Relative system response elevator dispatcher system using "artificial intelligence" to vary bonuses and penalties

Description: TECHNICAL FIELD [Detailed description of the invention]

[Industrial application field] The present invention relates to an elevator group management device and a group management method for allocating a plurality of cars to a floor where a hall is called. More specifically, the present invention assigns hall calls to selected elevators in a group of elevators commonly used at landings on each floor of a building, based on weighted relative device response (RSR) considerations. Regarding. These RSR considerations include factors that take into account the operating characteristics of the equipment according to an operating plan that includes multiple desirable factors, and in determining which car should be assigned to the hall call by the computer algorithm. It is based on the relative balance between the factors that essentially allocate "bonus" and "penalty". Similarly, in more detail, the present invention presents the number of people waiting in support of a hall call (number of people waiting for a hall call), the amount of boarding / alighting predicted at a stop, and the predicted car load at the hall calling floor. Based on this information to use artificial intelligence (AI) technology based on real-time and historical traffic forecasting, such as predicting, and then to more fairly distribute car load and car outage. It relates to a group management device and a method for optimally allocating a car based on a vehicle allocation algorithm having variable bonuses and penalties that change RSR bonuses and penalties.

[Conventional technology] -General information- In elevator equipment that uses Relative Device Response (RSR) to assign an elevator car to a hall call, the car's hall call run time is represented by various time-related penalties. Both of these penalties are added together and summed up by various penalties that penalize undesired driving characteristics. Bonuses are granted to the desired operating conditions, and these are subtracted from the total penalty resulting from the relative device response or RSR value. These values are calculated for each car for a given hall call and the car with the lowest RSR value is assigned to respond to the hall call. The penalties and bonuses selected for various time delays and driving characteristics are fixed or varied, for example, based on the average hall call latency of the last 5 minutes and the current hall call registration time. The vehicle allocation plan treats all hall calls equally, regardless of the number of people waiting for hall calls. They also handle all cars equally, regardless of the current car load, if the car is not fully loaded. It only considers the current car load, not the predicted car load when the car reaches the hall nominal floor. As a result, the car assigned in one cycle is often not assigned later because the car fills up later, and another car is assigned. Often the assigned car does not have the proper capacity. Therefore, when the car is stopped and people are picked up, some of them are left behind, and the remaining passengers need to re-register the hall call, resulting in increased waiting time and user irritation. Extra cars must be sent there, thus increasing the number of car outages and reducing the transport capacity of the device. When many people are waiting, one or more cars are needed to help the waiters, but traditional RSR devices still allocate only one, resulting in service delays and long waits for many people. Will give you time. When the car stops on the middle floor, the passenger transfer time is determined by the number of people getting on and off the car. By using a fixed car stop penalty, delays due to premature stops are only partially penalized. Numerous stoppages mean that the car is likely to be delayed, the car fills up before it reaches the hall call, and the car is additionally due to the car call that occurred in the middle hall call. The car call will be stopped. These are detrimental to the performance of the device, as they often result in reassignment of hall calls, but conventional ones have not been properly penalized. Often a heavily loaded car stops to carry one or two people. This will increase service time for a large number of people. Traditional RSR devices lack information on the number of people waiting to call the hall and the number of people expected to get in and out of the car at a stop, and therefore the expected load when the car reaches the hall call floor. Due to lack of recognition, it is not possible to effectively distribute the car load and the car stop. For further general background information on RSR-type elevator car allocation devices with fixed or variable bonuses and penalties, the assignee's United States Patent No. 4,363,381 granted to Joseph Bitter on December 14, 1982. See No. and US Application No. 07 / 192,436. These methods are further discussed below in a subsection entitled "Conventional RSR Allocation". -Invention prediction method- In contrast to the conventional methods of interest, the invention preferably collects traffic data and predicts traffic levels on all floors of the building at any time of the working day based on historical traffic forecasts and real-time traffic forecasts. Uses the "artificial intelligence" principle system. It calculates the amount of passengers getting off (the amount of getting off) when the car is stopped and the amount of passengers getting on (the amount of getting on) when the hall is stopped. These boarding / alighting loads and the current car load are used to predict the car load and remaining capacity as the car reaches individual or specific hall nominal floors. These predictions and other factors are used to appropriately change the RSR penalties and bonuses for assigning each hole call to one or more cars. In order to make accurate predictions, that is, to accurately predict traffic demand at any working day, some of the vehicle allocation plans of the present invention use either a single exponential smoothing method and / or a linear exponential smoothing method. It should be noted that some of the general prediction or prediction techniques of the present invention are "Prediction Methods and Applications" by Jyon Willy & Sun, 1978 by Spiros McClidakiss and Stephen Sea Wheellight. In particular, as discussed in Section 3.3: "Single Exponential Smoothing" and Category 3.6: "Linear Exponential Smoothing" in general (but not in the context of elevators or similar). is there.

[Problems to be solved by the invention] In the group management device using the conventional RSR method described above, only the current car load is considered, and the predicted car load when the car reaches the hall nominal floor is not considered. Therefore, the cars assigned in one cycle are assigned. Since the car fills up later, it is often not assigned later and another car is assigned. However, often the assigned car does not have the proper abilities, so even if the car is stopped, some will be left behind and they will have to re-register the ball call. Therefore, as a result, in addition to increasing the waiting time and frustrating the user, an extra car must be sent there, which increases the number of car stops and reduces the transport capacity of the entire device. When a large number of people are waiting, one or more cars are needed to help the person waiting, but with the prior art, only one is assigned, resulting in a delay to the large number of people. There is an inconvenience of resulting in poor service and considerable waiting time. An object of the present invention is to provide an elevator group management device and a group management method for improving various inconveniences of the above-mentioned conventional device.

[Means to solve problems] The present invention and suitable algorithms thereof have arisen from the need to fairly distribute car loads and car outages in order to minimize service time and waiting time to passengers and improve transport capacity. This distribution is achieved, for example, by "knowing" the number of people waiting to call the hall, the number of people expected to get on and off at various car stops, and the currently measured car load by traffic prediction. The configuration adopted by the group management device for elevators according to the present invention has one group of elevator cars used at a plurality of landings capable of calling halls, and for each registered hall call, the entire car can be applied to the car. Responds to the car's relative device response (RSR) factor sum signal, which represents the relativity of the hall call assignments to the car according to the device response plan, resulting in the lowest total relative device response for each registered hole call. The fastest to a registered hall call by assigning a car and weighting the relative device response factor by other response factors that indicate that this allocation has increased the predicted time that the car is expected to respond to the hall call. A group management device for elevators equipped with vehicle allocation means for allocating cars to improve overall equipment response rather than allocating responsive cars, said vehicle dispatching means being active on working days. Traffic data detection means for detecting traffic data including information on the amount of passengers getting into the car, the number of hall call stops, the amount of getting off the car, and the number of car call stops in each direction of up and down on each floor in the time zone. And, as a function of the traffic data regarding at least a short time period before a specific hall call to be assigned, a hall call waiting number predicting means for predicting the number of people waiting in support of the hall call, and this prediction. Allocation means to allocate at least one car to a particular hall call based on the number of hall call waits and the predicted car load expected when the car arrives at the hall call floor, and the predicted hall call wait Bonuses and bonuses assigned to relative device response factors weighted for each car based on the number of passengers, the predicted car load predicted when the car arrives at the hall call floor, and the predicted ride and disembarkation at a stop. The allocation means include a variable bonus and a penalty means that change the penalty according to a change in the predicted number of people waiting for the hall call and the car load expected when the car arrives at the hall call floor. Configured to assign the car selected by penalty means to the hall callIt is characterized by having done it. In addition, the elevator group management method according to the present invention has one group of elevator cars used at a plurality of landings capable of hall calling, and for each registered hall call, the entire device response applicable to the car. Responds to the total signal of the car's relative device response factors, which represents the relativity of the hall call assignments to the car according to the plan, and assigns the car with the lowest total relative device response to each registered hole call, as well as A car that can respond most quickly to a registered hall call by weighting the relative device response factor by another response factor that indicates that this allocation has increased the predicted time that the car is expected to respond to the hall call. It is a group management method of an elevator equipped with a vehicle allocation means for allocating a car so as to improve the overall response of the device instead of allocating the vehicle, and the vehicle allocation means is used on each floor during active hours of a working day. A traffic data detection step that detects traffic data that includes information about the number of car rides and hall call stops in each of the up and down directions and the number of exits from the car and the number of car call stops, and should be assigned. As a function of the traffic data for at least a short time period before a particular hall call occurs, a hall call waiting number prediction step that predicts the number of people waiting in support of the hall call and this predicted hall call Based on the number of people waiting and the estimated car load expected when the car arrives at the hall call floor, the allocation step to assign at least one car to a particular hall call, and the estimated number of people waiting for the hall call and the car Predict the bonuses and penalties assigned to the relative device response factors weighted for each car based on the predicted car load predicted when arriving at the hall call floor and the predicted occupancy and disembarkation at stop. The allocation step includes a variable bonus and penalty step that changes according to a change in the number of people waiting for the hall call and the predicted car load expected when the car arrives at the hall call floor, and the allocation step is based on this variable bonus and penalty step. I'll assign the selected basket to the hallIt is characterized by being configured in this way.

[Action] By using the information on the number of people waiting to call the hall, the amount of boarding / alighting, and the current car load, it is possible to calculate and predict the car load when the car reaches the hall call floor, and as a result, the remaining capacity of the car. Can be evaluated. This remaining capacity is matched with the number of people waiting for the hall to be called. If there is an inconsistency between the expected remaining capacity and the number of people waiting for a hall call, the hall call inconsistency penalty is used to allow or disallow the car to answer the hall call. .. Pauses on various floors are calculated using the predicted car load and passenger boarding / alighting. The car stop penalty and the hole stop penalty are changed as a function of the pause time and the number of people waiting for the hall to be called. Thus, car stops for hall calls and car calls are penalized based on the expected transfer time and the expected number of people waiting in support of the hall call to be allocated, resulting in a large number of. When a person is waiting, a car with fewer stops is selected. Stopping a heavily loaded car to carry a small number of people will increase service time for a large number of people. Therefore, in this case, for example, use a car load penalty that is proportional to the number of passengers in the car and changes without considering the number of people waiting for the hall call, that is, a car load penalty that emphasizes the number of passengers in the car. Penalties will be imposed. These penalties are included in the RSR value calculation. Thus, the resulting RSR value is influenced by the car load on the hall nominal floor, the number of people waiting on the hall nominal floor, and the number of people getting in and out of the car at a stop. All of these values can be obtained by using a traffic prediction principle system based on "artificial intelligence". The RSR algorithm improved by the present invention is thus more responsive to traffic conditions and can more effectively distribute car loads and outages, resulting in less latency and service time and higher transport capacity. .. Past device information is recorded in both a "history" database and a "real-time" database, and this stored information is used for yet other predictions. Thus, the present invention is "artificial intelligence" based on historical and real-time traffic predictions for predicting the number of people waiting to call a hall, the amount of boarding and alighting predicted at a stop, and the predicted car load at the hall calling floor. Based on a vehicle allocation algorithm with variable bonuses and penalties that use AI ") technology and change the RSR bonuses and penalties based on this information in order to distribute the car load and car outage more fairly. To the hall call floor. The exemplary approaches and other related RSR techniques that make up the above are detailed in the Examples section. The present invention can be practiced in a wide variety of elevator devices using known techniques in view of the teachings of the present invention, which will be discussed in detail later. Other features and advantages will be apparent from the specification and claims and from the accompanying drawings showing exemplary embodiments of the invention.

[Example] -Exemplary elevator applications- To elaborate on the exemplary uses of the invention, Bitter's United States Patent No. 4,363,381 (hereinafter referred to as the "'381 Patent") referred to above, as well as Donofio and May 18, 1982. The disclosure of the generally accepted United States Patent No. 4,330,836 entitled "Elevator Cab Load Measuring Device" of Games is incorporated herein by reference. A preferred application of the present invention is in a group management device for elevators using a vehicle allocation control device based on a microprocessor, which group management device is in a building serviced by a car under the control of the group management device. In order to determine the response to the hall call registered at multiple landings and the state of the car, the car is communicated, and the hall call is assigned based on the weighted total for each car while considering each call. "Bonus" and "penalty" in the weighted sum for multiple device response factors represented not only by the various car conditions unrelated to the assigned hole call, but also by the other car conditions associated with the assigned hole call. Is assigned. An exemplary elevator device and an exemplary car controller (in the form of a block diagram) are shown and described in detail in FIGS. 1 and 2 of the '381 patent, respectively. It should be noted that FIGS. 1 and 2 are substantially identical to the same drawings of the '381 patent and the United States co-pending application 07 / 192,436. For brevity, FIGS. 1 and 2 are simply outlined or generally shown below, as made in the co-pending application, while still other desired operational details are described above. It can be obtained from Patent No. 381, as well as other prior patents of the assignee. FIG. 1 shows a plurality of exemplary elevator hoistways, namely hoistway A 1 and hoistway F 2, the rest not shown for simplicity. In each hoistway 1,2, elevator cages or cabs 3,4 are guided for vertical movement on rails (not shown). Each car 3,4 is driven in both directions (up and down) or held in a fixed position by drive pulley / motor / brake structures 7,8, idler or return in the vertical hole of the hoistway. Suspended on steel cables 5 and 6 guided by pulleys 9 and 10. Also, each of these cables 5, 6 typically supports counterweights 11,12, the weight of which is the weight of the car, which normally supports half of the permissible load. Almost equal. Each of the cars 3 and 4 is connected by traveling cables 13 and 14 to the corresponding car controllers 15 and 16 usually located in the machine room at the top of the hoistway. As is known in the art, each of these car controllers 15 and 16 controls the operation and operation of the car, and constitutes a "car motion control means". In elevator group management devices equipped with a plurality of cars, it has been common for a long time to provide a group controller 17 as a "vehicle allocation means". The group controller 17 receives the up and down hall calls registered by the hall call buttons 18 to 20 on each floor of the building, allocates various cars in response to each of these hall calls, and further performs group operation. The car is distributed to each floor of the building according to one of several different modes of. This mode of group operation can be partially controlled, for example, by a lobby panel 21 connected to the group controller 17 by, for example, usually appropriate building wiring 22. As shown in FIG. 2, the group controller 17 is configured as a microcomputer 17a, and is composed of a microprocessor 17b, RAM17c, ROM17d, I / O ports 17h to 17k, and the like. In addition, the car controllers 15 and 16 "up" and "down" the response lamp (hole lantern) 23 provided corresponding to the car 3 and the response lamp 24 provided corresponding to the other car 4. Control some hoistway functions associated with the corresponding car and specify hoistway doors so that they are lit accordingly. Position of the car 3 and 4 in the hoistway 1 and 2, de be detected by a main position transducer (PPT) 25 and 26 as possible. Such transducers 25,26 are driven by appropriate sprockets 27,28 in response to steel strips 29,30, the steel strips 29,30 having both ends connected to the car. At the same time, it is provided so as to pass through the idler sprockets 31 and 32 in the vertical hole of the hoistway. Similarly, although not required for elevator devices such as those carrying out the present invention, detailed location on each floor (floor) for more preferred door control and confirmation of floor location information detected by PPT25,26. Information can be detected by sub-position converters (SPT) 33,34. Alternatively, if necessary, the elevator device in which the present invention is carried out can use the inner door area and outer door area hoistway switches of the types known in the art. The above is a description of the general configuration of an elevator device, and with respect to the description is the same for elevator devices known in the art and exemplary elevator devices to which the teachings of the present invention can be incorporated. All functions of the car itself are directed or communicated by the cab controllers 35,36 according to the invention, and serial time multiplex communication with the car controllers 15,16, as well as the car controller by running cables 13 and 14. It can be equipped with communication by direct wiring with 15,16. Each of these cab controllers 35,36 can monitor, for example, a car call button, a door open button and a door close button, and other buttons and switches in the car. In addition, each of the cab controllers 35 and 36 can control the lighting of the button so as to display the car call, and can also control the floor indicator in the car that displays the approaching floor. The cab controllers 35 and 36 are also connected to load detectors that detect weight information used to control car operation, operation and door function. As the load detector used in the present invention, the device disclosed in the '83 6 patent cited above can be used. An additional function of the cab controllers 35,36 is to safely control the opening and closing of the door under conditions where safety is required. The structure of the microcomputer device such that it can be used to run the car controllers 15,16, group controllers 17 and cab controllers 35,36 is easy according to known techniques as described in various commercial and technical presentations. You can choose from the components available for or similar. The software structure for carrying out the present invention and the peripheral features that can be disclosed herein can be organized in a wide variety of modes. -RSR allocation by conventional method- As noted above, the previous early car allocation device (group management device) using the RSR method detailed and established in the '381 patent is based on the relative device response (RSR) factor. It assigns hall calls to cars and is equipped with an elevator controller that has the ability to assign hall calls based on relative foundations rather than absolute foundations, that is, has the ability to make assignments based on the relative relationship of the state of each car. It is equipped with a control device, for which special preset values are used to assign RSR "bonuses" and "penalties". However, since these bonuses and penalties are fixed and preselected, the waiting time is often long depending on the condition of the device. Therefore, although the invention of the '381 patent was a substantial advance in the art, there are such improvementable points, and the improvementable points are as follows, and the simultaneous pending application No. 07 / 192,436 described above. Achieved in the issue. In the invention relating to the simultaneous application, the bonus and the penalty are different from being preselected and fixed as in the invention of the '381 patent, for example, the relatively current density of traffic in the building (traffic). It is changing as a function of the latest past average hall and waiting time that can be used to measure the quantity) and the current hall call registration time. An exemplary average time cycle that can be used is 5 minutes, and a time cycle of that degree is preferred. During the operation of the device, the average hall call latency for the selected past time cycle is, for example, the total number of hall calls answered during the selected time cycle and the clock time in the hall call registration. Evaluate using and. The hall call registration time is calculated from the time when the hall call is registered until the time when the hall call is assigned. The penalties and bonuses in the invention relating to the simultaneous application are selected to be suitable for hall calls that remain registered for a long time in relation to the average latency of the past selected time period of the hall calls. Has been done. When the hall call registration time is longer than the average wait time of the selected time cycle in the past, the long waited hall call has a high priority, thus, for example, a matching car call stop or an adjacent stop. It does not wait for a car with a car, nor does it wait for a car with less than the allotted acceptable number of calls, and does not activate the MG (motor generator) that drives the car to stop the car. Thus, due to these circumstances, bonuses and penalties can be altered by reducing them. When the hall call registration time is short compared to the average waiting time of the selected time cycle, the opposite situation is justified, and the bonuses and penalties are increased due to those situations. Can be changed. The working relationship used to select bonuses and penalties is, for example, associating the ratio of hole call registration time to the average hole call wait time of the selected time cycle in the past with increasing and decreasing values of bonuses and penalties. It was. As a variant to the above, bonuses and penalties may be reduced as a measure of current traffic density based on the difference between the current hall call registration time and the average hall call wait time of the selected time cycle in the past. , Or can be increased. -Variable bonuses and penalties based on exemplary "AI"- The AI principles used in the present invention and the application of the present invention in detailed exemplary embodiments will be discussed first, and then the exemplary embodiments will be further discussed in relation to the drawings. For example, between 6 am and 12 midnight during an active working day, the following traffic data in each direction on each floor of the building will be available for a short period of time (short time cycle), for example, each Collected at 1 minute intervals. That is, The number of hall call stops made and The number of passengers (passenger volume) to board the car on each floor determined by car load measurement, The number of car call suspensions made and The number of passengers getting off the car (amount of getting off) on each floor calculated by car load measurement, Is collected. At the end of each time cycle, the data collected during the last three time cycles on different floors regarding the number of passengers and car stops is analyzed. If, for example, a car stop occurs in either direction on any floor during 2 minutes of the last 3 minutes, and the car has stopped during at least 2 of the last 3 cycles. If the traffic data indicates that there are an average of two or more passengers getting on or off each car on that floor and in that direction, real-time predictions for that floor and in that direction are initialized. The traffic pattern (traffic demand) for the next two or three cycles of direction and traffic type (boarding or disembarking) on that floor is preferably predicted by using the linear exponential smoothing model. Both the number of passengers and the number of car stops (hall stop or car stop) are thus predicted. Traffic patterns are preferably also predicted for some preemptive cycles beyond the next cycle. That is, the traffic demand for three or more cycles ahead may be predicted. Heavy traffic may be caused by normal traffic patterns that occur on each working day of the week, or by special events (events) that occur on special days. For example, the total number of car stops on the floor in that direction and in that traffic type is less than 2 in each of the four consecutive time cycles, and getting in or out of the car during each of those four consecutive time cycles. Real-time prediction ends when the average number of passengers is less than two. Whenever a significant amount of traffic is observed on a floor in a certain direction and real-time traffic forecasts are made, the traffic data collected in real time for various time cycles is stored in the history database at the end of the real-time forecasts. Will be done. The floors on which the direction and type of traffic with respect to the number of boarding or disembarking and the hall call stop or car call stop are observed are recorded in the history database. The day of the week and the start and end times of traffic are also recorded in the history database. Once a day, at midnight, the traffic data stored in the history database throughout the day is compared to the traffic data from the previous day. For example, the tolerance of start time and end time is within 3 minutes, and the margin of error of traffic change during the first 4 cycles and the last 4 cycles is within 15%, and the same traffic cycle is on each working day. If repeated in, the data for the current day is stored in the normal traffic pattern file. On the other hand, the traffic data does not repeat on each working day, but this traffic pattern is on the same day of the week, for example, during the first four and last four cycles with a three minute margin of error for start and end times. If repeated within a 15% tolerance of traffic change, the traffic data for the current day is stored in a regular weekly pattern file. Traffic data collected throughout the day is analyzed in this way and stored in regular pattern files and regular weekly pattern files, and then within those files for different floors, directions and traffic types. All data will be used to predict traffic for the next day. For each floor, direction and traffic type, different occurrences of historical patterns are identified one by one. For the occurrence of each such historical pattern, the next day's traffic is predicted using the data from the previous occurrence, the predicted data from the last occurrence, and the exponential smoothing model. All normal and normal weekly traffic patterns that are expected to occur the next day are thus predicted and stored in the historical forecast database for the current day. At the end of each traffic data collection cycle, the floor on which significant traffic (congestion) was observed and its direction are identified. After real-time traffic forecasts for traffic types during congestion are predicted, the current day's history forecast database identifies whether historical traffic forecasts were made for the same traffic type on this floor and direction for the next time cycle. Is checked. If so, then the two predicted values are combined to get the best prediction. These predictions give equal weighting to historical and real-time predictions and use half (0.5) weighting factors (coefficients) for both. However, if the traffic cycle is started once, the real-time prediction will differ from the historical prediction by more than 20% in 4 out of 6 1-minute cycles, for example, so the real-time prediction may be, for example, 4 Along with a weighting of 3/4 (0.75), historical predictions are weighted 1/4 (0.25) to get the best predictions. Real-time predictions are made with respect to the number of rides or disembarkations and the number of times the car stops at a hall or car call from the end of the current cycle to 3 or 4 minutes ahead. Historical forecast data up to 3 or 4 minutes ahead can be obtained from the previously generated database. Therefore, this combined prediction of the number of passengers (boarding / alighting amount) and the number of car stops can also be made for 3 to 4 minutes from the end of the current cycle. If historical predictions are not made on the same floor for the same direction and traffic type for the next few time cycles, the real-time predicted number of passengers and car stops for the next 3 or 4 minutes is optimal. Used as a prediction. Using this predicted data, the amount of boarding and disembarking on the floor where considerable traffic occurs is then calculated. The amount of passengers is calculated as the ratio of the total number of passengers boarding the car on that floor in that direction during the cycle to the number of hall call stops made on that floor in that direction during the cycle. The amount of disembarkation is calculated as the ratio of the number of passengers getting off the car in that direction in that direction to the number of car call stops made in that direction in that cycle. Ride and disembarkation for the next 3-4 minutes with respect to floors and directions where significant traffic is observed are thus calculated every minute. Ride if traffic on one floor and in one direction is not significant (not crowded), that is, for example, if there are less than two passengers on average in or out of the car. The amount and disembarkation amount are not calculated. Next, when the hall call is received, the predicted car load at the hall call floor is calculated for each car. The car load expected when the car arrives at the hall call floor has already been registered from the sum of the current car load and the total number of passengers expected to board the hall call stop in the middle of the allocation. It is equal to the value obtained by subtracting the total number of passengers who are expected to get off when the car is stopped. In the calculation of this car load, the traffic at either the hall call stop or the car call stop is not remarkable, so if it is not predicted, the passengers who get in the car at the hall call stop will also get off the car at the car call stop. Assume that there is only one passenger in each case. The calculated car load is used to calculate the remaining capacity in the car for passengers. The estimated occupancy at the hall nominal floor is compared to the remaining capacity. Penalties, called "hole call nonconformity penalties" (HCMs), are used to allow or disallow a car to respond to a hole call, as follows: If the floor on which the hall call occurs does not have significant traffic, as mentioned above, only one person is considered to be in the car on that hall call, so if the car is not fully loaded. For example, if the car has a load that does not exceed 80% of the car capacity, it is desirable to allocate the hole call. So if the calculated car load is less than 80% when the car reaches the current hall nominal floor, the HCM is set to zero. If the calculated car load exceeds 80%, the HCM is set to, for example, 200. The method according to the present invention uses the current car load to allow or disallow car allocation and does not take into account the amount of boarding and disembarking caused by the stop of hall call and stop of car call in the middle. It is different from the patented method. Thus, this technique minimizes hall call reassignment due to a car that is fully loaded with a stop in the middle. In addition, the group management device using RSR according to the '381 patent does not use the estimated number of people waiting at the hall call floor in order to select the car to be assigned. In the present invention, if the floor on which the hall call occurs has significant traffic, the car load on that hall call is calculated and then the remaining capacity of the car is calculated with respect to the number of passengers. If the predicted occupancy on the hall nominal floor is less than or equal to the "single car limit waiter" and the remaining capacity of the car is equal to or greater than the average occupancy on the hall nominal floor, then the car is Desirable for allocation, HCM is set to zero. If the average occupancy on the hall nominal floor is less than the size of a single car limit waiter, but the remaining capacity of the car is less than or equal to the average occupancy on the hall nominal floor, then the car is allocated for the hall nominal floor. Not desirable for Therefore, the HCM is set to, for example, 200. Here, the "single car limit waiting number" means the upper limit of the number of passengers who can ride in this single car when only one car reaches the hall nominal floor. Thus, it is possible to avoid a situation in which a plurality of cars are stopped to carry a small number of people. This improves car productivity by minimizing car outages. If the average occupancy on the hall floor exceeds the single car limit waiting number, and if the remaining car capacity is the "minimum occupancy (boarding) limit for multiple cars", that is, less than two people, that The HCM is set to 200 because the car is not desirable for allocation. Here, the "minimum passenger limit for a plurality of cars" means the lower limit of the number of passengers who can board one car when a plurality of cars arrive at the hall nominal floor. If the average occupancy on the hall floor exceeds the number of passengers waiting for a single car and the remaining capacity of the car is equal to or greater than the minimum occupancy limit for multiple cars, the HCM penalty is set to zero. Next, if the remaining capacity of the car is less than or equal to the average occupancy on the hall nominal floor, the car will generate a second car request signal (SCR). If the car with the lowest RSR value does not generate an SCR signal, then this car alone responds to the hall call. If the car with the lowest RSR value emits an SCR signal, the car with the next lowest RSR value will also respond to the hall call for support. The size of the single car limit waiting number and the minimum boarding limit of multiple cars are functions of the traffic density at that time. These values are known by the device and are varied, for example, every 5 minutes each. When the first car answers the hall call, if this first car is not fully loaded when closing the door on the hall call floor, the first car is boarded by all the waiting passengers. It emits an SCR cancel signal indicating that. The other car that was responding to the hole call by the SCR signal from the first car releases its own assignment regarding the hole call by the SCR cancel signal. The group management device using the RSR according to the '381 patent uses a fixed car stop penalty and a hole stop penalty. The typical value of this car stop penalty (CSP) is 10, and the typical value of the hole stop penalty (HSP) is 11. When traffic data is predicted and the car load on various car call stops and hall call stop floors is evaluated, the remaining capacity of the car and the predicted boarding and alighting volumes are based on, for example, real-world observations. It is used to calculate the required door downtime (car downtime) on the floor by using an appropriate mathematical model. That is, the door opening time (door suspension time) required for passengers to get on and off on the stop floor is the remaining capacity of the car on the stop floor and the number of passengers expected to board the car on the stop floor. Since it is a function of the number of passengers getting off the car at the stop floor, it can be obtained by incorporating these factors into a mathematical model created based on the observation of boarding / alighting time in an actual elevator system. Can be done. Therefore, with respect to each car call stop and hole call stop, the car stop penalty is incremented if, for example, the required car stop time exceeds 1 second and the hole stop time exceeds 3 seconds. For example, if the stop time is increased by 2 seconds each, the car stop penalty and the hole stop penalty are increased by, for example, 1. Thus, if the car is expected to consume more time on a premature stop to unload or load a large number of passengers, the car will be penalized with an appropriate car stop penalty. In addition, these car stop and hole stop penalties are preferably varied as a function of the number of people waiting in favor of the assigned hall call. This is because each stoppage increases the probability that an unpredictable incident will cause a delay in the car and an increase in the load on the car. In addition, if the car makes a halfway stop to call the hall, passengers who board the car at this halfway stop may make a new car call. As a result, the car is further delayed. Unpredictable delays and loads will result in later reassigned hall calls. Choosing a car with fewer stops means more reliable car arrival at the hall nominal floor. The car stop penalty increases with the size of the number of waiters, as it is desired by a large number of waiters that the hall call allocation is reliable and the probability of reassignment is low. Thus, when there are many people waiting in favor of a hall call, more penalties are imposed for stoppages, while when the number of people waiting is small, a lower penalty is imposed. This plan, which prevents a large number of people waiting by choosing a car with fewer stops, results in a smaller number of people with less waiting time, resulting in a lower average waiting time for the device. Will occur. The table below shows a typical increase in car stop penalties when the downtime is 1 second for car stop and 3 seconds for hole stop.<img file="JP2509727B2_D0001.tif" /> The increase in penalty is variable as a function of traffic density. In harsh traffic conditions, these penalties increase rapidly with the number of waiters, as fewer outages are desired to help a large number of people waiting for a hall call. For hall calls with few people waiting, a car with more stops is available for service. The number of people who support and wait for the hall call is predicted using the "artificial intelligence" technology of the present invention, and the car load when the car arrives at the hall call is calculated, the car load penalty ( CLP) is used to penalize a heavily loaded car stop in the absence of a car call stop that matches the hall call. This car load penalty is variable and increases in proportion to the number of people in the car. This increase is high when the number of people waiting in support of the hall call is small. When there are a large number of people waiting in support of the hall call, the car load penalty increases the car load by a lower amount. If the car has a car call stop that matches the hall call, the car load penalty is set to 0. The change in the car load penalty (CLP) between the car load and the person waiting in the hall is expressed by the linear correlation model as follows. CLP = a_Cld(c_ld-c_ld1) -b_phc* N_phc Here, "a_cld"And" b_phcIs a correlation coefficient, and "c"_ld"" Is the car load when the car reaches the hall nominal floor, and "c"_ld1Is the set car load limit, and "N"_phcIs the number of people waiting on the hall floor. "A_cldAnd "b_phcThe exemplary changes in "c" are in the range of 0.3-3.0 and 0.5-1.5, respectively._ld1Is 4 to 12. The car load is "c_ld1When it is smaller than, there is no car load penalty. This set car load limit is determined by the number of people waiting in support of the hall call. As can be seen from the above equation, the model selects a lightly loaded car to serve a small number of waiters. The car can only accommodate up to as many passengers as its remaining capacity. Therefore, the linear correlation model should not be used if more than the remaining capacity is waiting in the hall. This is handled by limiting car allocation. Thus, if the car does not have adequate remaining capacity, the hole call nonconformity penalty (HCM) eliminates the hole call assignment to the car or optionally allows one or more cars to answer the hole call. Assigned. Therefore, the car load penalty is the car load (c)._ld), But the number of people waiting in support of the hall call (N)_phc) Decreased by these "c_ld+ N_phcIs applied until the total value of "" approaches or reaches the capacity of the car. Thus, the CLP can be calculated by using the above equation. The above equation is "a_cld"And" c_ld1"And" b_phcSpecified with respect to the value with, and for example, "N" from 1 to 12._phc"A different value is used. "N_phcWhen "" exceeds 12, the equation for 12 passengers is used. Since the above-mentioned special example is used as an exemplary embodiment of the present invention, the logical block diagrams of FIGS. 3A and 3B include collecting traffic data, predicting traffic, and calculating the amount of boarding and disembarking. Show an exemplary principle system for. In steps 3-1 and 3-2, which are "traffic data detection means" or "traffic data detection step", the traffic data is "up" and "up" on each floor (in the figure, "floor" is referred to as "floor"). In the "down" direction, with respect to the number of passengers getting into the car, the number of hall call stops made, the number of passengers getting out of the car, and the number of car call stops, at least all of the working days during operation, for example. , Collected every minute cycle, during an appropriate time frame covering from 6 am to 12 midnight (step 3-2). Then, for example, the data collected in the latest hour is stored in the database as illustrated in FIGS. 4A and 4B and step 3-1a. In the next step 3-3, it is determined whether or not the collection of traffic types (cars arriving and departing in the ascending direction and cars arriving and departing in the descending direction) on each floor is completed, and "YES" in this step 3-3. If it is determined, the process proceeds to step 3-3a to determine whether or not it is 12 o'clock in the middle of the night. When it is judged as "YES" in this step 3-3a, the traffic is predicted for the next day, and when it is judged as "NO", it ends. In the next step 3-4, it is determined whether or not the prediction processing is in progress, and if the prediction processing is not performed, the process proceeds to step 3-4a, and whether or not the car has arrived or departed for 3 of the 4 cycles. Is determined. If "YES" is determined in step 3-4a, the process proceeds to step 3-5. In step 3-5 as the "congestion detection means" or "congestion detection step", by analyzing the data, for example, in two of the three 1-minute cycles, the car stop is "up" and "down". It determines if it was done on any floor in the direction, and determines, for example, whether, on average, two or more passengers got off or boarded from each car during those cycles. If so, step 3-5 is judged as "YES" and it is considered that congestion has occurred. For example, traffic for the next 3-4 minutes is generally described in the text of Macridakis and Wheellights cited above, especially in Category 3.6, to the elevator car dispatch in the specification of the parent application mentioned above. Using real-time data and a linear exponential smoothing model, as applied, predictions are made at steps 3-6 on that floor in that direction. Therefore, if the traffic of "today" changes significantly from the traffic of the previous day, this change will be used immediately in the prediction. Here, the steps 3-6 correspond to the "hole call waiting number prediction means" or the "hall call waiting number prediction step". Further, this step 3-6 corresponds to the "step of predicting ... less than a few minutes" in claim 22. If this traffic pattern repeats on this floor each day of the week or each same day, the traffic data is stored in a historical database, and the data with a 2-minute or 3-minute cycle, for example, the moving average method, or in particular. The traffic for this day is predicted the night before, preferably by using a single exponential smoothing model. The model is generally described in the text of the Macridakis and Wheellights cited above, in particular in Category 3.3, as well as in general as applied to elevator car dispatch in the parent application specification described above. Be explained. Step 3-9 determines if there is historical traffic data for the next 3-4 minutes predicted the night before, and if such a prediction is available, say "YES". Once determined, steps 3-10 are taken to combine historical and real-time predictions for optimal prediction. Forecasting has the following relationship between both real-time forecasting and historical forecasting, that is, X = ax_h+ bx_rObtained by combining according to. Where "X" is the combined prediction, "x"_hIs a short time cycle history forecast for the floor, "x"_r"Is a short time cycle real-time forecast for a floor, and" a "and" b "are increasing factors. First, 0.5 is used for the values of "a" and "b". If the real-time prediction differs from the historical prediction by 20% or more in some cycles, for example, the value of "a" is decremented and the value of "b" is increased, as described above. Here, the steps 3-9 and 3-10 combine "at least some past ... predicting steps" in claim 23 or "real-time prediction and historical prediction" in claim 24. The relationship between "real-time prediction and historical prediction" in "Steps for obtaining optimal prediction" or claim 25, X = ax_h+ bx_rCorresponds to the "step of joining according to". If historical forecasts are not available, real-time forecasts are used for optimal forecasts, as shown in steps 3-11. Next, in steps 3-12, each floor and direction in which considerable traffic occurs is predicted, and the average occupancy is, for example, the number predicted to ride in the car during that cycle and the hall call stop made during that cycle. Calculated as a ratio to a number. The average disembarkation amount is calculated in steps 3-13 as the ratio of the number predicted to disembark from the car during the cycle to the number of car call stops made during the cycle. Here, these steps 3-12 and 3-13 are described in claim 26 as "in the cycle ... Corresponds to the "step to find the average disembarkation amount". These boarding and disembarking amounts are calculated for the next 3 to 4 minutes in step 3-13a and stored in the database. If "YES" is determined in step 3-4, it means that the prediction process is being performed, so the process proceeds to step 3-7, and it is determined whether or not the number of car stops in 4 cycles is less than 2. To do. If "NO" is determined in step 3-7, the process proceeds to step 3-6, and if "YES" is determined, the process proceeds to step 3-8. In this step 3-8, it is determined whether or not the average boarding amount or the average disembarking amount (average boarding / alighting amount) is less than 2, and if it is 2 or more, it is determined as "NO" and the process proceeds to step 3-6. .. On the other hand, when it is determined as "YES" in step 3-8, the process proceeds to step 3-14 as "traffic data storage means" or "traffic data storage step", the traffic data is stored in the history database, and the above step 3 Return to -3. Here, the correspondence between the configuration of claim 3 or 21 that stores traffic data when congestion is detected and the embodiment will be described based on the flowchart and the above-exemplary "AI". As mentioned in the Variable Bonus and Penalty section, when congestion is predicted in steps 3-5, real-time prediction is started in step 3-6, and this real-time prediction is said to be congestion relief in steps 3-7 and 3-8. It will be done until it is judged. Then, when the congestion is cleared and the real-time prediction is completed, the traffic data collected in real time is stored in the history database in step 3-14. Next, when received from a floor with a hall call, the RSR value for each car is calculated taking into account the hole call nonconformity penalty, the car stop and hole stop penalties, and the car load penalty, these penalties. Are all changed based on the number of people expected to support and wait for the hall call, the predicted car load on the hall call floor, and the predicted boarding and disembarking volume at the stop. The above is substantially the same as the initial principle system of the application submitted at the same time (Japanese Patent Application No. 2-52572). With reference to the logical block diagram in Figure 5, which shows a model principle system for calculating the hole call nonconformity penalty, in step 5-1 for certain car calls and hole calls, the car load on the hall call floor is , The total amount of boarding due to the stop of the hall in the middle is added to the current car load, and the total amount of disembarkation due to the stop of the car in the middle is subtracted from this result. In the figure, the stop in the middle is abbreviated as "EN route". Here, this step 5-1 corresponds to the current car load ... step for obtaining the predicted car load in claim 27. In step 5-2, if traffic is not predicted in the current hall nominal, it is judged as "NO", and in step 5-3, the predicted car load is, for example, less than 80% of the car capacity. Judge whether or not. If it is 80% or more, it is judged as "NO" and moves to step 5-5, and the hole call nonconformity penalty (HCM) of this car is set to a high value so that the hole call is not assigned to this car, for example. Set to 200. If this is not the case, the hole call nonconformity penalty is set to zero in steps 5-4 because the predicted car load is less than 80% of the car capacity and is determined to be "YES" in step 5-3. On the other hand, in step 5-2, if traffic is predicted for the current hall call, it is judged as "YES", the process moves to step 5-6, and the hall call is supported and waited. Determine if the expected number of people is less than or equal to, eg, 5 or less, the number of people waiting for a single car limit. If so, this step 5-6 determines "YES" and the logic branches to steps 5-7. In this step 5-7, it is judged whether or not the remaining capacity of the car is less than the number of people waiting for the hall call, and if the remaining capacity of the car is larger than the number of people waiting, it is judged as "NO" and the process proceeds to step 5-9. The HCM is set to zero. Otherwise, it is judged as "YES", step 5-8, and HCM is set to 200. In steps 5-6, when the number of people waiting for a hall call exceeds the number of people waiting for a single car limit and a judgment of "NO" is made, the process proceeds to step 5-10, and the remaining capacity of the car is larger than the minimum boarding limit for multiple cars. Judge whether or not. If the remaining capacity is larger, it is judged as "YES", HCM is set to 0 in step 5-11, and if the remaining capacity is smaller, it is judged as "NO", and step 5 Move to -12. In this step 5-12, the HCM is set to 200 to eliminate the assignment of this car to this hole call. If necessary, i.e., if it is determined in step 5-13 that the car capacity is less than the number of waiters supporting the hall call, then in step 5-14 the RSR value is calculated and the second car request signal ( SCR) is issued. Other detailed steps or features, as can be seen in the figure, are included in the algorithm of FIG. 5, but appear to those skilled in the art to be self-evident. Here, steps 5-4, 5-5, 5-8, 5-9, 5-11, and 5-12 in FIG. 5 are the predicted number of people waiting for a hall call and the car is a hall call in claim 28. Corresponds to the step of calculating the hall call nonconformity penalty based on the predicted car load when arriving on the floor. Refer to Figure 6, which shows the exemplary principle system used to calculate the variable car stop and hole stop penalties, if each car is scheduled to stop in the middle, step 6-1. The current car load, the predicted amount of boarding at the stop of the hall call in the middle, and the predicted amount of disembarkation at the stop of the call of the car in the middle are determined to be "YES" in step 6-2. It is used to calculate the car load when the car is loaded, the remaining capacity after the passengers get off, and the number of passengers transported. Then, in step 6-3, by using these parameters, that is, the remaining capacity, the total boarding amount and the total getting-off amount on the floor, and an appropriate mathematical model based on the real-world observation, the door pause time (getting on and off) is used. Time) is calculated. Here, the "appropriate mathematical model based on real-world observation" means a mathematical model introduced based on the observation of actual elevator traffic. In steps 6-4 as "Variable Bonus and Penalty Means" or "Variable Bonus and Penalty Steps", each car stop penalty (CSP) and hole stop penalty (HSP) of the car is, for example, the table shown above. Number of people waiting in support of hall calls using_phc) Is calculated by adding to the nominal increase in these penalties. In steps 6-5, such calculated penalties are further added with a "1" for each additional 2 seconds of rest time, for example, a minimum of 1 second for car call stop and a minimum of 3 seconds for hall call stop. And adjusted. With reference to the graph in Figure 7, a typical change in car load penalty due to car load and the number of people waiting in support of the hall call is shown for a car with an exemplary capacity of 4000 lbs. And here "N_phcThat is, the number of people waiting at the hall call floor varies from 1 to 12. This graph is based on the equation discussed above. Here, from this graph, it can be seen that the car load penalty constitutes part of the "variable bonus and penalty means" or the "variable bonus and penalty step" (the car stop penalty and the hole stop penalty are also variable). Since the car load penalty itself is a known technique, the flowchart is not shown. The penalties calculated in this way are used in RSR algorithms with other bonuses and penalties to calculate improved RSR values. The RSR algorithm with the variable bonuses and penalties of patent application 07 / 192,436 referenced above can be used for the improvement of the present invention. Thus, traffic predicted using the "artificial intelligence" principle system of the present invention can be used to vary bonuses and penalties and calculate the resulting RSR value. Using this technique, when a car is assigned to a hall call by an assignment program (not shown) as an "allocation means" or "allocation step" built into the group controller 17, which is a vehicle allocation means, the car stops and the car load is reduced. It is distributed more fairly, resulting in better service. That is, as described above, the car load penalty, the car stop penalty, and the hole stop penalty are variably adjusted according to the number of people waiting for the hall call and the predicted car load. Then, the hole call is assigned to the car in which the total value of the total of four penalties including these three variable penalties and the hole call nonconformity penalty is minimized. The allocation program itself as an "allocation means" or "allocation step" for assigning a hole call to a car having the minimum penalty is self-explanatory and is not shown. That is, it is only necessary to obtain the aggregated value of each of the penalties for each car and assign the hole call to the car having the minimum value among them. Although the present invention has been demonstrated in connection with detailed exemplary examples, it should be understood by those skilled in the art that various variations in shape, details, principle system and / or method of the present invention. It means that it can be done without departing from the spirit and scope.

[Effect of the invention] As described in detail above, according to the present invention, the car load when the car arrives at the hall nominal floor, the remaining capacity of the car, and the amount of boarding and disembarking at the halfway stop are predicted, and based on each of these factors. Therefore, the hall call nonconformity penalty, the car stop penalty, the hole stop penalty, and the car load penalty are obtained, and the car with the minimum RSR value caused by each of these penalties is assigned to the hall call. It is possible to improve the serviceability by shortening the time, and it is possible to improve the responsiveness of the entire device and improve the transportation capacity.

[Simple explanation of drawings]

FIG. 1 is a simplified partial rupture schematic block diagram of an exemplary elevator device to which the present invention can be applied. FIG. 2 is a simplified schematic block diagram showing an exemplary group controller that can be used in the apparatus of FIG. 1 and can carry out the present invention. FIGS. 3A and 3B simplify an exemplary algorithm of the principle system used to collect and predict traffic and boarding and disembarking volumes on different floors in a preferred embodiment of the present invention. Logical flow diagram, FIGS. 4A and 4B array the collection of real-time data used in the exemplary embodiment of the invention showing the collection of "up" rides and "up" hall stops on various floors. The overall matrix diagram shown in, FIG. 5 is a logical flow diagram that simplifies the exemplary algorithm for the principle system used in the calculation of Hall call incompatibility in the exemplary embodiment of the present invention. FIG. 6 is a simplified logical flow diagram of the exemplary algorithm of the principle system used to calculate the variable car stop and hole stop penalties in the exemplary embodiment of the present invention. FIG. 7 is an explanatory diagram showing a graph showing a typical change in the car load penalty depending on the number of people waiting in support of the car load and the hall call used in the exemplary embodiment of the present invention. In the figure, reference numerals 3 and 4 are elevator cars, 15 and 16 are car controllers, 17 are group controllers, 18 to 20 are hall call buttons, and 35 and 36 are cab controllers.