Lift Report 2016; 42(2): 59-68 Introduction to Elevator Group Control (METE XI) Lutfi Al-Sharif Professor, Mechatronics Engineering Department, The University of Jordan, Amman 11942, Jordan INTRODUCTION AND OVERVIEW OF ELEVATOR GROUP CONTROL When elevators are placed in the same group, they will respond to the internal car calls as normal, but the landing calls are allocated to the most suitable elevator The algorithm for allocating the landing calls to an elevator is based on minimising a certain parameter (e.g., waiting time, travelling time, energy consumption) A dedicated group controller is used in order to allocate the landing calls to the individual elevators in the group 1.1 Definition of Elevator Group Control Elevator group control is the algorithm that is used in order to assign/allocate the landing calls to the various cars in the elevator group in order to minimise a certain cost function As the group controller receives more information, it can make better decisions The pieces of information that can be made available (in addition to the new landing call to be allocated) are, in order of importance: Position of all cars Direction of each car Items and above are essential as a minimum to allow a sensible allocation The type of the prevailing traffic pattern (e.g., up peak traffic, down peak traffic… ) Some elevator group control algorithms use the type of traffic pattern prevailing in order to alter the group control algorithm The following three types of data are generally easy to communicate to the group controller (usually via a serial bus connection) and not require expensive extra hardware Loading in each car (number of passengers in each car) Car calls registered in each car (i.e., the destinations of the passengers already in the cars) Landing calls already allocated to various cars © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 The following two pieces of data require expensive hardware The data they provide can make a significant improvement to the group control algorithm The number of passenger waiting behind a landing call This can be done by the use of cameras fitted in the landing that use automatic passenger counting software (using, for example, image processing techniques) The destination of all passengers waiting in the lobby This is usually achieved by the use of destination control systems, where passengers register their destinations in the lobby before boarding the elevator It is worth noting that the use of a destination control system provides both pieces of data in and above (i.e., the number of the landing call passengers as well as their destination) So even if two passengers are heading to the same floor, they should enter their request twice in order to the let the system know that two passengers are waiting 1.2 Incoming (Up-peak) Elevator Group Control Algorithms The following is a high level description of the types of elevator group control algorithms that are used under incoming traffic conditions (up-peak) Static sectoring, static allocation, where sector sizes are fixed, and elevator allocations to sectors are fixed A good example is a 10 floor building in which the building is divided into two sectors: to and to 10 In the up peak period, elevators served the lower sector (1 to 5) and the other elevators serve the upper sector (6 to 10) This is an example of static sectoring and static allocation It is called static sectoring because the size and floor allocation to each sector not change; it is called static allocation because the elevator allocation to the sectors are also fixed Static sectoring, dynamic allocation is an algorithm in which the sectors are fixed, but elevators are dynamically allocated to the sectors each time Otis Channelling is a good example of this up-peak elevator traffic control algorithm Dynamic sectoring and dynamic allocation: In this algorithm, the size of the sectors changes continuously and the allocation of the elevators to sectors is also dynamically allocated An extreme case of this algorithm is the destination control system, where the sectors are dynamic and continuously changing and the elevators allocated to them are continuously changing Moreover, in the destination control system, the floors within each sector need not be contiguous 1.3 Building Zoning In order to reduce the number of stops and the passenger travelling time and in order to save space in a building zoning is used When a building is zoned, then a group of elevators permanently serves a lower zone of floors and another group of elevators serves a different zone of floors Zoning is fixed into the structure of building (i.e., elevator will physically serve a zone and cannot serve other floors as they not © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 have any doors opening onto that floor) For an upper zone, the elevators from the lower sub-zone cannot serve it as their machine room stops at the mid-point in the building A transfer floor is usually used by which the lower zone serves this floor as well as the upper zone This is usually the highest floor in the lower zone or the lowest floor in the higher zone From the above, it can be seen that zoning is fixed by hard arrangements Sectoring is flexible and variable (e.g., can be changed by change of software, wiring and extra fixtures) 1.4 Advanced Techniques in Elevator Traffic Analysis Under general traffic conditions, a number of different elevator group control algorithms can be used In general, all of these algorithms attempt to minimise a certain cost function Some of the cost functions that are minimised: Waiting time Total journey time (waiting time and travelling time) Some of these systems are also applied to destination control systems, whereby the destinations of the passenger are also known This leads to a better allocation and much more efficient elevator group control system These systems are very efficient when applied to double-decker elevator group control systems In some cases, the cost function could be a combination of different parameters, with different weights given to different parameters For example, the cost function could be a combination of waiting time, travelling time and energy consumed, whereby more weighting is given to the waiting time and little weight is given to the energy consumed These weightings could even be varied depending on the time of the day Some systems employ advanced techniques to allocate the calls, such as fuzzy logic and genetic algorithms The rest of this article discusses in more details the concepts that were briefly introduced in this article PURPOSE OF GROUP CONTROL The function of efficiently distributing landing calls to individual elevator cars in a group is basically the same for both large and small groups Thus, a two-car elevator group can benefit from the use of group control as much as an eight-car group This is called group control The purpose of group control is to allocate (or assign) the landing calls in an optimum way to the various individual elevators in the group When there is only one elevator, then there is no need for a group controller The word optimum above is a very difficult word to define A number of definitions have been suggested as to what is optimised For example, optimising could mean any one of the following: • minimises passenger waiting time • minimise system response time (system response time being the time between the registration of the call until it is answered; this will be equal to the waiting time of the first passenger, i.e., the passengers who registered the call) © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 • minimises passenger journey time • reduce bunching • minimise the variation (or the variance in statistical terms) in passenger waiting time (or system response time) To achieve any one of the above various algorithms have been developed What makes these algorithms even more complicated is the fact that some algorithms are more suitable than others under different types of traffic For all of these algorithms to work they need to receive information about the elevator system and the traffic it is subjected to For example, the data which any group controller will need are: • All the landing calls • Position of all elevator cars • The status of each elevator car (moving up, moving down, stationary, out of service) Further improvement to the performance of the system is achieved if the following variables are also fed to the group controller: • All the car calls registered in the elevators • The type of prevailing traffic (e.g., up peak, down peak, inter-traffic ) • In some cases the destination of each passenger prior to boarding the elevator (as in Hall Call Allocation systems, to be discussed later in this article) As a general rule, the more information the group controller knows about the elevator system, the better the performance of the group controller in allocating the landing calls, and optimising the relevant parameter (e.g., passenger waiting time ) It is also worth mentioning that due to the fact that the up peak traffic has usually been the most stretching type of traffic for elevator systems, most the traditional algorithms are built around that type of traffic Moreover, a lot of the terminology and the methodology used in elevator design, even to this day, relies heavily on the concept of meeting a heavy up-peak influx in five minutes, by circulating elevators at the main terminal, delivering the passengers, and expressing back again to the main terminal EVALUATION OF THE GROUP CONTROL PERFORMANCE Several measures have been put forward as a method of determining the performance of an elevator system This is a measure of two combined variables: The performance of the group controller in assigning the landing calls, and the performance of the individual elevators in answering their assigned calls One of the important methods of measurement is the use of Average Waiting Time (AWT) The definition of Bunching in the Elevator Micropedia (Barney) is as follows: “A traffic pattern, where a number of elevators move around a building together, instead of being separated about the building, often caused by a sudden heavy traffic demand or an inadequate traffic supervisory system.” © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 versus System Loading INT in the discussion below refers to the Interval (i.e., the time between elevator arrivals at the main terminal) Figure shows a typical curve describing the relationship between the loading factor (some authors use the more generic term, "system utilisation", [8], pp 21), and an index of performance The ratio of AWT to INT has been proposed as a measure of performance of the elevator system ([8], pp 201) The loading factor is expressed as the ratio between the number of passengers and the rated capacity The index of performance is taken as the ratio between the Average Waiting Time (AWT) and the Interval (INT), where division by the Interval is a method of normalising the measure of performance This performance curve has been derived for a number of the traditional algorithms [8] Figure 1: Relationship between System loading and AWT/INT, with a straight extrapolation of the linear part below 50% loading The curve is linear in the region of loading 0% to 50% As the system loading increases above 50%, the curve departs from linear behaviour, and increases rapidly The linear relationship existing below 50% loading can be extrapolated as shown in Figure by the dotted line So it is possible to draw a straight line as a continuation of the linear relationship which exists at loads below 50% It has been proposed by the late Dr Schroeder [24] that the departure from this linear behaviour at loads of more than 50% is caused by the phenomenon of bunching in the elevator system (discussed later in the article) Schroeder further proposes that the measure of bunching is the ratio between two quantities The numerator is the actual AWT is at a certain system loading The denominator is the value the AWT would have been if the linear relationship applied above 50% system loading This is given the name Bunching Factor, bx [4] CONTROL ALGORITHMS Elevator group control systems respond to the necessity of providing efficient control of a group of automatic elevators servicing a common set of landing calls The goal is to provide the maximum handling capacity and the minimum waiting and travelling time of passengers while using the most economical installation The criteria are to arrange © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 the best arrangement of landing calls and to allocate the best elevator car to serve the particular landing call Various algorithms have been developed over time 4.1 Up-Peak Algorithms All of the conventional algorithms attempt to meet the up-peak traffic situation, as the most demanding on the elevator system resources With the advent of flexitime working practices, and the introduction of restaurants at the top or middle of buildings, the most stretching traffic pattern is now not necessarily the up peak traffic pattern 4.1.1 Conventional up-peak control Until recently, this has been the only available method It simply works as follows When an up peak traffic pattern is detected, all cars are sent to the main terminal The cars are then dispatched one after the other as they fill up The car will deliver the passengers to their floors and then return express to the main terminal, usually ignoring the down calls, or servicing them occasionally [9] In some buildings, the up peak is switched in via a timer, or manually by an attendant In more advanced systems, the system itself detects an up-peak condition by measuring the number of up calls from the main terminal, or the number of passengers in the cars (using load weighing devices) 4.1.2 Up peak sub-zoning The principle of this method is to try to reduce the number of stops an elevator car during a round trip by allowing certain elevators to serve specific zones in the building during the up peak period For example, elevators 1, and will serve floors to 5, while elevators 4, and will serve floors to 10 In this configuration, the elevator assignment and the zone assignments are fixed When the up peak period finishes, the elevators revert to the normal mode of operation, under which all the elevators serve all floors 4.1.3 Up-peak Dynamic Zoning (Otis “Channelling”) Channelling2 was introduced in the 90’s to alleviate lobby congestion during heavy uppeak traffic periods It is similar to the up-peak zoning method, with the difference that the elevator assignments to various zone is not fixed Elevators make fewer stops per round trip, with the result that cars return to the lobby faster Cars are assigned to groups of contiguous floors while the main terminal is often referred to as Floor As an elevator returns to the main terminal during an up-peak period, it is assigned to service one of the groups Passengers can then easily determine which floors each car is serving by checking the information display system screens located next to or above each elevator entrance The advantages of channelling are summarised below: • • • • • • Particularly effective under heavily demanding up-peak period or when some cars are out-of-service Cars usually load all of the passengers waiting to travel to the assigned floors of a group Average service time reduced Average round trip time reduced High flexibility of operation as the system is only activated when needed Mode activation can either be local or remote Referred to as up-peak sectoring by Barney 1992 © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 There are disadvantages and imperfection in channelling, summarised below: • • • • Average waiting time increased under pure up-peak mode of traffic Low flexibility of sectoring as floor assignment has been fixed during design stage System not adaptable to the ever changing traffic patterns System not designed for down-peak condition The idea of channelling was further developed to provide fully dynamical channelling for up-peak, down-peak and inter-floor traffic conditions A new term, dynamic sectoring, has been adopted This system makes use of known traffic patterns from other means, such as statistical, electronic or artificial intelligence, so that the optimisation of a cost function is carried out During up-peak and down-peak conditions, the overall round-trip time and handling capacity of each elevator car are optimised in order to produce the best assignment of elevator cars to different sectors The assignment is totally dynamic as it can be changed from time to time based on the real-time traffic patterns Under inter-floor traffic conditions, the probability function that represents the possible demand from one floor to another floor is used for carrying out the optimisation on another useful cost function, the equivalent round trip time A solution of the best floor assignment to different sectors can be obtained so that the information is displayed on screens at all landings During a specific period of time, the cars service the assigned sectors When the traffic patterns change, another sectoring arrangement will be implemented so that the best performance can be achieved 4.2 Down Peak Group Control Similar to the conventional up peak algorithm, in this mode the elevator system will only answer the down calls, while only one elevator in the group is left to answer the up calls 4.3 General Group Control Algorithms Apart from the up and down peak algorithms, several traditional and modern algorithms have been developed to deal simultaneously with all types of traffic 4.3.1 Static sectoring This method is quite simple to implement, and has been a widely used method especially in relay controlled installations Static sectoring divides the building into a number of sectors, and assigns one car to each sector That car will always go and park inside the sector, and when a landing call is registered from that sector, then the assigned car is dispatched there If the car is delayed or out of service, another car is assigned and sent As is obvious from the above, this method deals poorly with traffic which is concentrated in one zone It is probably more suitable for lightly used buildings with inter-floor traffic The parking aspect of this system is wasteful of energy, owing to unnecessary elevator movement and changes in floor usage This results in poor service and long waiting times at some floors It may be better to define the main terminal only as a parking floor and leave all other elevators at their last stopping floor 4.3.2 Dynamic sectoring © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 Dynamic sectoring is similar to static sectoring, but the sectors, as the name suggests, change in relation to both the position and direction of each car Figure shows how the sectors are allocated to each car, depending on its position and direction The “nearest” car is allocated the call This algorithm deals poorly with a number of consecutive calls in the same direction, which will all be allocated to the same car, with the other cars remaining idle The implementation of this system is usually referred to in the jargon as the “ring selector”, and used to be implemented historically using relays G Figure 2: Example to show dynamic sectoring 4.3.3 Estimated Time of Arrival This computer controlled method works by calculating a time function which represents the best estimate of when each car can respond to a landing call, and assigning that call to the car with the lowest estimated time of arrival Although this is an algorithm in its own right, the concept has been applied in many other computer controlled systems 4.3.4 Hall Call Allocation Systems As mentioned previously, the more a group controller knows about the prevailing traffic, the better an allocation it can make This concept has been taken to its extreme in this system, where the group controller knows the destination of each passenger before he/she boards the elevator Instead of the usual two up/down buttons, a panel of touch buttons are required at every landing, with a button for every possible destination floor Thus, a passenger arriving at a particular floor and wishing to travel to another floor presses the button of the destination floor The passenger then receives a message indicating which car has been allocated to this call The algorithm obviously requires that every passenger register his/her call Car buttons are not necessary and an indicator inside the car shows the floors at which the car is stopping © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 Every time a new call is registered, the computer allocates it in turn to each of the cars and evaluates the cost of each allocation The allocation giving the lowest cost is then adopted Suitable cost functions are passenger average waiting time, passenger average journey time or a combination of both The advantages of the system are: • • • • • Passengers not need to translate their intention to travel to a specific floor into a request for an up or down command Passengers not need to rush to the elevator whose hall lantern is on as they just go to the pre-assigned elevator after pressing the landing call button The supervisory system receives full information regarding the destinations of all passengers and thus it can make more intelligent decisions Handling capacity increases by about 35% to 60% The time to destination can be reduced by about 25% to 40% The disadvantages of the system are: • • • If the lobby is crowded, the passengers may not have time to get to the preassigned elevator, thus missing their elevator and having to re-register their call There may be a long delay after an assignment of car is made and the arrival of the car to service the passenger No flexibility is allowed during this period of time (i.e., the assignment cannot be altered) The system is effective for up-peak traffic and it has not been reported that it works equally well for down-peak traffic or inter-floor traffic USE OF ARTIFICIAL INTELLIGENCE IN GROUP CONTROL There are many ways to define the field of Artificial Intelligence (AI) One possible definition is the study of the computations that make it possible for the computer to perceive, reason and act From the perspective of this definition, AI differs from most of psychology because of the greater emphasis on computation and AI differs from most of computer science because of the emphasis on perception, reasoning and action The engineering goal of AI is to solve real-world problems using AI as an armamentarium of ideas about representing knowledge, using knowledge and assembling systems The scientific goal of AI is to determine which ideas about representing knowledge, using knowledge, and assembling systems explain various sorts of intelligence There are numerous applications of AI in elevator engineering A number of areas in AI have been applied in elevators, as follows: 1- Expert System Control: The philosophy of supervisory control based on traffic sensing and rule-based expert systems was developed in 1992 The system was implemented using standard packages, built on a spreadsheet in the first instance Simulated input traffic was generated and dynamically linked to the simulator, showing car movements An expert system linked to the traffic sensing system continuously calculated optimal car movements 2- Fuzzy control: The application of fuzzy logic in elevator systems was first implemented in Japan where the appropriate rule was selected right after any hall call button was pressed A fuzzy logic dispatching system reduces waiting time by operating in an active mode The dispatcher uses fuzzy rules based on past experience to predict how many people will be waiting for elevators at various © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page of 17 Lift Report 2016; 42(2): 59-68 times of the day rather than simply reacting to calls When several fuzzy features are included in the dispatching decision process, the result is a more effective approach to elevator dispatching than systems based on conventional digital logic 3- Artificial Neural network control: Artificial neural networks have been used to select the appropriate traffic patterns so that the traffic control module could choose the best hall call assignment algorithm A destination-oriented car allocation service has also been developed to improve services during rush hours 4- Optimal Variance Method: A statistical approach involving variance analysis has been adopted where the variance of hall call response time could be decreased by computerised elevator dispatch systems utilising cost function minimisation The idea is to improve the variance performance by sacrificing the mean response time METHODS OF DETECTING TRAFFIC PATTERNS AND PEAK TRAFFIC Until the 1980s, office working hours were relatively stable Incoming and outgoing traffic peaks could be predicted and simple time clocks used to switch the mode of operation of the group control The installation of analogue computer circuits to measure the numbers and direction of landing calls provided additional discrimination Changes in working practices to more flexible and staggered office hours defeat these simple strategies for handling peak traffic To compound the problem, building population densities have often increased beyond the original designed capacity of the elevators using non-computer-based systems For these reasons, it has become necessary to enable the controller to detect a type of traffic when it happens The behaviour of an elevator system is actually highly non-linear and it is very difficult to predict the status of an elevator system, in general, even minutes prior to it taking place For up-peak and down-peak conditions, it is comparatively easier to get closer to the real situation since a complete round trip can be clearly defined under these circumstances The difficulties to model are that the operation is based on the passengers' demand while the demand is purely a random process Furthermore, it is at present quite impossible to describe, in details, the operation of an elevator system on a real-time basis by means of simple equations Hence, real-time simulation seems to be the most effective means of testing new control algorithms while macroscopic modelling can help during the design stages More research inputs must be elaborated to build up more accurate models for the real-time operation of elevator systems Supervisory control algorithms to handle different types of peak traffic are available and proved to be reasonably effective One key point is to identify the occurrence of a particular type of peak traffic Manual monitoring is one way but it is not economical Different methods of detecting traffic patterns are deemed necessary so that the appropriate supervisory control algorithm can be called upon to handle that particular type of traffic a- The load weighing device: In most cases these devices give the estimated weight in discrete steps (steps of 20%, 25% or even 50%) of full load, and thus only give a rough estimate of the number of passengers The conversion process from absolute weights to passenger numbers assumes a fixed weight figure per passenger, which could vary widely, due to largely differing weights, different © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 10 of 17 Lift Report 2016; 42(2): 59-68 countries and other effects (e.g., passengers carrying objects, or pushing trolleys etc.) b- Using Photocell signals: This method is used to identify the number of passengers leaving or entering the car ([25], [26]) In cases where the elevator responds to both a landing call and a car call, it is not clear how a single photocell could distinguish between in-going and out-going passengers, unless two displaced photocells are used and the two resulting signals are combined to derive the number of transferring passengers Moreover, this system could give erroneous results if the door width is such that more than one passenger can board or alight at the same time c- Using a sensitive pad on the floor of the car which identifies the number of passengers from the shape of the footprints [14] d- Imaging systems, which use artificial intelligence techniques to identify the number of passengers e- An inverse formula based on the relationship between S (number of stops) and P (number of passengers) can be inverted and used to derive the probable number of passengers from the elevator activity This method has been called the I-S-P method (inverse S-P) The I-S-P method can be used in conjunction with the centred moving average to filter out disturbances in the derived traffic patterns, and statistical techniques can be used to derive upper bounding and lower bounding curves [3] f- Neural networks: Artificial neural networks have been employed to identify five types of common traffic patterns, namely up-peak, down-peak, off-peak, one-way and two-way The networks are under supervised training mode, being tested with the high-rise zone of a real commercial building Once the type of traffic pattern has been identified, the appropriate control algorithm can be called in to perform the most efficient service Another approach monitors the number of passengers entering and leaving the car by sensing the changes in car loading These numbers are fed into a neural network to identify different types of traffic patterns Continuous training of the neural network is carried out and for each distinguished traffic pattern, the corresponding operation is executed BUNCHING To explain bunching, it is best to take an example from buses in real life If someone has waited for 30 minutes for the bus to arrive, then four buses arrived at the same time, then the passenger has waited four times the optimum time This real life example taken from buses is similar to the situation which takes place in elevators The ideal situation in elevator traffic control is to keep the elevators in the group as far apart as possible, when they are circulating in up-peak This is the ideal situation in which the elevators should be in order to achieve the optimum performance of the elevator system However, if the elevators are not spaced equally apart in the building, the performance of the system starts to suffer, and manifests itself in increased waiting time Many of the early control system have tried to avoid that by dispatching the elevators from the main terminal at fixed intervals of time, in order to keep the elevators equally apart during their trips ([8], pp 37), but this will start to affect the handling capacity of the elevator system © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 11 of 17 Lift Report 2016; 42(2): 59-68 As seen from the above, bunching in elevator systems occurs when the intervals between elevator arrivals at main terminals vary widely, and this increases the waiting time of the passengers Very often, the rapid increase of average waiting time at loads above 50% is caused by the effects of bunching A typical case of bunching can be seen in elevator systems when the elevators start following each other (or even leapfrogging), as they answer adjacent calls in the same direction This has a detrimental effect on passenger waiting time and the ultimate case is when all the elevators in the group move together, acting effectively as one huge elevator with a capacity equal to the sum of the capacities of all the elevators in the group 6 3 2 Long wa iting G G Figure 3: Two diagrams showing an ideal situation and an extreme bunching situation One way to identify is to find out the time in seconds between the departure of a car numbered i and the next car numbered (i+1) as t i,i+1 Then, the difference between this time and an ideal time, say the desirable round trip time divided by the number of cars of the whole system for instance, can be an effective measure of how much bunching exists in the elevator system Two bunching coefficients can then be evaluated, namely the first order and the second order while the second order places a heavier penalty on large deviations from the reference [4] The concept of bunching has provided the idea for the so-called “Adaptive Round Trip Time” algorithm During real operation, the round trip time of each car can deviate significantly from the desirable figure, thus giving large first order and second order bunching coefficients when bunching is actually not very serious In this case, real-time condition monitoring is necessary to update the round trip time continuously The concept is to record the total time taken of each round trip for a car under either an uppeak or down-peak traffic condition This time value is used to finely adjust the average round trip time by using a proportional-integral-differential algorithm, hence the term adaptive round trip time © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 12 of 17 Lift Report 2016; 42(2): 59-68 COMPUTER VISION BASED CONTROL Conventional supervisory control dispatches elevator cars based on the landing call registration where the number of passengers demanding services cannot be established Passenger counting is a very important feature in improving the performance of a supervisory control system At present, loading devices, photocells, sensitive pads and even Inverse S-P method have been used to find out the number of passengers but their level of accuracy is not satisfactory Computer vision is a combination of image processing and image understanding The first generation of computer vision based application counts the number of passengers based on the cross-sectional areas of each passenger with the aid of appropriate fuzzy membership functions Cameras needed to be installed just below the ceiling vertically above the passengers The algorithm has later been improved so that the constraint of camera position can be removed by introducing two new techniques, namely algebraic reconstruction and optical flow METHODS OF IMPLEMENTATION 9.1 Use of Computers Computers allow raw data to be collected from the system for processing to enable sophisticated approaches to group control This allows fast, flexible and adaptive responses to traffic situations in the building Correctly designed algorithms can adapt to changes in the use of floors, the building and the population without the need for modifications by the manufacturer The computer can process the following information which can be made available via inputs in order to provide optimal group control for the elevators Operational information for each elevator, as follows: • • • • • • • • • operational mode (in or out of the group) running state (moving or stationary) position planned or actual direction of movement if stationary, door status (opening, open, closing, closed) number of car calls car load rated speed of the elevator (some manufacturers can mix elevator speeds in the group) current system response time to each landing call (time to travel to that landing) Landing information, as follows: • • • number and direction of landing calls waiting time of each landing call computed traffic intensity at each floor Group information, as follows: © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 13 of 17 Lift Report 2016; 42(2): 59-68 • • measured maximum landing call waiting times measured average landing call waiting times 9.2 Centralised and Distributed Control The landing calls are registered at the landing and transferred to the elevator system As there are usually a number of elevators in the group, there are a number of methods by which the calls are collected and processed: 9.2.1 A dedicated group controller This system employs a dedicated controller which collects and registers the calls and allocates them to the required elevator, according to a certain algorithm The problem of this method is that the group controller needs extra space, and if it fails, it jeopardises the whole system Group controller Landing Calls Lift Lift Lift Lift Controller Controller Controller Controller Figure 4: Dedicated group controller 9.2.2 A soft master-slave configuration In this configuration, all elevator controllers receive the landing calls All elevators also communicate with each other via serial links One elevator assumes the master-ship of all others, and develops a scoring table to decide to which elevator each call is assigned If that elevator goes out of service, another elevator will assume master-ship, and take over the call allocation © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 14 of 17 Lift Report 2016; 42(2): 59-68 Lift Controller Lift Controller Serial link Landing Calls Lift Controller Lift Controller Figure 5: The soft master-slave configuration 10 USE OF DATA LOGGING METHODS FOR EVALUATING ACTUAL PERFORMANCE Data logging is essential in facilitating routine maintenance as well as a verification and basis of improvement to a design On-line data logging can provide useful information for the intelligent control of operations Before any control algorithm is executed, adequate information reflecting the current status of every car within the system must be retrieved and this relies on an advanced digital monitoring system A data-network system has been developed in which microprocessor modules are placed inside the machine room, inside each car and at each landing lobby All these modules are linked together by a serial-transmission network Network distributed design could increase both system sophistication and performance A similar concept utilising the concept of network layer protocol for elevator communication have also been developed The advantages of this approach are the flexibility to use any types of protocol in any layer and that once it is designed the network program module is applicable to any node in an elevator system In addition to remote status monitoring, a diagnostic system has been developed for preventive maintenance The system can pick up hidden failure symptoms which could not be noticed even by a skilled maintenance engineer because it has been found that even such hard-to-detect abnormalities in some equipment might cause serious trouble when amplified by factors such as wear and deterioration 11 GROUP CONTROL BACKUP With computer-based systems it is easy to provide backup to normal operation to accommodate failures in the controller which could otherwise cause complete loss of elevator service The first level of group control backup, however, should not be a 'bus stop service' in which landing calls are ignored and the elevators move continuously between floors, stopping at each floor to pick up any waiting passengers Backup service of this kind is electrically inefficient and gives very poor elevator service to the building Using high-speed communications channels between computers, redundancy can be built into a group control to provide backup Many approaches are possible Two computers connected together can be used to control a group Either computer can control the group, but only one runs the group control at any one time The © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 15 of 17 Lift Report 2016; 42(2): 59-68 second computer monitors the operation of the first Should the first computer fail, the second reports the failure and takes over the group control Computers capable of providing both single elevator control and group control can be used For example, on a four-car group this means that there are four computers able to run the group controller If the computer currently running the group control fails, one of the other computers automatically takes over Even if all the group control programs were out of action the elevator control could still perform a basic collective operation in response to the calls because the landing calls are registered by each elevator computer In practice, this means that group control reliability is very close to 100% REFERENCES AND BIBLIOGRAPHY [1] Alani A.F., Mehta P., Stonham J and Prowse R "Performance optimisation of knowledge-based elevator group supervisory control system", Elevator Technology 6, IAEE, 1995, pp 114-121 [2] Al-Sharif L and Barney G C Bunching Factors in Lift System Control Systems Centre Report, UMIST, No 749 and No 754, 1992 [3] Al-Sharif, L New concepts in lift traffic analysis, The inverse S-P method Elevator Technology 1992; 4:8-17 The International Association of Elevator Engineers [4] Al-Sharif, L Predictive methods in lift traffic analysis Ph.D Thesis, UMIST, Control Systems Centre, 1992 [5] Amano M., Yamazaki M and Ikejima H "The latest elevator group supervisory control system", Elevator Technology 6, IAEE, 1995, pp 88-95 [6] Aoki K "Information networks in elevator systems", Elevator Technology 4, IAEE, 1992, pp 18-27 [7] Bailey S and Clarke T "Monitoring systems and their role in lift maintenance", Elevator World, May, 1994, pp 58-63 [8] Barney G.C and Dos Santos S.M Elevator Traffic Analysis, Design and Control, IEE, Peter Peregrinus, London, 1985 [9] Barney, G.C., 1992, “Uppeak revisited”, p39, Proceedings of Elevcon ‘92, Elevator Technology 4, IAEE, 1992 [10] Chan W.L and So A.T.P "Dynamic zoning for intelligent supervisory control", International Journal of Elevator Engineering, Vol 1, 1996, pp 47-59 [11] Chan W.L and So A.T.P "Dynamic zoning in elevator traffic control", Elevator Technology 6, IAEE, 1995, pp 132-140 [12] Chenais P and Weinberger K "New approach in the development of elevator group control algorithms", Elevator Technology 4, IAEE, 1992, pp 48-57 [13] Halpern J.B "Statistical analysis of modern elevator dispatch hall call response times", Elevator Technology 6, IAEE, 1995, pp 96-103 [14] Haraguchi, H., 1991, "Detecting System", Patent No 4951786, 28/8/1990, Elevator World, March 1991 [15] Iwata S., Terazono N and Suzuki S "Data network-based elevator control system", Elevator 4, IAEE, 1992, pp 266-275 [16] Kawano S., Yuminaka T and Kobayashi N "Remote and intelligent diagnostic system for elevators incorporating preventive maintenance", Elevator World, May, 1994, pp 66-71 [17] Kim C.B., Seong K.A., Lee-Kwang H., Kim J.O and Lim Y.B "A fuzzy approach to elevator group control system", IEEE Transactions on Systems, Man and Cybernetics, Vol 25, No 6, 1995, pp 985-990 © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 16 of 17 Lift Report 2016; 42(2): 59-68 [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] Kubo S., Nakai S., Imasaki N and Yoshitsugu T Elevator group control system with a fuzzy neural network model, Elevator Technology 6, IAEE, 1995, pp 1120 Pang G.K.H "Elevator scheduling system using blackboard architecture", IEE Proceedings D, Vol 138, No 4, 1991, pp 337-346 Powell B.A "Important issues in up peak traffic handling", Elevator Technology 4, IAEE, 1992, pp 207-218 Powell B.A and Sirag D.J "Fuzzy logic", Elevator World, September, 1993 Prowse R.W., Thomson T and Howells D "Design and control of lift systems using expert systems and traffic sensing", Elevator Technology 4, IAEE, 1992, pp 219-226 Schroeder J "Advanced dispatching, destination hall calls plus instant car-tocall assignments: m10", Elevator World, March, 1990 Schroeder, J., 1990, "Elevator Traffic: Elevatoring, Simulation, Data recording The Data compatibility problem and its solution", Elevator World, June 1990 Siikonen M-L and Kaakinen M "Using artificial intelligence to improve passenger service quality", Elevator Technology 5, IAEE, 1993, pp 238-246 Siikonen, M.L., 1991, "Simulation - a Tool for Enhanced Elevator Bank Design", Kone Elevators Research Centre, Elevator World, April 1991 So A.T.P and Chan W.L "Comprehensive dynamic zoning algorithms", Elevator Technology 8, IAEE, 1997, pp 98-107 So A.T.P and Liu S.K "An overall review of advanced elevator technologies", Elevator World, Vol XLIV, No 6, 1996, pp 96-101 So A.T.P., Beebe J.R., Chan W.L and Liu S.K "An artificial neural-network based traffic patterns recognition system", International Journal of Elevator Engineering, Vol 1, 1996, pp 35-46 So A.T.P., Chan W.L., Kuok H.S and Liu S.K "A comprehensive solution for computer vision based supervisory control", Elevatori, Vol 23, No 4, 1994, pp 61-77 So A.T.P., Kuok H.S., Liu S.K., Chan W.S and Chow T.Y "New developments in elevator traffic analysis", Elevatori, Vol 23, No 6, 1994, pp 59-74 Umeda Y., Uetani K., Ujihara H and Tsuji S "Fuzzy theory and intelligent options", Elevator World, July, 1989 © Copyright held by the author 2016: Prof Lutfi Al-Sharif Page 17 of 17