34]
[13]
[40]–[42], [44]
[39]
[7], [50]–[52], [61]–[65], [69], [70],
[72], [73], [77], [78], [80]–[82],
[84], [86], [88]
[91], [92], [96], [100], [101]
[103]–[112]
[114]–[117], [119], [120]

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  49


Table 4 Classification according to the criteria employed.

Criteria

Maximum WT
+ Number Average Load
Uniform Distance of Stops + Number of Preference
% of Round- + Number Number of Passengers + Service Handling
Reference AWT LWT Energy ATT AJT Trip Time of Stops RGP Loops per Trip Preference Capacity Interval
[35], [40], [44], [50],
[51], [61]–[64], [69],
[72]–[74], [77], [78],
[80], [82], [86], [91],
[92], [106], [107]
[36]
[34], [52], [108]
[13]
[41]
[42]
[70], [88], [96]
[81]
[7]
[65]
[84]
[100]
[101]
[103], [109], [110]
[104]
[105]
[111]
[112]
[114]
[116]
[115], [117], [119], [120]
AWT: average waiting time
%LWT: percentage of long waiting calls
ATT: average trip time
AJT: average journey time
Max WT: maximum waiting time
RGP: genetic network programming
Interval: average time for having an idle elevator ready to serve at the main floor.

names the adjacent deck that was not directly considered Multicar Elevator Systems
by the algorithm for optimizing the landing-call assign- Multicar elevators (MCEs) consist of several cars that are
ment as “other-cage.” The system handles 12 different items placed in a single elevator shaft. This architectural variation
or features that represent the state of the system. has recently attracted significant attention among high-per-
The authors of [102] suggest another genetic network formance transportation systems. However, the accumulat-
programming approach. First, this algorithm optimizes the ed knowledge regarding single-car elevators is not readily
genetic network programming for solving the double- applicable to MCEs due to their specific operation and differ-
decker system problem. Next, its performance is compared ent performance. An MCE system’s main problem is avoid-
with the performance obtained by using the conventional ing conflict between the two cars. Therefore, an MCE group
methods for ratification. control method that can efficiently allocate service calls

50  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


Table 5 Classification according to the methodology used.

Methodology
Particle
Historic Fuzzy Neural Genetic Ant Colony Tabu Multi- Search Petri Dynamic Immune Swarm Cellular
Reference Memory Logic Net Algorithm Optimization Search agent Tree Net Programming Algorithm Algorithm Automata
[26]
[37]
[74]
[13],
[34]–[36],
[40], [44],
[50]–[52]
[41],
[61]–[65],
[69], [70],
[80], [100],
[111], [112]
[42]
[77], [78]
[88]
[81], [82]
[7]
[72], [73],
[92], [109]
[86]
[84]
[101]
[115]
[116], [117]
[85], [119],
[120]

while preventing interference among the cars is required. In However, the main objective is usually to reduce the average
[103], an MCE group control algorithm is proposed that im- total waiting time, as is the case of [107], which proposed a
plements an interference risk evaluation, a schedule-comple- method that minimizes the waiting times based on an algo-
tion-time minimization function, and idle-car parking rithm that allocates requests to elevator cars by using a local
strategies. The algorithm achieves an efficient allocation of search method to reduce the wasted time of idle cars. Other
calls and prevents interference without restricting the opera- techniques, such as [108], consist of defining dynamic zones
tion of elevator cars. The simulation results showed good in which the elevators can operate without conflict. More
performance under low and medium passenger arrival complex approaches use other artificial intelligence tech-
rates. The study in [104] also proposes a method to avoid col- niques such as reinforcement learning in [109], where a con-
lisions between cars in an MCE system, but in this method, trol agent selects the best of four strategies, namely:
collision avoidance is treated as a dynamic optimization 1) transportation strategy, 2) passenger strategy, 3) zone strat-
problem (to determine which floor each car should visit next egy, and 4) difference strategy, according to the traffic pat-
under the constraint that collisions never occur). In this ap- tern. Similar approaches, such as [110], use an average
proach, the group controller gives basic car travel schedules. reward learning method to reduce service completion time
Other contributions, such as [105], follow a mathematical ap- and prevent interference. In [111], a continuously running
proach to study the technical conditions, relevant parame- real-time genetic algorithm is used, and in [112] a genetic
ters, and engineering implementation issues for handling programming network is selected to operate the cars at
and overstepping the cars in the shaft. Some approaches equal time intervals to obtain a better performance. This ge-
[106] are traffic sensitive and depend on the traffic pattern to netic network utilizes different setting parameters, the most
give preference to a set of different optimization criteria. relevant being the maximum predicted waiting time, the

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  51


A double-decker system consists of two cabins that are
joined and travel together along the same shaft.

number of passengers to be transported, the number of al- problem, which results in improved service quality. How-
located hall calls to the candidate car, and the predicted ever, EGCSs can perform other secondary tasks to improve
riding rate (number of passengers/car capacity) of the candi- dispatching, such as reducing the amount of uncertainty
date car when it arrives at the assigned hall call. by gathering proper data or by detecting traffic patterns,
and calibration for better algorithm performance. A trend
Dynamic Sectorization Management exists regarding the development of EGCSs that are able to
One of the most frequently employed and efficient meth- integrate and fulfill these entirely different tasks.
ods for improving the handling capacity of a vertical trans- Although the main aim of the research has been focused
port installation during severe up-peak traffic patterns is to on improving the quality of service, another aim also exists
divide the building into zones. In this way, each zone can to improve the quantity of service that is offered by im-
only be attended to by the elevator that is already assigned proving the definition of the boundaries in dynamic sector-
to it; see Figure 6. Roughly, the sectorization methods can ization during up-peak periods. Consequently, the elevators
be classified as static or dynamic when the boundaries of can carry more passengers in less time. The traffic detec-
the zones are fixed or flexible, respectively. Dynamic sec- tion function is not only relevant to reduce the amount of
torization performs better because it is capable of adapting uncertainty but helps the system to autotune its own pa-
to changes in traffic flow using, for example, artificial intel- rameters according to the traffic demand. This function is
ligence algorithms. generally easily identified in research proposals because
The efficiencies of the algorithms that define the vertical transport systems are designed to function under
dynamic sectorization are related to the performances of the maximum up-peak capacity. This design results in an
the traffic-detector systems. The problem can be solved oversized system for the remaining traffic patterns, which
using dynamic programming ([113] and [114]), and the can be advantageous for minimizing criteria other than the
authors of [115] designed an algorithm for the optimal defi- typical waiting time, such as power consumption. Howev-
nition of zones based on cellular automaton. In [116] and er, although the main objective in landing-call allocation is
[117], a control system is presented that (by means of an to reduce the AWT to provide optimum services for the
artificial immune algorithm) improves the handling capac- passengers, several articles have included the tenants’ pref-
ity of the vertical transport installations, minimizes the erences in the optimization process, such as giving service
round-trip time of each installation, and maintains a uni- to preferential floors or reducing power consumption.
form round-trip time (see [118] for background on artificial Regarding power consumption, although there is a lot of
immune algorithms). In [119] and [120], the optimization of interest in sustainable development, it has not been consid-
the zone definitions is suggested by using a particle swarm ered in many articles regarding vertical transport systems.
algorithm. For a wide background on particle swarm opti- The few references that have addressed the problem do so
mization algorithms and a guide to selecting algorithm using artificial intelligence heuristics. However, the com-
parameters for successful convergence, see [121]. putational load of each algorithm and the cost of the pro-
duction relative to the benefits of their implementation are
Relevant Literature Classification typically not addressed. Furthermore, no studies exist that
Tables 3–5 summarize and classify the most relevant litera- relate these factors for modern controllers or old-fashioned
ture discussed in this article. Moreover, the references are controllers from an economic viewpoint (those based on
classified according to their aims, methodology, and the simples logic rules). Another significant aspect in vertical
criteria used. transport research is the lack of benchmarks and the non-
existence of comparisons among the EGCSs that are sug-
Conclusions and Further Research gested in the different articles. This fact, together with the
This article contains a complete survey of the EGCS lack of data, such as the computational time or the environ-
research that has been undertaken in the past. These results mental conditions set for the simulations, complicate the
are represented and classified in Table 3. Furthermore, this task of properly evaluating all of the presented approaches.
article summarizes the literature related to EGCS tasks and The dynamic sectorization problem is typically solved on
notes the relationships among the different subjects in the the timescale of several seconds, so there is usually enough
vertical transport field. Consequently, this study can serve time to calculate and evaluate all of the possible solutions
as a starting point for new research. Currently, most before choosing the best one. In those cases, dynamic sector-
research has aimed to optimize the landing-call allocation ization problems can be solved with algebraic methods

52  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


rather than artificial intelligence heuristics, the use of which board member for a number of scientific journals. He has
can unnecessarily complicate the design. been a project leader of many RTD projects funded by pub-
Several future areas of research may arise from the lic and private administrations and has been responsible
EGCSs for vertical transportation systems. One area of re- for numerous technology transfer projects. He is the inven-
search may be related to the development of special archi- tor of three international patents in vertical transportation.
tectural solutions for new EGCSs, as in the case of He has been general manager of the Office for Research
double-decker elevators, which are quite common in sky- and Technology Transfer of the University of Seville and
scrapers, but there are few references in the literature. A now is general manager of the Association for Industrial
similar area of research arises with twin systems, which is Cooperation of Andalusia (AICIA). He is a Member of the
a novel elevator architecture. Twin systems are placed on IEEE. He can be contacted at Dpto. Organización Industrial
the same rails but the decks are not connected in such a y Gestión de Empresas II, Escuela Técnica Superior de Ing-
way that the current and future positions of the other decks eniería, University of Seville, Camino de los Descubrimien-
define the maximum movement of the deck. Recently, tos s/n, E-41092 Sevilla, Spain.
MCEs (representing the general extension of twin topolo-
gies, including the possible installation of three or even References
more cars in the same shaft) have become common in new [1] M. Gerometta, P. Kazmierczak, M. Lacey, P. Oldfield, and A. Wood, “Tall
buildings in numbers. World’s tallest 50 urban agglomerations,” CTBUH J.,
skyscrapers. Consequently, a scientific literature regarding no. 2, pp. 48–49, 2009.
EGCS of MCEs is now being constructed. These two facts [2] G. Barney, Elevator Handbook Theory and Practice. London: Spon Press, 2003.
are especially relevant considering that such skyscrapers [3] P. Cortés, J. Guadix, and J. Muñuzuri, “A state of the art on the most rel-
evant patents in vertical transportation in buildings,” Rec. Pat. Comput. Sci.,
usually represent the most complex vertical transport vol. 2, no. 2, pp. 96–110, 2009.
problem. Also, from this viewpoint, the main aim of future [4] M. Thornton, “Skyscrapers and business cycles,” Quart. J. Aust. Econ.
research and development should focus on keeping pro- Spring, vol. 8, no. 1, pp. 51–74, 2005.
[5] Y. Liu, Z. Hu, Q. Su, and J. Huo, “Energy saving of elevator group control
posals and designs as simple and efficient as possible to based on optimal zoning strategy with interfloor traffic,” in Proc. 3rd Int.
facilitate real-life implementation in the industry. Improve- Conf. Information Management, Innovation Management Industrial Engineering,
ments should be numerically measured in terms of the pro- 2010, vol. 3, article 5694743, pp. 328–331.
[6] M. Z. Hasan, R. Fink, M. R. Suyambu, and M. K. Baskaran, “Assessment
vided service and the computation time that is required to and improvement of elevator controllers for energy efficiency,” in Dig. Tech.
reach the solutions. In this way, a novel proposal may be Papers—IEEE Int. Conf. Consumer Electronics, 2012, article 6241747, pp. 1–8.
properly evaluated according to its actual improvements [7] M. Hamdi and D. Mulvaney, “Prioritised A* search in real-time elevator
dispatching,” Control Eng. Pract., vol. 15, no. 2, pp. 219–230, 2007.
(whether in terms of the manufacturing costs or the ser- [8] P. Cortés, J. Muñuzuri, and L. Onieva, “Design and analysis of a tool
vices provided) rather than according to the originality of for planning and simulating dynamic vertical transport,” Simul.: Trans. Soc.
the model. Hence, it is economically beneficial to define a Model. Simul. Int., vol. 82, no. 4, pp. 255–274, 2006.
[9] P. Cortés, J. R. Fernández, J. Guadix, and J. Muñuzuri, “Fuzzy logic based
standard vertical transport simulation environment that controller for peak traffic detection in elevator systems,” J. Comput. Theor.
exactly quantifies the validity of every model and its com- Nanosci., vol. 9, no. 2, pp. 310–318, 2012.
petitiveness against other models. [10] M. M. Rashid, N. A. Rashid, A. Farouq, and M. A. Rahman, “Design and
implementation of fuzzy based controller for modern elevator group,” in Proc.
IEEE Symp. Industrial Electronics Applications, 2011, article 6108794, pp. 63–68.
AUTHORS [11] Z. Hu, Y. Liu, Q. Su, and J. Huo, “A multi-objective genetic algorithm de-
Joaquín R. Fernández is a postdoctoral researcher in in- signed for energy saving of the elevator system with complete information,”
in Proc. IEEE Int. Energy Conf. Exhibition, 2010, article 5771661, pp. 126–130.
dustrial engineering at the University of Seville. His back-
[12] T. Tyni, J. Ylinen, M. Matela, and T. Rintala, “Genetic allocation method
ground includes a B.Sc. in telecommunication engineering, for an elevator group,” U.S. Patent 7140472 B2, Nov. 28, 2006.
an M.Sc. in industrial engineering, and a Ph.D. in artificial [13] J. R. Fernández, P. Cortés, J. Guadix, and J. Muñuzuri, “Dynamic fuzzy
logic elevator group control system for energy optimization,” Int. J. Inform.
intelligence applied to vertical transportation, all from the
Technol. Decis. Making, vol. 12, no. 3, pp. 591–617, 2013.
University of Seville. His main field of research, with sev- [14] M. Z. Hasan, R. Fink, M. R. Suyambu, and M. K. Baskaran, “Assessment
eral contributions in diverse international journals, is fo- and improvement of intelligent controllers for elevator energy efficiency,” in
Proc. IEEE Int. Conf. Electro Information Technology, 2012, article 6220727, pp. 1–7.
cused on vertical transportation and elevator group control
[15] CIBSE Guide D: Transportation Systems in Buildings. London: Chartered
systems. He has worked on several projects with private Inst. Building Services Engineers, 2010.
vertical transportation companies and has two patents re- [16] K. Thangavelu, “‘Artificial intelligence’ based crowd sensing system for
elevator car assignment,” U.S. Patent 5 022 497, June 11, 1991.
lated with this field.
[17] H. Kulju and A. Vurri, “Procedure and means for counting objects
Pablo Cortés (pca@us.es) is a professor in industrial en- within a given area,” Kone elevator GB Patent 2 191 358, Dec. 1987.
gineering at the University of Seville. He obtained a special [18] H. Haraguchi, “Load distribution detecting system for elevator,” Otis
doctoral award for his Ph.D. in industrial engineering at the elevator, U.S. Patent 4 951 786, 1990.
[19] N. Kameli and J. M. Collins, “Elevator downpeak sectoring,” U.S. Patent
University of Seville in 2000. His research interests focus 5 480 006, Jan. 2, 1996.
on the fields of soft computing, computational intelligence, [20] J. Kim and B. Moon, “Adaptive elevator group control with cameras,”
operations research applied to transport and logistics, and IEEE Trans. Ind. Electron., vol. 48, no. 2, pp. 377–382, 2001.
[21] L. Lin, Z. Xiong, M. A. Finn, P. Y. Peng, M. Atalla, P. Kang, M. Misra,
vertical transportation. He is the author and referee of and C. M. Netter, “Video aided system for elevator control,” GB Patent 2 447
academic articles in many journals, as well as an editorial 829(A), Sept. 24, 2008.

August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  53


[22] N. Yamaguchi and T. Suzuki, “Functional variable remote monitoring [48] D. Nikovski and M. Brand, “Method and elevator scheduler for sched-
system and remote monitoring method for elevator,” JP Patent 2 008 081 299 uling plurality of cars of elevator system in building,” WO Patent 04 113 216
(A), April 10, 2008. (A2), Dec. 29, 2004.
[23] R. Deplazes and E. Cortona, “Three-dimensional monitoring in the [49] M. J. Atalla, A. C. Hsu, P. B. Luh, G. G. Luther, and B. Xiong, “Group
area of an elevator by means of a three-dimensional sensor,“ U.S. Patent 20 elevator scheduling with advanced traffic information,” WO Patent 2 006
060 037 818, Feb. 23, 2006. 113 598 (A2), Oct. 26, 2006.
[24] T. Nagatani and K. Tobita, “Effect of periodic inflow on elevator traffic,” [50] J. Fernández, P. Cortés, J. Munuzuri, and J. Guadix, “Dynamic fuzzy
Phys. A, vol. 391, no. 18, pp. 4397–4405, 2012. logic elevator group control system with relative waiting time consider-
[25] S. Makridakis, S. Wheelwright, and V. McGee, Forecasting Methods and ation,” IEEE Trans. Ind. Electron., vol. 61, no. 9, pp. 4912–4919, 2014.
Applications. New York: Wiley, 1998. [51] C. Kim, K. Seong, and H. Lee-kwang, “Design and implementation of
[26] K. Thangavelu, “Artificial intelligence based learning system predict- a fuzzy elevator group control system,” IEEE Trans. Syst. Man, Cybern., vol.
ing peak-period times for elevator dispatching,” U.S. Patent 5 241 142, 1993. 28, no. 3, pp. 277–287, 1998.
[27] N. Kameli, “Elevator traffic predictions using historical data checked [52] T. Ishikawa, A. Miyauchi, and M. Kaneko, “Supervisory control for el-
for certainty,” U.S. Patent 5 272 288, Dec. 21, 1993. evator group by using fuzzy expert system which also addresses traveling
time,” in Proc. IEEE Int. Conf. Industrial Technology, 2000, vol. 2, pp. 87–94.
[28] N. Kameli, “Predictor elevator for traffic during peak conditions,” U.S.
[53] N. Imasaki, S. Kubo, S. Nakai, T. Yoshitsugu, K. Jun-Ichi, and T. Endo,
Patent 5 276 295, Jan. 4, 1994.
“Elevator group control system tuned by a fuzzy neural network applied
[29] K. Thangavelu, “‘Artificial Intelligence’ based learning system predict-
method,” in Proc. 4th IEEE Int. Conf. Fuzzy Systems and 2nd Int. Fuzzy Engi-
ing ‘peak-period’ times for elevator dispatching,” U.S. Patent 5 035 302, July
neering Symp., 1995, vol. 4, pp. 1735–1740.
30, 1991.
[54] M. Amano, “Elevator control apparatus using evaluation factors and
[30] Y. Xu and F. Luo, “Pattern recognition of traffic flows in elevator group
fuzzy logic,” U.S. Patent 5 233 138, Aug. 3, 1993.
control systems based on SVM,” J. South China Univ. Technol., vol. 33, no. 6,
[55] J. O. Kim, “Group management control method for elevator system em-
pp. 32–35, 2005.
ploying traffic flow estimation by fuzzy logic using variable value prefer-
[31] F. Luo, Y. G. Xu, and J. Z. Cao, “Elevator traffic flow prediction with ences and decisional priorities,” U.S. Patent 5 679 932, Oct. 21, 1997.
least squares supports vector machines,” in Proc. 4th Int. Conf. Machine [56] D. J. Sirag, Jr., “Elevator car assignment conditioned on minimum crite-
Learning Cybernetics, Aug. 18–21, 2005, pp. 4266–4270. ria,” U.S. Patent 5 260 526, Nov. 9, 1993.
[32] T. Tyni and J. Ylinen, “Method and apparatus for allocating passengers [57] D. J. Sirag, Jr., “Using fuzzy logic to determine elevator car assignment
by a genetic algorithm,” U.S. Patent 20 056 913 117, July 5, 2005. utility,” U.S. Patent 5 248 860, Sept. 28, 1993.
[33] S. A. Sattar, J. K. Mizon, and D. J. Sirag, Jr., “Using fuzzy logic to deter- [58] K. Sasaki, K. Yokota, H. Hattori, and N. Sata, “Method and apparatus
mine the number of passengers entering and exiting an elevator car,” CA for controlling a group of elevators using fuzzy rules,” U.S. Patent 5 022
Patent 2 085 087, June 18, 1993. 498, June 11, 1991.
[34] M. Siikonen and J. Leppala, “Elevator traffic pattern recognition,” in [59] J. A. Stanley, H. Honma, D. S. Williams, T. Mori, and P. Simcik, “El-
Proc. 4th IFSA Congr., 1991, pp. 195–198. evator car dispatching including passenger destination information and a
[35] S. Hikita and K. Komaya, “A new elevator group-supervisory control fuzzy logic algorithm,” U.S. Patent 20 070 045 052 (A1), Mar. 1, 2007.
system using fuzzy rule-base,” J. Soc. Instrum. Control Eng., vol. 25, no. 1, [60] M. M. Rashid, B. Kasemi, A. Faruq, and A. Z. Alam, “Design of fuzzy
pp. 99–104, 1989. based controller for modern elevator group with floor priority constraints,”
[36] C. Kim and O. Jeong, “Group management control method for eleva- in Proc. 4th Int. Conf. Mechatronics: Integrated Engineering Industrial Societal
tor system employing traffic flow estimation by fuzzy logic using variable Development, 2011, article 5937196, pp. 1–8.
value preferences and decisional priorities,” U.S. Patent 5 679 932, 1997. [61] H. Kitano, Genetic Algorithms. Japan: Sangyo Tosho, 1993, p. 328.
[37] M. Hayase, H. Takeuchi, and Y. Suzuki, “COMMAND 6000L elevator [62] J. Alander, T. Tyni, and J. Ylinen, “Elevator group control using dis-
group supervisory control system with self-learning function,” Toshiba tributed genetic algorithm,” in Proc. Int. Conf. Artificial Neural Nets Genetic
Rev., vol. 39, no. 9, pp. 776–778, 1984. Algorithms, 1995, pp. 400–403.
[38] P. E. Utgoff and M. E. Connell, “Real-time combinatorial optimization [63] A. Miravete, “Genetics and intense vertical traffic,” Elevator World, vol.
for elevator group dispatching,” IEEE Trans. Syst. Man, Cybern. A, vol. 42, 47, no. 7, pp. 118–120, 1999.
no. 1, pp. 130–146, 2012. [64] T. Chen, Y. Hsu, and Y. Huang, “Optimizing the intelligent elevator
[39] T. Tobita, A. Fujino, K. Segawa, K. Yoneda, and Y. Ichikawa, “A param- group control system by using genetic algorithm,” Adv. Sci. Lett., vol. 9, pp.
eter tuning method for an elevator group control system using a genetic 957–962, Apr. 2012.
algorithm,” in Proc. IEEE 22nd Int. Conf. Industrial Electronics, Control Instru- [65] A. Fujino, T. Tobita, K. Segawa, K. Yoneda, and A. Togawa, “An eleva-
mentation, 1996, vol. 2, pp. 823–828. tor group control system with floor-attribute control method and system
[40] S. Oh and D. Park, “Self-tuning fuzzy controller with variable universe optimization using genetic algorithms,” IEEE Trans Ind. Electron., vol. 44,
of discourse,” in Proc. IEEE Int. Conf. Systems Man Cybernetics, 1995, pp. no. 4, pp. 546–552, 1997.
[66] S. Tsuji, “Group managing system for elevator cars,” U.S. Patent 5 780
2628–2632.
789, July 14, 1998.
[41] T. Tyni and J. Ylinen, “Evolutionary bi-objective optimisation in the
[67] E. L. Koh, J. O. Kim, and P. H. Hahn, “Group management control meth-
elevator car routing problem,” Eur. J. Oper. Res., vol. 169, no. 3, pp. 960–977,
od for elevator,” U.S. Patent 6 000 504, Dec. 14, 1999.
2004.
[68] T. Tyni and J. Ylinen, “Evolutionary bi-objective optimisation in the
[42] J. Liu and Y. Liu, “Ant colony algorithm and fuzzy neural network
elevator car routing problem,” Eur. J. Oper. Res., vol. 169, no. 3, pp. 960–977,
based intelligent dispatching algorithm of an elevator group control sys-
2006.
tem,” in Proc. IEEE Int. Conf. Control Automation, 2007, pp. 2306–2310.
[69] P. Cortés, J. Larrañeta, and L. Onieva, “Genetic algorithm for control-
[43] J. Jamaludin, N. Rahim, and W. Hew, “Self-tuning fuzzy-based dis-
lers in elevator groups: Analysis and simulation during lunchpeak traffic,”
patching strategy for elevator group control systems,” in Proc. Int. Conf. Appl. Soft Comput., vol. 4, no. 2, pp. 159–174, 2004.
Control, Automation Systems, 2008, pp. 339–344. [70] B. Bolat, P. Cortés, E. Yalcin, and M. Alisverisci, “Optimal car dispatch-
[44] J. Jamaludin, N. Rahim, and W. Hew, “An elevator group control system ing for elevator groups using genetic algorithms,” Intell. Automat. Soft Com-
with a self-tuning fuzzy logic group controller,” IEEE Trans. Ind. Electron., put., vol. 16, no. 1, pp. 89–99, 2010.
vol. 57, no. 12, pp. 4188–4198, 2010. [71] M. L. Siikonen and J. Ylinen, “Method for controlling an elevator sys-
[45] S. Wu and G. Wu, “A novel elevator group control scheduling algorithm tem,” WO Patent 05 100 223 (A2), Oct. 27, 2005.
based on pseudo differential feedback,” in Proc. IEEE Int. Conf. Automation [72] T. Shintaro, “Elevator control apparatus using neural net,” UKGB Pat-
Logistics, 2012, pp. 111–115. ent 2246214-A, Jan. 22, 1992.
[46] S. Kuzunuki, Y. Morita, K. Yoneda, T. Ueshima, and T. Tobita, “Group- [73] S. Kenji, M. Sandor, and N. Masami, “Elevator group supervisory con-
control method and apparatus for an elevator system with plural cages,” trol system using neural network,” Elevator World, vol. 44, no. 2, pp. 81–86,
U.S. Patent 4 947 965, Aug. 14, 1990. 1996.
[47] T. Tyni and J. Ylinen, “Method and apparatus for allocating landing [74] C. Imrak and G. Barney, “Application of neural networks on traffic con-
calls in an elevator group,” U.S. Patent 5 932 852, Aug. 3, 1999. trol,” in Proc. ELEVCON, 1998, pp. 140–148.

54  IEEE CONTROL SYSTEMS MAGAZINE  »  August 2015


[75] B. L. Whitehall, D. J. Sirag, Jr., and B. A. Powell, “Elevator control neural [100] J. Sorsa, M. Siikonen, and H. Ehtamo, “Optimal control of double deck
network,” U.S. Patent 5 672 853, Sept. 30, 1997. elevator group using genetic algorithm,” Int. Trans. Oper. Res., vol. 10, no. 2,
[76] B. L. Whitehall, T. M. Christy, and B. A. Powell, “Method for continuous pp. 103–114, 2003.
learning by a neural network used in an elevator dispatching system,” U.S. [101] L. Yu, J. Zhou, S. Mabu, K. Hirasawa, J. Hu, and S. Markon, “Elevator
Patent 5 923 004, July 13, 1999. group control system using genetic network programming with ACO con-
[77] I. Naoki, K. Susumu, and N. Shoji, “Elevator group control system sidering transitions,” in Proc. SICE Annu. Conf., 2007, pp. 1330–1336.
tuned by a fuzzy neural network applied method,” in Proc. IEEE Int. Conf. [102] K. Hirasawa, T. Eguchi, J. Zhou, L. Yu, J. Hu, and S. Markon, “A double-
Fuzzy Systems, 1995, pp. 1735–1740. decker elevator group supervisory control system using genetic network
[78] C. Gu, G. Zhao, and Z. Liu, “Adaptive fuzzy logical elevator group con- programming,” IEEE Trans Syst. Man Cybern. C, vol. 38, no. 4, pp. 535–550,
trol system based on artificial neural net,” Comput. Meas. Control, vol. 11, no. 2008.
1, pp. 947–949, 2003. [103] A. Valdivielso and T. Miyamoto, “Multicar-elevator group control
[79] M. Nakagawa, K. Sasaki, S. Markon, I. Nagashima, T. Midoritani, T. algorithm for interference prevention and optimal call allocation,” IEEE
Tanabe, and N. Ohta, “Elevator group supervisory control system,” U.S. Trans Syst. Man Cybern. A, vol. 41, no. 2, pp. 311–322, 2011.
Patent 5 767 461, June 16, 1998. [104] S. Tanaka and M. Watanabe, “Optimization-based collision avoidance
[80] S. Mabu, L. Yu, J. Zhou, S. Eto, and K. Hirasawa, “A double-decker ele- in multi-car elevator systems,” in Proc. ICROS-SICE Int. Joint Conf., Fukuo-
vator systems controller with idle cage assignment algorithm using genetic ka, Japan, 2009, pp. 764–769.
network programming,” J. Adv. Comput. Intell. Intell. Inform., vol. 114, no. 5, [105] J. Liu, Y. Wu, J. Dai, M. Liu, C. Wu, and E. Gao, “Overstep control
pp. 487–496, 2010. analysis for multi-car elevator,” in Proc. Int. Conf. Modelling, Identification
[81] Y. Ogoshi, H. Kimura, S. Hirose, and N. Osato, “Elevator group con- Control, Wuhan, China, 2012, pp. 254–259.
trol system using multiagent system,” Syst. Comput. Japan, vol. 34, no. 1, pp. [106] K. Ikeda, H. Suzuki, S. Markon, and H. Kita, “Traffic-sensitive con-
45–58, 2003. trollers for multi-car elevators; design, multi-objective, optimization and
[82] D. Muñoz, C. Llanos, M. Ayala-Rincón, R. van Els, and R. Almeida, analysis,” in Proc. SICE Annu. Conf., Takamatsu, Japan, 2007, pp. 2655–2662.
“Implementation of dispatching algorithms for elevator systems using [107] T. Fujimura, S. Ueno, H. Tsuji, and H. Miwa, “Control algorithm for
reconfigurable architectures,” Eng. Applicat. Artif. Intell., vol. 21, no. 8, pp. multi-car elevators with high transportation flexibility,” in Proc. IEEE 2nd
1309–1320, 2008. Global Conf. Consumer Electronics, 2013, pp. 544–545.
[83] Z. T. Wang, D. F. Ji, F. Qiao, and Q. D. Wu, “Modeling and scheduling [108] H. Ishihara and S. Kato, “Multi-car elevator control using dynamic zon-
elevator group control system based on MAS and negotiation mechanism ing,” in Proc. IEEE 2nd Global Conf. Consumer Electronics, 2013, pp. 546–549.
of agents,” Contr. Decis., vol. 22, no. 10, pp. 1184–1188, 2007. [109] M. Ikuta, K. Takahashi, and M. Inaba, “Strategy selection by reinforce-
[84] D. Pepyne and C. Cassandras, “Optimal dispatching control for eleva- ment learning for multi-car elevator systems,” in Proc. IEEE Int. Conf. Sys-
tor systems during uppeak traffic,” IEEE Trans. Control Syst. Technol., vol. 5, tems, Man, Cybernetics, 2013, pp. 2479–2484.
no. 6, pp. 629–643, 1997. [110] A. Valdivielso and T. Miyamoto, “Multicar elevator group control: Av-
[85] B. Bolat, O. Altun, and P. Cortés, “A particle swarm optimization al- erage reward learning method for service completion time reduction and
gorithm for optimal car-call allocation in elevator group control systems,” interference prevention,” in Proc. IEEE Int. Conf. Control Applications Part
Appl. Soft Comput. J., vol. 13, no. 5, pp. 2633–2642, 2012. IEEE Multi-Conf. Systems Control, Yokohama, Japan, 2010, pp. 234–239.
[86] Z. Qun, S. Ding, C. Yu, and L. Xiaofeng, “Elevator group control system [111] S. Markon, H. Suzuki, K. Ikeda, and H. Kita, “Direct control of multi-
modeling based on object oriented petri net,” Elevator World, vol. 49, no. 8, car elevators with real-time GA,” in Proc. INES 11th Int. Conf. Intelligent En-
pp. 99–105, 2001. gineering Systems, Budapest, Hungary, 2007, pp. 191–194.
[87] K. Verma and J. Goubault-Larrecq, “Karp-Miller trees for a branch- [112] L. Yu, S. Mabu, and K. Hirasawa, “Multi-car elevator system using
ing extension of VASS,” Discrete Math. Theor. Comput. Sci., vol. 7, no. 1, pp. genetic network programming for high-rise building,” in Proc. IEEE Int.
217–230, 2005. Conf. Systems Man Cybernetics, 2010, pp. 1216–1222.
[88] B. Bolat and P. Cortés, “Genetic and tabu search approaches for op- [113] W. Chan, A. So, and K. Lam, “Dynamic zoning in elevator traffic con-
timizing the hall call car allocation problem in elevator group systems,” trol,” in Proc. Int. Conf. Elevator Technologies, Hong Kong, China, 1995, pp.
Appl. Soft Comput., vol. 11, no. 2, pp. 1792–1800, 2011. 132–140.
[89] P. Cortés, L. Onieva, J. Muñuzuri, and J. Guadix, “A viral system algo- [114] A. So and W. Chan, “Comprehensive dynamic zoning algorithms,”
rithm to optimize the car dispatching in elevator group control systems of Elevator World, vol. 45, no. 8, pp. 99–109, 1997.
tall buildings,” Comput. Ind. Eng., vol. 64, no. 1, pp. 403–411, 2013. [115] Y. Xu, F. Luo, and J. Wang, “A new modeling method for elevator group
[90] Chartered Institution of Building Services Engineers, Transportation control system with cellular automata,” in Proc. 5th World Congr. Intelligent
Systems in Buildings CIBSE Guide D. London: CIBSE, 2005. Control Automation, 2004, vol. 4, pp. 3596–3599.
[91] F. Nowak and J. Luce, “Multicompartment elevator call assigning” U.S. [116] Z. Li, “Traffic based on artificial immune algorithm research on dy-
Patent 4 632 224, Dec. 30, 1986. namic zoning of elevator traffic based on artificial immune algorithm,” in
[92] K. Sasaki, S. Markon, and M. Makagawa, “Elevator group supervisory Proc. Control Automation Robotics Vision Conf., Guangzhou, China, 2004, vol.
control system using neural networks,” Elevator World, vol. 44, no. 2, pp. 3, pp. 2170–2171.
81–90, 1996. [117] Y. Xu, F. Luo, and X. Lin, “Hybrid destination registration elevator
[93] T. Mori, Z. Bahjat, J. Stanley, M. Ross, M. Sahara, and H. Honma, “Con- group control system with artificial immune optimization algorithm,” in
trol system and control method for reassigning the cars of a double-deck Proc. 8th World Congr. Intelligent Control Automation, 2010, pp. 5067–5071.
elevator,” U.S. Patent 6 419 051, July 16, 2002. [118] U. Aickelin, D. Dasgupta, and F. Gu, “Artificial immune systems,” in
[94] M. Kostka and K. Steinmann, “Method of controlling elevator installa- Search Methodologies: Introductory Tutorials in Optimization and Decision Sup-
tion with multiple cars,” U.S. Patent 6 508 333 B2, Jan. 21, 2003. port Techniques, E. K. Burke and G. Kendall, Eds. New York: Springer, 2014,
[95] M. Siikonen, “Procedure for control of an elevator group consisting pp. 187–211.
of double-decker elevators which optimizes passenger journey time,” U.S. [119] Z. Li, H. Z. Tan, and Y. Zhang, Particle Swarm Optimization Applied to
Patent 6 237 721, May 29, 2001. Vertical Traffic Scheduling in Buildings (Lecture Notes in Computer Science).
[96] M. L. Siikonen, U.S. Patent 401874 B2, June 11, 2002. Berlin Heidelberg, Germany: Springer-Verlag, 2007, pp. 831–838.
[97] M. L. Siikonen and J. Sorsa, “Optimal control of double-deck eleva- [120] L. Fei, Z. Xiaocui, and X. Yuge, “A new hybrid elevator group control
tor group using genetic algorithm,” Int. Trans. Oper. Res., vol. 10, no. 2, pp. system scheduling strategy based on particle swarm simulated annealing
103–114, 2003. optimization algorithm,” in Proc. 8th World Congr. Intelligent Control Auto-
[98] J. Ylinen and T. Tyni, “Genetic procedure for multi-deck elevator call mation, 2010, pp. 5121–5124.
allocation,” U.S. Patent 6,293,368 B1, Sept. 25, 2001. [121] I. C. Trelea, “The particle swarm optimization algorithm: Conver-
[99] M. Ruokokoskia, J. Sorsaa, M. L. Siikonen, and H. Ehtamo, “On the gence analysis and parameter selection,” Inform. Processing Lett., vol. 85, no.
quality of the assignment model for the elevator dispatching problem,” 6, pp. 317–325, 2003.
submitted fpr publication.


August 2015  «  IEEE CONTROL SYSTEMS MAGAZINE  55

You might also like