ArticlePDF Available

Towards A Global Traffic Control (Dispatcher) Algorithm - Requirements Analysis


Abstract and Figures

This paper presents an analysis of the requirements of a Global Dispatcher Interface for the control of a group of lifts. The information passed to and from the interface is defined as well as the common processing which will be executed on that information in order to generate the response. Using recognised software development methods, requirements are elicited from a consideration of the significant use cases and the architectural configurations which must be supported by the interface. The analysis concludes by defining the roles and responsibilities of the key objects of the software. A final section proposes relevant standard open technologies that avoid proprietary and potentially incompatible and high maintenance solutions.
Content may be subject to copyright.
Towards A Global Traffic Control (Dispatcher) Algorithm -
Interface Requirements Analysis
Jonathan Beebe
Jonathan Beebe Ltd., 2 Heap Bridge, Bury, BL9 7HR UK,
Keywords: Global dispatcher, standard interface, group control, requirements analysis, use case
Abstract. This paper presents an analysis of the requirements of a Global Dispatcher Interface for
the control of a group of lifts. The information passed to and from the interface is defined as well as
the common processing which will be executed on that information in order to generate the response.
Using recognised software development methods, requirements are elicited from a consideration of
the significant use cases and the architectural configurations which must be supported by the
interface. The analysis concludes by defining the roles and responsibilities of the key objects of the
software. A final section proposes relevant standard open technologies that avoid proprietary and
potentially incompatible and high maintenance solutions.
This paper presents an analysis of the requirements for a Global Dispatcher Interface for controlling
a group of lifts, a concept which was earlier proposed by Peters[1]. While definitions of standard
group control algorithms have been documented[2], the reality of group control to date is that
manufacturers have created proprietary designs that are inextricably linked to their own lift equipment
so that it is not possible accurately to compare or predict performance of different control policies
during the design phase of a building or in advance of a refurbishment of the lifts. The benefit of a
standard interface is that it would be possible to supply the dispatching capability in a component
form that could be "plugged" into any group of lifts that conforms to the interface.
Secondly, a dispatcher design which has been configured and validated using simulation can be
transferred directly into a physical installation with confidence, if both the simulator and the real lifts
use the same standard interface.
Additionally, as lifts become better integrated with the other services of so-called "smart buildings"
and with the introduction of applications that allow passengers to register requests for lift travel via a
variety of channels[3] including personal mobile devices[4], an interface that allows simplified and
standardised access to the group call assignment mechanism becomes increasingly desirable.
The objective of the current paper is not to deliver a complete design for such a control system instead
a structured statement of the requirements that such a system must satisfy followed by an analysis of
those requirements is presented as the initial phase of development of such a system. It is therefore a
guide for designers proposing to implement a Global Dispatcher Interface.
Furthermore, it is the standard requirements of the dispatcher interface that are analysed here omitting
analysis of any specific dispatcher (i.e. the logic that determines which is the best lift in the group to
be assigned to answer a call, which is sometimes referred to as the group or traffic "control
A key consideration for this paper, in addition to established conventions of lift system design, is a
review of current trends and possible future developments in lift group controller technology so that
the requirements identified are sufficiently broad and flexible to avoid the analysis becoming
prematurely outdated.
The software requirements analysis process that has been used here forms an early phase of a
lightweight Software Development Lifecycle Process (SDLC) and is supported by a UML Modelling
tool[5] to undertake:
- Use case identification and analysis[6] for the significant processes which the interface must
leading to:
- Domain analysis[7] by requirement identification and refinement
which is achieved through analysis of published reference works [9],[2].
The paper starts with an analysis of the operation of a generic group of lifts from the perspective of
the travelling passenger. This is presented as the passenger use cases. There follows a section which
presents the main dispatcher system use cases, which are the result of analysing the passenger use
cases. The analysis concludes with overview definitions of the key collaborating objects, deduced by
analysing the dispatcher system use cases. These are necessary for the Global Dispatcher Interface to
be globally applicable in a number of identified potential interaction modes. Finally, under a number
of topic headings, a discussion is presented of the requirements that will enable a Global Dispatcher
Service to operate safely, securely, flexibly and manageably in a networked environment with
recommendations that open standards (that are non-specific to the environment of passenger lift
systems) are adhered to. Thus, by the conclusion of the paper we have an analysis of the Global
Dispatcher Interface, which is a suitable starting point for the design of specific implementations.
In order to present a paper of the size and detail that is suitable for publication, the current document
is necessarily a summary of a large body of work consisting of diagrams and documentation relating
to a Global Dispatcher Service, which has been collected in the form of a UML model[8]. References
are included in this paper to on-line documents which have been generated from the model and which
provide more detailed elaboration supporting the concepts presented here. However, it is intended
that it should be possible simply to understand the discussion presented here without the need to make
cross-reference to such elaborations.
The paper begins with an overview of some relevant terms in the domain of lift control which makes
reference to definitions in earlier publications [9],[10].
Landing call (LC)a request for travel from a landing. This may be for
- a specified destination floor
or simply
- to travel in specified direction.
Although it may be possible to infer a direction of travel from a landing call which only calls a lift to
a floor with no information about intended travel, this is not considered relevant to the analysis
presented here, because it would thereby be converted into one of the above cases.
The process of registering a landing call may take place on the landing or at a physically remote
location (via dedicated hardware or possibly a personal mobile device). It is helpful to group landing
call devices in order to take account of the approximate time taken for passengers to travel from one
to the other after registering their call. These groups have traditionally been referred to as "risers"
because of the cabling used, but in the context of this paper the grouping covers a broad range of
configurations according to their distance from the entrance(s) of lifts to which a call might be
Group of lifts may be several cars, each travelling vertically in an independent dedicated shaft or,
as in recently developed examples, multiple cars capable of both horizontal and vertical travel (or a
combination), occupying the same shaft concurrently, with the ability to transfer between shafts.
Carsmay carry passengers on one or more decks (stacked vertically) so that passenger arrival and
destination floors must be coordinated to minimise or remove the possibility of a car stop where the
passengers on one of the decks see no-one enter or leave, which would be frustrating or possibly
Algorithm the intelligence, in whatever appropriate form, which determines which lift will be made
responsible (i.e. "assigned") for servicing a passenger's landing call. Often called a "Group Control
Algorithm" or "Traffic Control Algorithm"[9]. This intelligence is encapsulated in software that is
referred to in the current paper as "Cost Function + Heuristics" or simply the "dispatcher".
Standard Elevator Information Schema (SEIS) [10] is a standard for communicating static
information, such as configuration details, dynamic information such as current floor and registered
calls and events such as call registrations and car trips, between all manner of systems and users of
passenger lifts. It comprises a set of definitions of complex and simple data types and the structures
in which they may be used. This paper makes frequent references to the schema, both for data types
of parameters passed in messages and also for the internal data structures of the dispatcher interface,
and these are indicated by text in CamelCase which includes a hyperlink to the definition on the
website where the schema is published.
Global Dispatcher Serviceis a standard, non-proprietary mechanism for assigning a landing call
to the lift car most suited to serving that call. It therefore includes an element - the "dispatcher" -
which can produce the optimal assignment decision. The Global Dispatcher Interface encapsulates
the dispatcher and is the route by which all access to the Global Dispatcher Service is made.
This section simply identifies the use cases that are of key relevance to the Global Dispatcher
Interface. The complete set of detailed use case documentation, which provides the "Use Case Story"
is published separately[11]. The text of each use case story is essential to understanding the operation
of the Global Dispatcher Interface. However, only the most significant use cases are included in the
current paper for reasons of size.
Users (passengers) interact with the system (Global Dispatcher Interface) in two separate (although
loosely interlinked) processes, which are defined as the "Passenger Use Cases":
- Request to travel (Landing Call)
- Travel to destination
These processes must be supported by the Global Dispatcher Interface and are described and
elaborated using use case analysis.
In each use case the passenger undertakes a different activity and since each use case process depends
on the independent and unpredictable actions of individual human beings, we must conclude that
these use cases represent concurrent and unsynchronised (asynchronous) processes.
3.1 Use Case Analysis - Passenger Use Cases
Figure 1 Passenger Use Cases
The Passenger use cases describe the flow of interactions between the passenger and the elements of
the lift system to achieve the use case objective. They are a view of the interaction of the lift system
with its external environment (see Figure 1).
The main actor of these use cases is "the passenger" but it is important to note that many instances of
these use cases may be active concurrently and asynchronously so they may exist in as many different
stages of completion. One passenger, having partly completed one of the above use cases may affect
the experience of another passenger or be affected by another passenger who is at a different stage of
completion of the same or a different use case. This means that other passengers are referenced in the
use case text who are not the main actor and should not be considered as actors in the use case in any
The flow of events of each use case in the achievement of its specific goal is described in words by
the Use Case Story:
3.1.1 Request to travel (Landing Call) - Use Case Story Main Flow
The passenger approaches a call station or activates an application on a smart device.
The passenger registers a request to travel which includes:
- Origin floor
- Destination direction OR destination floor
Optionally other information may be supplied in the request such as:
uc Passen ger Use Cases
Lift System
Passe nge r
Request to travel
(Lan din g Call )
Travel to de stination
Call station/device
Car stati on
- Priority status of the passenger or requested journey
- Number of passengers travelling as a group
- Access security details
- The expected delay before the passenger will be ready to enter the lift
The dispatcher responds via the call station (or smart device) to the passenger to:
- Confirm the request (e.g. illuminated button) if not already activated
plus optionally:
- Inform the passenger which car to travel in
The dispatcher also communicates the assignment directly to the assigned car.
The car responds by adding the assigned call to its travel plan and in due course departs with a
destination of the passenger's origin floor.
On arrival at the passenger's origin floor the car cancels the passenger's call and informs the dispatcher
of the call cancellation.
The dispatcher indicates the cancellation to the passenger via the call station and optionally other
indicator devices (including possibly the smart device, if in use).
The dispatcher records the System Response Time for performance analysis purposes.
If the dispatcher has not been informed of the passenger's destination floor the passenger's call is then
deleted from the list of current calls, although the dispatcher retains an expectation that one or more
new car calls may be generated as a result of passengers entering the car. However, if the destination
floor is known then the call is retained but its status is changed to "Answered".
3.1.2 Travel to destination - Use Case Story Main Flow
The car arrives at the passenger's origin floor and opens its doors to allow:
- any arriving passengers to exit from the car
- the waiting passengers to enter the car.
A delay is initiated for the car doors to stay open in order to allow these transfers.
If the dispatcher already knows the passenger's destination floor it will establish with the arriving car
a car call for the destination floor, which may be displayed on the car station.
If the dispatcher does not know the passenger's destination floor, then the passenger registers their
destination floor via the car station.
Whichever mechanism is used to register the car call results in a stop at the passenger's destination
floor being included in the car's travel plan (although this may already be in place due to the requests
of other travelling or waiting passengers).
After allowing passengers to transfer to and from the current stop floor, the car closes its doors and
departs for the next destination in its travel plan.
On arriving at the passenger's destination floor the car cancels the car call for that floor and informs
the dispatcher.
The car opens its doors to allow the arriving passenger to depart and then any further waiting
passengers to enter the car to begin travel to their own destination floors.
3.2 Use Case Analysis - Dispatcher System Use Cases
Now that the Passenger Use Cases have been established it is possible to extract from them the roles
and responsibilities of the Dispatcher, namely:
- To monitor and retain the current state of each car (floor position, motion, doors,
registered/cancelled calls).
- To accept passenger landing call requests and inform passengers, by various means what
response to expect.
- To assign (and it is reasonable to assume possibly to re-assign) landing calls to the most
appropriate cars.
From these the main system use cases of the dispatcher can be developed. Figure 2 shows the
dispatcher system use cases and the actors which initiate them.
N.B. The two "Register client" use cases have been inferred since the dispatcher will necessarily need
to know what clients are available (which may be a dynamic property). However, these use cases are
not felt to be concerned directly with the primary activity of the dispatcher in making call assignments
so have been omitted here for brevity, although they are documented elsewhere[11].
Figure 2 Main Dispatcher System Use Cases
The dispatcher system use cases describe the flow of interactions between the elements of the lift
system in order to support the fulfilment of the Passenger use cases. In that sense they are a simple
view onto the internal operation of the lift system. (N.B.: there is not a simple relationship between
the Passenger and System use cases).
The next sub-sections contain the use case story of each of the two most significant dispatcher system
use cases:
- Update Client Car Status
- Assign Call
N.B. References to "Car" in the following text are purely abstract and do not relate to the physical
car or its control equipment. Similarly for references to the "dispatcher".
uc M ain Sy stem Use Case s
Dispat cher
Register Client
Update Client Car
Assi gn Call
Main Dispatcher System Use Cases
This di ag ram s hows the mai n use
cases of the Global Dispatcher
service. Although the text of these
use cases refers to requests and
responses it is not necessarily the
case that there is a one-to-one
relations hip between a us e cas e a nd
a s ing le reques t.
Canc el Call
Register Client - Car Register Client -
Passe nger Call Stati on
Client Car
Client Passenger Call
3.2.1 System Use Case - Update Client Car Status Main Flow
A client car announces a change of its state to the dispatcher by providing details of its dynamic
information at the time they occur:
- The ID supplied by the Service in response to the Introduce Client request
- The client's own identifier
- Current Floor
- Direction
- Drive state
- Door state
- Travel Plan
- Shaft
- Currently registered calls (Car, Landing, Park)
All this information may be supplied in a single request (CarDynamicData) or, preferably a sequence
of incremental requests as each change of state event occurs (LogEventType). All dynamic data and
events are time-stamped to avoid errors due to communication delays.
At the discretion of the specific dispatcher algorithm (not the dispatcher interface) an update of car
status may initiate the re-assignment of calls, which is shown by the extends relationship to the Assign
Call use case.
3.2.2 System use case - Assign Call
Any client (including a passenger signalling device) may imply a request for a CallAssignment in the
form of a CallRegistration event (see Figure 3).. The event includes the following information:
- Call Floor
- Direction or Destination Floor
- Registration time
Figure 3 Structure of CallRegistration event
(N.B. inclusion of StartTime means this request may (re)occur at any time after the initial call
registration event and may therefore be a request for a call to be re-assigned or simply a delayed
request if no car was available for assignment earlier. In the case of re-assignment the dispatcher
will have a record of the currently assigned car and this information may influence the result of the
new assignment, though this would be a characteristic of the specific dispatcher algorithm and not
of the dispatcher interface.
At the discretion of the specific dispatcher algorithm (not the dispatcher interface) an Assign Call
request may initiate the re-assignment of other previously assigned calls, which will result in multiple
executions of the Assign Call use case (i.e. once for each call to be re-assigned). Main Flow
The dispatcher calculates the cost of assigning the call to each of the registered client cars according
to its own internal algorithm design. The term "cost" is not restricted to a purely financial cost and
may be evaluated in terms of one or more criteria such as:
- waiting time,
- system response time,
- energy consumed,
- etc
as a function of the increase in the value of that parameter after the call has been assigned compared
to the cost before it was assigned to the car. The algorithm may include penalties or incentives that
are derived from a logical analysis which will modify the simple cost.
If an overriding criterion is included in the algorithm, such as never assigning a call that would cause
the car to become overloaded, then that car will be marked as "blocked" (N.B. this type of logic is
later termed a "heuristic").
N.B. Availability for assignment may be determined by a variety of properties such as:
the operating mode of the car
whether the car is able to service the call floor(s)
However, the availability decision is a characteristic of the specific dispatcher algorithm and not of
the dispatcher interface.
The response is broadcast by the dispatcher to all registered clients and includes (but is not limited
- the CallRegistration call event updated with the cost of assignment, where cost analysis is
made in terms of the cost-function that is specific to the algorithm used by the dispatcher
plus optionally:
- a CallAssignment event where the Assigned To element is populated with the minimum-cost
assignment details.
- a TravelPlan (see Figure 4) if the dispatcher is configured to hide assignments so that the car
will by-pass an assigned call until it becomes the next landing call for the car.
- if the call was previously assigned to a different car the response will also include a
CallDeassignment event.
The Assignment is only considered to be complete after a positive acknowledgement has been
received from the designated car.
N.B. An important conclusion that can be drawn from this use case is that a single CallRegistration
event message results in multiple further requests to more than one client of the local dispatcher
interface. The secondary messages are requests in their own right and should not be confused, in the
context of the local dispatcher interface, with response messages. Alternative Path 1)
If no acknowledgement or a negative acknowledgement is received from an assigned client car then
the call assignment will be repeated with that car excluded. Alternative Path 2)
The dispatcher may be distributed as a number of collaborating instances, each responsible for a
unique set of one or more registered client cars. In this case each dispatcher instance will respond
with a "bid" publicising the minimum cost of assignment for adding the call to the travel plan of one
of its registered client cars. If having bid, the dispatcher instance is notified by another instance of a
lower bid, the assignment will be relinquished in favour of the lower cost bid.
In this mode, the Assignment is only considered to be complete after a positive acknowledgement has
been received from the dispatcher responsible for the assigned car.
Figure 4 Structure of TravelPlan
Throughout the analysis process, requirements are identified and these are catalogued[15] to provide
a precise definition of the functions, features and capabilities that the system must support once it has
been developed. The requirements catalogue therefore forms the basis of the contract between
developers and users, as well as the basis for the suite of tests which the system must pass before it
can be declared to be finished.
This section addresses the architectural requirements that are derived in part from the use cases and
also from study of published reference works.
4.1 Controller Interaction Modes
A lift group controller architecture is required that will enable the dispatcher system use cases to
provide the basic interactions between the car and dispatcher allowing:
- the dispatcher to notify the car of its currently assigned LC(s)
- the car to update the dispatcher with its current status (changes) and call
registration/cancellation events.
Controller interaction modes to be supported include (though may not be limited to) the following:
Dedicated Group Controller (Simple Hierarchy)[9] where a Group Controller monitors
landing call stations, assigns LCs to cars and cancels the call at the landing when informed by
the car controller that the assigned car has answered the call. This architecture permits the use
of very simple car control, allowing standard hardware to be used across single and multi-lift
installations and which might, if desired, be implemented in a non-computerised technology.
It should be remembered that this architecture is extremely vulnerable as the service of entire
group of lifts is dependent on a single component and so a backup control policy that provides
a minimal service must be provided in case of failure.
Master/Slave[9] - only car controllers exist but all are capable of assuming the role of Master,
which oversees the assignment of LCs to specific cars. A decision process is defined for
designating which car controller is the current master and also selecting a new master in case
of failure. This interaction mode provides a resilient solution, where failure of one or more
elements leads to a "graceful degradation" of service.
Assignment Bidding (N.B. This is a suggestion by the author as a logical development of the
master/slave mode; no published reference available.) every car controller is capable of
preparing a bid for responding to an LC and the one offering a minimum "cost" is awarded
the assignment, either directly by all other controllers or by a designated group master. This
option becomes more attractive as the number of cars in a group and the number of floors
served increases, where the number of possible solutions to the assignment task results in a
highly demanding computing load that is too great for a single processor to deal with while
providing a timely response. The complexity of calculations and consequent computing
demand will increase significantly with the introduction, recently announced[12], of vertical
transportation systems based on multiple small rope-less cars that share and can transfer
between shafts[13]. It is likely that an architecture supporting this interaction mode will
become more widely used.
Some algorithms operate by generating assignments (and re-assignments) for all currently
registered LCs at the same time, i.e. in a single complete execution of the algorithm, for
example when employing so-called "genetic" technology. In this case the assignments must
be only for those client cars which are registered locally with the dispatcher instance, thereby
excluding the Assignment Bidding architecture since the bidding process is effectively
internal to the algorithm (distribution of computational processes, with its associated benefits
of scalability and resilience is not excluded but becomes proprietary). However, triggering of
the assignment process will be caused, as in all of the architectures by a change of state in
some significant data - primarily but not exclusively, the registration of a new LC at a landing
call station. The cost of assignment is still the fundamental driver for the decision process and
should be reported as an element of each optimised assignment because it is therefore critical
to any auditing or analysis of the performance of the algorithm.
The above interaction modes place requirements on the software architecture of the Global Dispatcher
Interface. However, it remains a subsequent design decision as to whether the software is to be
implemented in a hardware architecture that exactly replicates that of the software, with network
communications between distributed electronic components or whether, in reality, it is aggregated
within a single device/computer (or indeed any other intermediate level of integration). The advantage
of considering the software architecture separately from the hardware is that the designer is then free
to choose a distribution of software functionality and a configuration of hardware elements that meets
other requirements, for example for resilience and fault-tolerance, communication latency and
bandwidth as well as physical location. Again, with the introduction of machine-room-less lifts, the
control components and communication distances may be very significant, requiring specific
4.2 Key Collaborating Objects
Careful examination of the text of the dispatcher system use cases[11] exposes the names, roles and
responsibilities of the key structural elements that will come to constitute the eventual solution (both
external and internal to the dispatcher). It is important however to understand that these elements are
not optional nor are they specific to a particular implementation of the Global Dispatcher - they are
necessary for the Global Dispatcher Interface to be truly global and to operate according to the system
use cases in any one of the interaction modes.
The aim of this analysis is to understand and define, in abstract terms, the essential elements of
software of the Global Dispatcher Service independently of a consideration of constraints, such as
the hardware of existing products or limitations of current technologies, in order to provide an optimal
breakdown of the overall system into its fundamental components, which can then readily be
implemented in many different configurations with the greatest flexibility and a minimum of
N.B. A further phase of analysis will lead to the production of a set of Use Case Realisations[6] in
the form of interaction diagrams that demonstrate the operation of each of the architectural
configurations described above (to be discussed in a subsequent paper).
The key collaborating elements (objects) are identified in the above dispatcher system use cases,
which together comprise the Global Dispatcher Service.
4.2.1 Dispatcher Interface
The dispatcher interface provides an encapsulation of the entire dispatcher service and is its
presentation to the external environment. For example, in addition to fulfilling assignment requests
in the form of call registration events and accepting status update events from lifts, this interface also
provides access to performance monitoring and fault logging functionality.
The dispatcherinterface acts in a similar fashion to a database in that it processes events with generic
Create/Read/Update/Delete (CRUD[17] ) operations rather than function calls with names that relate
specifically to the operation of a dispatcher.
The assignment bidding interaction mode requires that several instances of the dispatcher interface
must be able to co-exist and interoperate as parts of the same Global Dispatcher Service though this
interoperation may be via a different (internal) interface (i.e. not the Global Dispatcher Interface).
4.2.2 Heuristics + Cost Function
This object is responsible for estimating the cost for the identified car of adding one call to its list of
assigned calls, by considering the current state of all cars and other assigned calls. The estimation is
initially calculated by applying a Cost Function. This calculation is modified and possibly overridden
(e.g. to block the car completely from bidding for the assignment) by the application of rules or
So the capability of this object supports all manner of assignment algorithm mechanisms - from
simple relay ladder logic (as in a Programmable Logic Controller) to neural networks and genetic
4.2.3 Bid Manager
The Bid Manager is responsible for eliciting bids for all the cars that are local to the dispatcher and,
for the assignment bidding interaction mode, capturing bids for all cars that are registered with remote
dispatcher instances. Then selecting the minimum cost bid.
Finally, the Bid Manager must notify all the local registered cars plus remote dispatchers of the
winning bid, in the form of a CallAssignment event to the winner and CallDeassignment events to
the losers.
4.2.4 All Cars State
Responsible for holding the current state of all cars in the group (i.e. not just those that are registered
locally). The stored information includes both static and dynamic data for each car. The StaticData
of all cars (local and remote) in the group is populated and held in the AllCarsState entity for future
reference by the dispatcher while servicing subsequent client requests.
When a state change occurs to a "local" lift, this must be replicated in the data of the AllCarsState
object. The change is signalled by a LogEventType request message sent via the local dispatcher
interface. This LogEventType message must be replicated to all "remote" instances of the dispatcher
interface. In a complimentary manner, the remote instances of the dispatcher interface must signal all
change of state events to the "local" instance of the dispatcher interface so that the state of the local
instance of AllCarsState can be updated. The LogEventType message is an efficient means of
signalling only the information which has changed and implies an event-driven mode of operation.
Communication/interaction between asynchronous processes is well suited to the Event Driven
Pattern[16] of programming - for example: change of next possible stopping floor, call registration,
4.2.5 Current Landing Calls State
Responsible for holding the current state of all currently registered landing calls. Destination landing
calls continue to be held after the assigned lift car arrives at the passenger's origin floor until the car
arrives at the passenger's destination floor. The call state includes registration time, cancellation time
and allocated car. The CallCycle (see Figure 5) is an efficient means of storing the complete landing
call state.
Figure 5 Structure of CallCycle
4.2.6 RegisteredClients - Cars
The list of registered car clients. Local car clients register directly with the dispatcher interface with
which they are associated. Remote car clients register indirectly via the dispatcher interface with
which they are associated.
4.2.7 Registered Clients - Passenger Call Stations
The list of registered passenger call station clients. Call stations are grouped (as "risers") to enable
the dispatcher to factor the time needed for a passenger to arrive at the entrance of the assigned lift.
The term "riser" is derived from conventional lift installations but in this context may designate a
group of mobile devices that might not even be in the same building as the lifts.
Local call station clients register directly with the dispatcher interface with which they are associated.
Remote call station clients register indirectly via the dispatcher interface with which they are
4.2.8 Car Gateway
A further software element is needed although it is not an integral part of the Global Dispatcher
Service. This is referred to as the Car Gateway and it is required to ensure that call assignment
decisions from the dispatcher can be effected by the lift car to which the assignment has been made.
The objective of the Global Dispatcher Service is to provide, via the Global Dispatcher Interface,
assignments to a wide variety of lift equipment from a range of manufacturers and so a standard car
interface must be defined. In addition to requiring car status information to be supplied in a
standardised form, the dispatcher interface requires the car controller to appear to operate a
Directional Collective (also known as Simplex Collective see [9] sec control policy, which
allows multiple concurrent LC assignments to a car. If the actual car controller being used does not
operate in this way, it will be necessary to create a gateway software component which is specific to
the car controller concerned and which then offers the standard, Directional Collective interface to
the dispatcher.
Amongst other duties, the gateway will accept multiple assignments and then select one that is the
most immediate LC to pass on to the underlying car controller. This aspect of the gateway
functionality would be most effectively and efficiently based on the car's Route data element (of type:
TravelPlan)[10], though this is not in itself a requirement.
It is noteworthy that the complex type :FloorStop(), which is part of the TravelPlan type, optionally
includes shaft occupancy as well as floor level information and is therefore forwards-compatible with
(ropeless) vertical transportation systems based on multiple cars that share and can transfer between
shafts. Whether or not a car's TravelPlan is communicated, a change of the shaft currently used by
each car is, in any case, communicated via the CarShaftEvent type.
The gateway could also assemble statistical information on car static data, for example:
- door timings (DoorProfile)
- inter-floor trip times (SpeedProfile)
- energy consumption (ReferenceEnergyProfile)
for use both by the dispatcher cost function and also for more general logging and maintenance
Whilst it is common practice for data-logging functionality, which supports remote monitoring of lift
status, door operation, fault conditions, etc to be included in group control equipment and even though
all such information is available to the dispatcher, this capability is not specific to the dispatcher and
so will not be described here. However, there is a definite requirement for the dispatcher to report:
- bid preparation (assignment cost calculation and application of rules)
- bid selection
- error and exception conditions (e.g. preventing assignment, client registration failures,
This capability is essential since the dispatcher is expected to be globally compatible with all
configurations of groups of lifts and without it, the dispatcher would remain a "black box" making
inexplicable and apparently arbitrary decisions.
It is the responsibility of the dispatcher interface to pass the reports to the external environment,
though a separate communication channel from that which handles assignment requests, etc may be
employed actually to pass the data.
The dispatcher interface will normally be connected to a network, so all the elements of the Global
Dispatcher Interface must be protected from infiltration and attack by malevolent agents. Precautions,
such as firewalls, authentication (built into the client registration process) and possibly encryption of
communication channels should be implemented using standard components which can be obtained
from third parties specialising in these complex technologies and who will provide maintenance
updates to protect against newly discovered threats and weaknesses. It is the intention of this paper
to note the requirement rather than to present the details of such measures.
There is as yet no standardisation, either of the names or the nature of the parameters that are required
for configuring a dispatcher to match the building in which it will operate or the lifts it will control.
Furthermore, different parameters will be required, depending on the design of the algorithm and the
implementation of a particular manufacturer. However, there remains a requirement for the dispatcher
interface to provide a mechanism whereby parameters may be set during installation and possibly
also modified during the service lifetime of the lifts, for example if the building occupancy or
utilisation changes, or simply an unusual event occurs - a conference is scheduled, an emergency, etc.
It is to be expected that new versions of the dispatcher software will be released, either to fix errors
or deficiencies in the current version or simply to introduce new features and improvements. While
this update capability might be undertaken by a technician on site substituting physical components
with replacements that have been pre-loaded with the updates, it would be quicker and easier to
control the process if such updates could be performed via secure remote access to an update
mechanism. This is increasingly the method of choice for applications running on general purpose
computers and mobile devices. While not an essential requirement of the Global Dispatcher Interface,
a remote software installation and update facility remains a highly desirable feature.
Lift systems exist to provide a service to the users of buildings (passengers, managers, maintenance
technicians and owners) and increasingly will need to integrate with other sophisticated services of
smart buildings and cities within which they are located. In order to achieve such integration it will
be necessary to publish the capabilities and to control access to dispatcher services as resources in a
standardised form. For example, information such as the physical location of the building and the lifts
within it and the configuration of dispatcher services needs to conform to agreed standards. Third
party registries (e.g. Waher[18]), which provide categorisation and publication of resources already
exist and are subject to ongoing development, with inbuilt security. These repositories should be used,
in preference to proprietary solutions, so as to avoid inconsistencies and resulting fragmentation of
the potential for integration of the Global Dispatcher Service.
While taller buildings and novel configurations of lifts place greater computational demands on the
call assignment mechanism (dispatcher), lift manufacturers develop increasingly sophisticated group
control algorithms in order to optimise the available resources. The task is further complicated with
the introduction of new passenger signalling devices and modes of interaction, including mobile
devices. Currently, there is no accurate means by which the algorithms of different manufacturers
may be compared nor is there a reliable method for integrating the lifts of one manufacturer with the
group control policy of another. In this paper a Global Dispatcher Interface has been described that
brings standardisation to this problem domain and thereby allows architects and building owners to
design, manage and maintain modern buildings with greater confidence.
The first phase of the process of defining the Global Dispatcher Interface, that of capturing and
analysing the requirements has been described, using established software development methods,
through the documentation of Passenger Use Cases and Dispatcher System Use Cases. The
architecture of modern lift control systems has also been discussed yielding further requirements and
leading to the identification of key software elements of the Global Dispatcher Service with
descriptions of their roles and responsibilities. An overview of the analysis work has been presented
in this paper, while references are given to the complete details in the form of a structured UML
model and supporting documents generated from that model. The referenced documentation is
published on the worldwide web.
A further paper is planned to describe the design and implementation of a prototype Global Dispatcher
Interface, which is currently the subject of ongoing research and development.
The material presented in this paper is, by its nature, a proposal and has yet to be implemented
commercially. The author welcomes comments and questions, via the Editor, regarding possible
improvements, errors and omissions.
[1] Peters, P. (2016). Global Dispatcher Interface - Proceedings 6th Symposium on Lift &
Escalator Technologies (pp. 18-1 18-9). Northampton: CIBSE.
[2] G. C. Barney, G.C. and Al-Sharif, L. (2016) Elevator Traffic Handbook, Second edition,
Chap.12, Routledge, Abingdon UK, 2016, ISBN 978-1-138-85232-7.
[3] thyssenkrupp Elevators Corporate Website . AGILE., From:
[Accessed February 16, 2018]
[4] KONE Corporate Website, "KONE DESTINATION SOLUTIONS", 19/02/2016,
[Accessed February14, 2018].
[5] Sparx Systems Corporate Website. Sparx Enterprise Architect. From: [Accessed Jan, 2018].
[6] Bitner, K and Spence, I. (2003). Use Case Modelling, Pearson Education Inc., Boston USA,
[7] Eriksson, H.-E., & Penker, M. (1998). UML Toolkit - Domain Analysis, (pp. 247 & 324 - 325).
John Wiley & Sons, Inc.
[8] Beebe, J. (2018). Global Dispatcher Service - UML Model. Retrieved from:
[Accessed February 17, 2018]
[9] CIBSE. (2015). CIBSE Guide D. 2015 Transportation Systems in Buildings. The Chartered
Institution of Building Services Engineers.
[10] Beebe, J. "Standard Elevator Information Schema",
[Accessed February 17, 2016].
[11] Beebe, J. (2018). Global Dispatcher - Use Cases., From: Jonathan Beebe - Global
[Accessed February 17, 2018]
[12] thyssenkrupp Elevators Corporate Website. MULTI., From:
[Accessed January 23, 2018]
[13] Gerstenmeyer, S., & Peters, R. (2016). Multicar Dispatching. 6th Symposium on Lift &
Escalator Technologies - Proceedings Vol 6 (pp. 8-1 - 8-12). Northampton: The CIBSE Lift
[14] Emre Oner Tartan, Cebrail Ciftlikli.( 2016) A Genetic Algorithm Based Elevator Dispatching
Method For Waiting Time Optimization; IFAC-PapersOnLine; Volume 49, Issue 3(pp. 424-429)
[15] Beebe, J. (2018). Global Dispatcher - Requirements Catalogue., From: Jonathan Beebe -
Global Dispatcher:
[Accessed February 17, 2018]
[16] Richards, M. (2015). Software Architecture Patterns. Chap. 2. Event-Driven Architecture,
O'Reilly Media, Inc.
[17] Wikipedia, CRUD definition,,_read,_update_and_delete, [Accessed May 26, 2018]
[18] Waher, Peter. (2018), Mastering Internet of Things, Chap 10 The Controller, Packt
Publishing Ltd. (,ISBN 978-1-78839-748-3.
... A previous paper [3] (hereafter referred to as "Paper1") analysed the requirements for a Global Dispatcher Interface (GDI) via which a group of lifts could be controlled. Current trends and possible future developments in lift group controller technology were reviewed so that the identified requirements are sufficiently broad and flexible to avoid the analysis becoming prematurely outdated. ...
Full-text available
This paper presents an overview of the design and development of a prototype Global Dispatcher Interface (GDI) for the control of a group of lifts. The role of the dispatcher is to assign passenger calls to the optimal lift in a group, as decided by a dispatcher algorithm. The GDI is independent of the underlying algorithm, which may be distributed remotely, and provides a standard means through which all interactions with the dispatcher may occur. To warrant the “Global” appellation the GDI must support any of the currently available, as well as anticipated, call station modes, types and configurations of cars, topology of control equipment and buildings. The design process is a continuation of a recognised Software Development Lifecycle, centred on Use Cases in a UML model, the initiation of which is covered in a previous paper. Significant diagrams from the model are presented and discussed to illustrate the evolution of the prototype design. One of the requirements, resulting from analysis of the Use Cases, identifies that the GDI design must be compatible with a publish-and-subscribe architecture and a RESTful interface is selected for this purpose. Where possible, the prototype design uses open standards with an emphasis on demonstrating those aspects that are specific to lift system dispatcher operation, while attempting to demonstrate independence from implementation details such as programming language, network protocols, etc. The Standard Elevator Information Schema is particularly relevant and fulfils these objectives. The operation of the working prototype, in conjunction with simulated lifts and passengers, is presented as a validation of the design.
... This example is discussed in more detail in a separate publication [45]. ...
Full-text available
A generational change is taking place in building transportation systems as manufacturers and maintenance companies begin to integrate their products and services with the technologies of smart buildings and smart cities. Frequently this integration relies on the Internet of Things and cloud services. The diverse and heterogeneous nature of such collaborations requires a common shared semantic understanding of the complex and dense information that may be generated by transportation systems in buildings. The Standard Elevator Information Schema (SEIS) provides this in a format which is both machine and human readable. The role of the schema is to provide the ‘vocabulary’ for these collaborations. At the same time the schema specifies the properties, relationships and validation rules that define the information model, which could form the foundation upon which all elements of building transportation control and monitoring functions are constructed. SEIS is published under the Collective Commons licence and is free to download and incorporate into any product with the objective of reaching the broadest audience. This chapter discusses the origins and features of SEIS and provides a varied set of example applications. Consideration is also given to the issues of cyber security and data protection.
Full-text available
Elevator group control is one of the important issues in vertical transportation systems in buildings. For an efficient group control algorithm it is required to overcome some optimization problems. From the perspective of quality of service two fundamental parameters needed to be optimized are waiting time and journey time. Elevator group control algorithms can be developed based on soft computing methods used for optimization. In the last decade Genetic Algorithm(GA) has attracted researchers’ attention as the suitability for the encoding car dispatching problem. In this study we enhance a previous study suggested for optimization of waiting time. The proposed method reduces average waiting time and uses a simpler encoding approach which results in efficiency in terms of computational cost. We explicitly demonstrate the method on the referred scenario for the purpose of visualization and insight.
This second edition of this well-respected book covers all aspects of the traffic design and control of vertical transportation systems in buildings, making it an essential reference for vertical transportation engineers, other members of the design team, and researchers. The book introduces the basic principles of circulation, outlines traffic design methods and examines and analyses traffic control using worked examples and case studies to illustrate key points. The latest analysis techniques are set out, and the book is up-to-date with current technology. A unique and well-established book, this much-needed new edition features extensive updates to technology and practice, drawing on the latest international research.
  • Kone Corporate Website
KONE Corporate Website, "KONE DESTINATION SOLUTIONS", 19/02/2016, [Accessed February14, 2018].
UML Toolkit -Domain Analysis
  • H.-E Eriksson
  • M Penker
Eriksson, H.-E., & Penker, M. (1998). UML Toolkit -Domain Analysis, (pp. 247 & 324 -325). John Wiley & Sons, Inc.
Global Dispatcher Service -UML Model. Retrieved from:
  • J Beebe
Beebe, J. (2018). Global Dispatcher Service -UML Model. Retrieved from: [Accessed February 17, 2018]
Standard Elevator Information Schema
  • J Beebe
Beebe, J. "Standard Elevator Information Schema", [Accessed February 17, 2016].
Global Dispatcher -Use Cases
  • J Beebe
Beebe, J. (2018). Global Dispatcher -Use Cases., From: Jonathan Beebe -Global Dispatcher: [Accessed February 17, 2018]
From: thyssenkrupp-elevator
  • Multi Thyssenkrupp Elevators Corporate Website
thyssenkrupp Elevators Corporate Website. MULTI., From: [Accessed January 23, 2018]
Global Dispatcher -Requirements Catalogue
  • J Beebe
Beebe, J. (2018). Global Dispatcher -Requirements Catalogue., From: Jonathan Beebe -Global Dispatcher: [Accessed February 17, 2018]
Software Architecture Patterns. Chap. 2. Event-Driven Architecture
  • M Richards
Richards, M. (2015). Software Architecture Patterns. Chap. 2. Event-Driven Architecture, O'Reilly Media, Inc.