This invention relates to apparatus for handling units of currency. Theinvention will be described mainly in the context of coin handling, but is alsoapplicable to apparatus which also or alternatively handles other units ofcurrency, such as banknotes, smart cards, payment cards or the like.
It is known to provide a coin handling apparatus which receives andvalidates coins of different denominations, and directs valid coins torespective containers each containing coins of a single denomination. It isalso known to dispense coins from these containers as change in an amountcorresponding to the difference between the value of inserted coins and theprice of a product or service obtained from a machine associated with the coinhandling apparatus.
It is also known to arrange for the level of coins in each container notto exceed a predetermined upper level. The apparatus would thus tend todirect coins of a particular denomination to an associated container until theupper level is reached, and then any further coins of the same denominationwould be sent to a cashbox, which would normally be of a type which doesnot permit the dispensing of coins therefrom.
Periodically, an operator will empty the cashbox. At this time, it iscommon for operators to adjust the levels of coins in the coin containers sothat each one will contain a number of coins corresponding to a so called"float" level for the respective container in an attempt to ensure that there isusually an adequate supply of coins in the container to be used as change ifthis is necessary.
In use, between visits by the operator, the levels of coins in thecontainers vary, and at some times there may not be sufficient coins in acontainer to supply the correct change for a given transaction, when themachine is said to be in "exact change only" mode. It is understood that when a vending machine only accepts the exact change for a selected item, sales arelost.
It is known from EP 0 653 084to attempt to minimise the risk that, in the periodfollowing servicing of the machine, there will be insufficient coins availablefor change by arranging for the apparatus to monitor certain parameters, tocalculate dynamically the likelihood that a particular denomination will berequired for dispensing.
US 5,356,333 relates to a coin escrow which is a store for temporarilystoring coins inserted into a coin-handling apparatus in a single transactionbefore they are either returned to the user of directed to a cashbox.
The invention provides a currency handling apparatus as set out inclaim 1 and a method of operating a currency handling apparatus as set out inclaim 33. Preferred features of the invention are set out in the dependentclaims.
An example of an apparatus in accordance with the invention will nowbe described with reference to the accompanying drawings, in which:
- Fig. 1 is a schematic diagram of the mechanical part of a coin handlingapparatus;
- Fig. 2 is a block diagram of the circuit of the coin handling apparatus;
- Fig. 3 is a diagram showing a removable cassette;
- Fig. 4 is a flow diagram;
- Fig. 5 is a diagram of a vending machine;
- Fig. 6 is a flow chart; and
- Fig. 7 is a flow chart.
- Fig. 5 shows a vending machine, indicated generally by the referencenumber 90. Referring to Fig. 1, the coin handling apparatus 2 in the vendingmachine includes a coin vaiidator 4 for receiving coins as indicated at 6.During the passage of the coins 6 along a path 8 in the validator 4, thevalidator provides signals indicating whether the coins are acceptable, and ifso the denomination of the coins. Various types of validators are known,including validators using optical, acoustic and inductive techniques.Examples of such validators are described in, amongst others, GB 1 397 083,GB 1 443 934, GB 2 254 948 and GB 2 094 008.
Acceptable coins then enter a coin separator 10, which has a numberof gates (not shown) controlled by the circuitry of the apparatus for selectivelydiverting the coins from a main path 12 into any of a number of further paths14, 15, 16 and 17, or allowing the coins to proceed along the path 12 to a path20 leading to a cashbox 21. If the coins are unacceptable, instead of enteringthe separator 10 they are led straight to a reject slot via a path 30.
Each of the paths 14, 15, 16 and 17 leads to a respective one of fourcoin tubes or containers 22, 24 and 26 and 28. Each of these containers isarranged to store a vertical stack of coins of a particular denomination.Although only four containers are shown, any number may be provided.
A dispenser indicated schematically at 29 is operable to dispense coinsfrom the containers when change is to be given by the apparatus. Thedispensed coins are delivered to a refund path 31.
As shown in Fig. 3, the tubes 22, 24, 26, 28 are provided in aremovable cassette 40, and the tubes themselves are removable from thecassette, as described in GB 2 246 897 A.
Referring to Fig. 2, the circuitincorporates a microprocessor 50 connected to data and addressbuses 52 and 54. Although separate buses are shown, data and address signalscould instead be multiplexed on a single bus. A bus for control signals couldalso be provided.
The microprocessor 50 is connected via the buses 52 and 54 to a read-onlymemory (ROM) 56 and a random access memory (RAM) 58. The ROM56 stores the program controlling the overall operation of the microprocessor50, and the RAM 58 is used by the microprocessor 50 as a scratch-padmemory.
The microprocessor 50, the ROM 56 and the RAM 58 are, in thepreferred embodiment, combined on a single integrated circuit.
The microprocessor 50 may also be connected via the buses 52 and 54to an EAROM 60 for storing a variety of alterable parameters.
The microprocessor 50 is also coupled via the buses 52 and 54 toinput/output circuitry indicated at 62. The circuitry 62 includes user-operableswitches, at least one level sensor for each of the coin containers 22, 24, 26,28, circuits for operating the dispenser 29 and the gates of the coin separator10, the circuitry of the coin validator 4, and a display visible to a user of theapparatus for displaying an accumulated credit value and an indication wheninsufficient coins are stored to guarantee that change will be available. Thecircuitry 62 is connected to a display 68 visible to the operator, and to akeypad 70 accessible only to the operator.
The input/output circuitry 62 also includes an interface between thecontrol circuit of the apparatus and a vending machine circuit board 64 towhich it is connected, and a further interface to an audit device 66.
In operation of the apparatus the microprocessor SO successively teststhe signals from the validator to determine whether a coin has been inserted inthe apparatus. When a credit has been accumulated, the microprocessor alsotests signals from the vending machine to determine whether a vending operation has been carried out. In response to various signals received by themicroprocessor 50, various parts of the program stored in the ROM 56 arecarried out. The microprocessor is thus arranged to operate and receivesignals from the level sensors of the coin containers 22, 24, 26, 28, and tocontrol the gates in the separator 10 in order to deliver the coins to therequired locations, and is also operable to cause appropriate information to beshown on the displays of the apparatus and to deliver signals to the vendingmachine to permit or prevent vending operations. The microprocessor is alsooperable to control the dispenser to deliver appropriate amounts of change.
The arrangement so far is quite conventional, and the details ofparticular structures suitable for using as various parts of the mechanism willtherefore not be described in detail.
The particular sequence of most of the operations carried out by themicroprocessor may be the same as in previous apparatus. A suitable programto be stored in the ROM 56 can therefore be designed by anyone familiar withthe art, and accordingly only the operations carried out by the particularlyrelevant parts of this program will be described.
The apparatus 2 is arranged to monitor each transaction performed bythe vending machine. More specifically, the microprocessor 50 registers, foreach transaction, the price of the item selected, the number and denominationsof the coins supplied as payment, and the number and denominations of coinssupplied in change. The processor 50 also registers when 'exact change only'events occur, and which coin tubes are empty and for how long. The numberand denomination of coins supplied as change is determined by a suitablechange algorithm in the microprocessor, as described, for example, in GB 2269 258. Theinformation registered by the processor 50 is stored in the audit device 66.
When the operator visits the machine, he causes the microprocessor 50to analyse data stored by the audit device 66 and calculate a desired float levelfor each denomination of coin by pressing appropriate keys on the keypad 70. Alternatively, the stored data can be downloaded to and processed by acomputer which the operator plugs into the processor 50 or the audit device66. Steps of the calculation are set out in Fig. 4. More specifically, themicroprocessor 50 calculates the total number of coins of that denominationthat have been inserted (step 72) and the total number of coins of thatdenomination (step 74) that have been dispensed in the previous seven days.The difference between the number of coins dispensed and inserted is thencalculated (step 76). That number is then multiplied by four, assuming thatthe operator visits the machine once a month, to arrive at an estimate for thedifference over a month, and hence an estimate for the float level. Themicroprocessor 50 then divides that estimated figure for the float level by themaximum number of coins that can be stored by a tube in the mechanism(step 78), rounds up the result (step 80) to the next whole number to arrive at anumber which represents the number of tubes of that denomination thatshould be supplied.
The microprocessor 50 performs similar calculations for eachdenomination of coin, and then decides what changes need to be made to thecoin tubes in the assembly at the time. A signal, representing instructionsregarding changes to the tubes, is generated and sent to the circuitry 62. Theinstructions are then displayed on the display 68.
For example, suppose over a period of 7 days, the coin mechanismdetects that 100 5p coins are inserted, and 130 5p coins are dispensed aschange. If the operator is to visit the machine once a month, using steps of thealgorithm set out above it is estimated a float of 120 5p coins should beavailable at the start. Assuming that a preformed tube stores at most 90 5pcoins, then it can be seen that at least two such tubes, each full of 90 5p coinsshould be present at each service to reduce the risk that an exact change onlymode will occur. Instead of supplying two full tubes, two tubes containing atotal of 120 or more coins could be provided.
In the example given above, the processor 50 is making predictionsabout the tubes to be required by predicting the likely distribution of coins tobe dispensed as change. This is done by monitoring the flow in and out ofeach denomination coin over a predetermined period of time, or, more simply,the difference between the number of coins of that denomination inserted andthe number dispensed. The number of tubes to be required is then estimatedusing the predetermined, known capacity of a tube.
The above example does not, for example, take account of fluctuationsthroughout the seven day monitoring period, such as periods when a sequenceof 5ps are dispensed as change, but none are inserted. More sophisticatedmethods of predicting a good float level and hence the number of tubesrequired are possible.
In an alternative approach, a rate of depletion for each coindenomination is calculated, whereRate (R) =(no. of coins dispensed - no. of coins accepted)capacity of coin tube(s) (taking into account only coins of the given denomination).
The rate is established over 50 vends, and is then continuouslyupdated, using a running sample of 50 vends in a sequence.
R tends to 0 with a slower depletion of coins in the coin tubes andtends to I with a faster depletion of coins. The calculated value R can be usedas an indicator for changes in the tubes. For example, if R > 0.80 for a givendenomination continuously for a certain number of vends, say 10, then themicroprocessor outputs a signal indicating that the number of coin tubes forthat denomination should be increased. Similarly, if R < 0.15, for example,the processor outputs a signal indicating that the number of tubes should bedecreased (the processor can be programmed to suppress such a signal if thenumber of tubes is one).
The information about the number and denomination of coins insertedinto the mechanism and monitored by the microprocessor 50 is useful on its own. For example, it can give an indication of changes in the relativepopulation levels of certain denominations of coins, or of the introduction ofnew coins into a coin set.
As discussed below, predictions for the number of tubes of eachdenomination to be provided can be made in otherways and using variables other than those in the examples above.
Information about the prices of items for sale is useful for makingaccurate predictions of change that will be necessary. Price information isespecially useful in such predictions when there is an alteration in price. Asan illustration only, suppose the vending machine is set up for use in the U.K.and comprises a single coin tube for each of 5p, 10p, 20p, and £1. Only oneitem is available for sale and its price is 95p. Using the information about theprice of the item for sale and of the acceptable currency set, the processor canwork out that, whatever coins are inserted, the only change that will ever bedispensed are 5ps and 10ps. On that basis, the processor issues instructions toremove the 20p, and £1 tubes and replace them with, for example, one extratube for 5ps, and one extra tube for 10ps, and direct all the remaining coinsstraight to the cash box.
Another useful parameter is the number and denominations of coinsdispensed as change. Such information can be used,for example, to request a new tube for coins of a higher denomination thanany already provided. Suppose the highest denomination of coin for which atube is provided is a 20p coin and transactions occur regularly (say more than5 times a day an average) where five or more 20p coins are dispensed aschange, then a £1 coin tube may be requested. Usually, it will be preferablealso to consider other parameters as well, such as number of denominations ofcoins inserted. In the above example, a £1 coin tube would be of limited useif few £1 coins are received because the tube would empty relatively rapidly.The change algorithm in use, which influences the likelihood that a coin willbe dispensed in change, may also be useful.
Other parameters, such as data indicative of the relative populationlevels of respective currency denominations in the area in which the apparatusis to be used, may be relevant. The relative population level of givendenominations are related to the probability that those denominations will beinserted into the machine. The relative population levels may vary within acountry, as. for example, in Southern Germany where there is a differentdistribution of coins from the rest of the country. Such data is known, and canbe supplied by a programming mechanism such as a computer, in the field, orit may be pre-programmed in, depending on the intended destination of themechanism. The mechanism might, for example, note where the use of smartcards and/or banknotes is on the increase and recommend changes in the cointubes as a result As mentioned above, a record is made regarding exactchange only' events. It is noted, for example, when exact change only eventsoccur, how long they last, and which tubes have been emptied (or have a levelprovided). Some or all of that information can also be used in predicting whatchange tubes should be provided.
While each of the parameters mentioned above may be used on itsown to provide an indication of changetubes to be required, it is usually advantageous to consider two or more ofthose parameters in combination.
In the present example, the change cassettes comprise removable tubes22, 24, 26, 28. The operator simply interchanges tubes of certaindenominations in accordance with the instructions on the display means 68.The operator uses the keypad 70 on the coin apparatus to inform themicroprocessor of the changes in the tubes in the mechanism, which alters thesettings in the microprocessor 50 accordingly so that coins are subsequentlydirected to and dispensed from the correct tube.
Other modifications to the apparatus described are also envisaged. Forexample, instead of using the information stored in the previous seven days,the microprocessor may analyse the information stored since the apparatus was installed. Alternatively, for example, the processor could be arranged tocarry out an analysis of the relevant information after a predetermined numberof transactions.
Instead of replacing certain tubes, the operator could replace the wholecassette, in accordance with suitable instructions from the apparatus, with acassette prepared with a predetermined arrangement of tubes. Cassettesprepared in such a way may be marked with a code, indicating the type anddistribution of tubes within the cassette, so that instructions for changing thetubes can be displayed using the appropriate code, and once the cassette isinserted, the code can be input via the keypad to inform the microprocessor ofwhich tubes are in use.
The prediction may be stored in the processor to be retrieved by theoperator at a later date. Alternatively, predictions may be obtained remotely,using known techniques, so that the operator knows in advance what tubes totake to the machine to service it. For example, the machine may transmitsignals down a telephone line, or mains line, to a computer in the servicingcentre. The transmitted signals may be data relating to the vend signals,which are then processed at the servicing centre to provide instructions aboutchanging the tubes, or signals representing the instructions. Data can also bedownloaded using a smart card and processed or read elsewhere.
The information regarding 'exact change only' events, that is recordedin the audit device has uses other than for predicting change tubes to berequired. As mentioned previously, an exact change only event is deemedequivalent to lost sales, and a record of such events is useful for analysts todetermine the occurrence and volume of lost sales. A log of exact changeonly events can also be useful in setting float levels, or for determiningwhether the operator needs to visit the machine more often to replenish thecoin tubes.
In the example given above, the rate at which the service operatorvisits the machine or the next date for a visit is predetermined and the visiting rate or date is used in working out the optimum number of stores for reducingthe risk of one or more tubes running out of coins. Alternatively, as describedin more detail below, the operator may visit the vending machine to replacethe change cassette or replenish the stores in response to instructions whichare output from the machine in advance, for example, instructions regarding adate on which a visit should be made.
In this example, the machine outputs instructions for the operator tovisit in accordance with a prediction of when one of the tubes is likely to runout of coins.
To make the prediction, the processor 50, automatically and once aweek, performs certain calculations, as explained below with reference to Fig.6, using information stored in the audit device 60.
First, the processor 50 calculates, for one of the denominations usedfor change, a rate of depletion r1 for the preceding week, as follows:r1 = b1 - a1where b1= no. of coins of the given denomination dispensed as change in thepreceding week, and a1 = no. of coins of the given denomination inserted aspayment in the preceding week.
The number of coins in the tubes for the relevant denomination at thetime at which the calculations are being carried out, N1, is divided by r1, andthe result rounded down to the nearest whole number, V1. The number V1 isan estimate of the number of weeks before the tube will empty.
The above calculations are repeated for each of the denominationsused for change, to result in a set of values Vi. The smallest non-negativevalue VS, which is the shortest time before one of the tubes is estimated to runout of coins, is selected.
The processor then informs the service centre, via a telephoneconnection, that the operator needs to visit the machine on a date in VS weekstime. On that visit, the operator replaces the change cassette with a newchange cassette, already containing coins at appropriate, predetermined levels.
Various modifications are alsoenvisaged. Instead of simply informing the operator of a date on which tovisit, the machine can also calculate and output the likely number of coins tobe held in each of the stores in VS weeks time so that the operator knows howmany coins of each denomination to bring to replenish all the stores. Thecalculations may be performed at other rates, for example, daily, or once amonth. Predictions can be made using other parameters, for example, thosedescribed above in relation to the first example, or using other calculations.Instead of calculating when the first tube is likely to empty, the processor mayinstead predict when a tube is likely to have a given number of coins, forexample, two. In the calculation of the date when the operator should visit, amargin of error of, say, a week, may be introduced (that is, the operator isinstructed to visit a week before the date on which the processor has estimatedthat the first tube will run out of coins). As before, the data could be suppliedremotely with the calculations being performed remote from the vendingmachine.
In another example, which is an embodiment of the claimed invention,instead of requesting a change to the tubes, the mechanism reallocates anexisting tube and uses it for a new denomination of coin.
In this example, the vending machine is set up for use in Brazil. Thus,the acceptance criteria within the validator for determining which coins areacceptable correspond to Brazilian currency and the tubes are configured tostore coins of Brazilian currency.
More specifically, the coin apparatus 2 accepts 25 centavo coins and50 centavo coins, amongst others. There are two tubes 22, 28 to which 25centavo coins are directed and 50 centavo coins are directed to the cashbox21. 25 centavo coins needed for change are dispensed from the first of thosetubes 22 unless it is empty in which case they are dispensed from the secondtube 28. Accepted 25 centavo coins are directed to the second tube 28 unlessit is full, in which case they are directed to the first tube 22.
The processor 50 registers how many coins are in each of the tubes 22,24, 26, 28 at the end of each transaction, and, more particularly, which of thetubes are empty. Using that information, the processor 50 works out for eachtube on the basis of the previous 50 transactions a value E, which is anindication of how frequently the tube is empty, as follows:E =no. of transactions at the end of which the tube is empty50
The more often a tube is empty, the closer E is to 1.
The more often a tube is empty, the less useful it becomes, and thusthe efficiency of the coin apparatus 2 is reduced. One of the reasons for a tubeemptying frequently is if the coins in the tube are often dispensed in change inmultiples, in which case the problem of a frequently empty tube can bereduced by providing a coin of a higher denomination for use as change.
In this example, when E exceeds the value of 0.6 for the first 25centavo tube 22, the coin mechanism is adapted so that 50 centavo coins areused for change where appropriate in place of the 25 centavo coins. Morespecifically, the processor 50 alters the routing of the coins so that 50 centavocoins are directed to what was previously the first 25 centavo coin tube 22,and 25 centavo coins are directed to second tube 28 or, if that tube is full, tothe cashbox 21. The shape and size of 25 centavo and 50 centavo coins aresimilar so that no change of tube is required.
Other ways of determining when to reallocate an existing tube arepossible, for example, by considering the rate of depletion of a tube, or, asdescribed above, in relation to changes in prices of products or in currencysets.
It may be that when a tube is reallocated coins of the previousdenomination remain in the tube. In that case, those remaining coins may bediverted to the cashbox, or else dispensed as change in the next transaction(even if that results in incorrect payment of change). Alternatively, becausethe mechanism knows how many coins of the previous denomination are in the tube at any time, it may continue to dispense change from the tube in thefirst denomination while filling the tube from the top with the seconddenomination.
In another example, past sales in a vendingmachine are analysed to predict future stock requirements. The apparatus issimilar to the apparatus first described above, but it alsostores in the audit device 66 what item of stock is dispensed at each sale. Theprocessor then analyses the stored data and uses it to produce predictions offuture sales for each product.
Using trends derived from the analysis of past sales and the currentstock position, the machine predicts the earliest time when one of the productswill go out of stock. Alternatively, for example, the machine predicts theearliest time when one of the products reaches a level of 5% of the maximumlevel of that product. This information is an indication of the latest time bywhich a serviceman should visit to replenish the machine. In a similar manneras described in earlier embodiments, the machine can also estimate and outputthe number of items of each product that will be required to refill the machineat that time. This information can lead to a reduction in the number of visitsto a machine by a serviceman needed.
Alternatively, the machine is provided with the date on which theserviceman will next visit, and the machine then estimates and outputs theamount of stock to be added of each product so that the machine will havesupplies of each product until the serviceman's next visit on that given date.This can prevent overstocking of a machine and also lead to fresher productsbeing provided in the machine.
Also, the machine can predict when any product is first likely to go outof stock, or reach a predetermined level, and using that information advisehow many items of each of the other products to provide so that they all goout of stock, or reach the given level, at around the same time.
In this example, calculations are performed daily, and informationis available at the vending machine on a display by pressing an appropriatekey. Alternatively, results can be supplied to a remote position. Also, thedata can be supplied to a remote position and processed there.
The algorithm used in this example is discussed below and shownin the flowchart of Fig. 7. The vending machines offers M different types ofproducts for sale, and the algorithm uses information about sales of all typesof products to make a prediction of future sales.
According to the algorithm, a number of variables are used to calculatea general trend and a variability component, which are then used to predictfuture sales over a desired future period.
In the following, the subscript n refers to the day number, where n=0for the day just finished, n=1 for the day before and so on up to n=6 for thelast day in the preceding week. Thus if today is Tuesday, then n=0 refers toMonday, n=1 refers to Sunday, etc., and so on up to n=6, referring toWednesday. Similarly, if today is Wednesday, n=0 refers to Tuesday, etc. Fand f are fade factors, representing the fact that, as time passes, informationabout a day's activity plays a smaller part in predictions. F is the fade factorper week and f is the fade factor per day. In this example, (1-f)7 = (1-F), andmore particularly, F = 0.5.
The variables An and xn are used, where An is the average sales forproduct A for day type n, relative to the running average over all days for allproducts, and xn is the running average daily total sales for all products ascalculated on day n. Instead of calculating the averages using all datacollected from when the vending machine first started operating, which wouldrequire a large amount of memory, the averages are calculated at the end ofeach day using algorithms, as set out below:Acurrent = A6(1-F) + FtA/x0 where tA = total number of sales for product A for the day just ended.xcurrent = x0(1-f) + fT where T is the total number of sales for all product types for the dayjust ended.
Now, a value W is calculated where W = xcurrent/x6. Thus, Wrepresents the running average just calculated relative to the same runningaverage as calculated on the same day the previous week. If W is less than 1,then it is changed to 1 for calculating stock requirements, because an increasein stock requirements is the main interest.
Another variable Rn is also calculated, where Rn is the average totalsales for a particular day type (eg Monday), relative to the running averageover all days as calculated the previous day.Rcurrent = R6(1-F) + FT/x0
Also a variance term VA is calculated where VA = (1-f)VA + fDA2
DA is the difference between the actual total sales of product A for theday just ended and the expected total sales for product A for the day justended.
Thus:DA=tA-EA where tA = total number of sales for product A for the day just endedand EA = the expected total number of sales for product A for the day justended. EA is calculated by the equationEA = A6x0 where A6 and x0 are as already calculated at the end of the precedingday.
If DA is negative, it is set to 0. This gives a different response to risesin product throughput compared with falls.
Having performed all the above calculations, the calculated values forx, A and R are renamed for use in subsequent calculations, that is calculationsat the end of the day to come, and also for calculations of the general trend,discussed below. More specifically, x6 is dropped, the values x5 to x0 arerelabelled x6 to x0 respectively and xcurrent is relabelled x0. Similarly, A6 and R6 are dropped and Acurrent and A0 to A5 are relabelled A0 to A6 respectivelyand Rcurrent and R0 to R5 are relabelled R0 to R6 respectively.
The general trend is then derived using these newly labelled values x0to x6, and W as already calculated as set out above. The general trend is givenby xnWm, where n is the day of the week, from n=0 to n=6, and m is the weeknumber, starting at m=1 for the week to come. Thus, the general trend for thecoming day is x6W, for the following day it is x5W and so on up to x0W. Theweek after starts with x6W2, x5W2 and so on.
The general trend as set out above gives a broad indication of likelyfuture sales. A respective variability component is also added for each productas an allowance for fluctuations.
The variability component for product A is the square root of (kdV
A)
where:
- d = no. of days ahead for the prediction;
- VA is the variance term as calculated above; and
- k is a constant.
In this example, k = 18, which has the effect of imposing an allowancesimilar in scale to three standard deviations for a normal distribution.
Other algorithms can be used to predict future product requirements.For example, a prediction for each type of product using only sales of therespective product is also possible. Also, for example, the calculationsdescribed above in relation to coins can be adapted for use with stock, themain difference being that new stock is not introduced at each end. In analternative example, a neural network is used to perform predictions offuture stock requirements.