JP2998589B2 - Elevator group control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an elevator group management control system for allocating calls using a neural network, and more particularly to improving the responsiveness and fault tolerance of the allocation process using a neural network. It is what it was.

[0002]

2. Description of the Related Art Conventionally, group management control of elevators has mainly been a call allocation method using an evaluation function or a call allocation control using an expert system using fuzzy logic. A new method of assigning calls using a neural network has been proposed.

A neural network is a network imitating the human brain. A plurality of neural cell models (neurons) are connected in a complicated manner, and the operation of each neuron and the connection form between the neurons are determined in a well-defined manner. The function and the knowledge processing function can be embedded. For example, "Nikkei Electronics" 1987
P115-P124 and 1 of August 10, issue (No. 427)
It is disclosed in the book "PDP model" published by Sangyo Tosho Co., Ltd. in February 989, and in particular, those in which neurons are arranged in a hierarchical structure can use an autonomous learning algorithm called "back propagation". There are features.

When this neural network is used, there is no need for a human to consider an assignment algorithm at all, and a decision system for deciding an optimum assigned car in response to various traffic conditions can be automatically generated. For example, Japanese Patent Application Laid-Open No. 1-275381 "Elevator group management control device" and Japanese Patent Application Laid-Open No. 3-31173 "Elevator group management control device" Japanese Patent Application No. 5-2438
No. 17, "Method of Learning Neural Network for Elevator Call Assignment" and the like.

FIG. 4 shows an example of a neural network for call assignment. As shown in FIG. 4, the neural network NN for call assignment includes an input layer NR1 corresponding to an input pattern (elevator system state data), an output layer NR3 corresponding to an output pattern (assignment suitability),
It is composed of neurons of an intermediate layer NR2 located between the input layer and the output layer.

The input pattern uses various data representing the state of the elevator system (floor and direction of landing call, position and driving direction of each car, car call, load state, etc.) as parameters necessary for call assignment. , Into a form that can be input to a neural network, and the number of neurons in the input layer corresponds to the total number of parameters.

When input data is given to each neuron in the input layer, a signal is sequentially transmitted to the output layer, and as a result, a certain value is output from each neuron in the output layer. In the output layer, there are neurons for the number of elevators, and for various input patterns, the neuron corresponding to the car that is optimal for allocation outputs “1” (maximum value), and the other neurons output “0”. Is learned in advance.

Accordingly, among the values of each neuron in the output pattern, a neuron that outputs a value closest to “1” (maximum value) indicates that it is optimal for allocation.
The unit corresponding to this neuron is selected as the assigned unit. Note that the number of neurons in the intermediate layer (one layer in the embodiment, but may be plural) is appropriately determined according to the number of elevators, the properties of the building, and the like.

Although not shown, a connection weight (synapse weight) indicating the strength of the connection between the neurons is set between the neurons. The connection weight is initially set to an appropriate value, but can be modified so that a more accurate call assignment can be performed using a learning algorithm called “back propagation”.

Since the back propagation is well known and will not be described in detail, a learning sample (an input pattern and a desired output pattern corresponding to the input pattern, ie, a teacher signal, are paired). An algorithm that compares the output pattern for the same input pattern with the teacher signal and corrects the connection weights to minimize the error. Initially, all weights are initialized (for example, random The input pattern of the learning sample is given to each neuron of the input layer. The output pattern at this time and the output pattern of the learning sample (teacher signal)
Then, using the difference (error), the values of the connection weights are sequentially corrected from the output layer side so that the difference becomes smaller.

If this is repeated using a large number of learning samples until the error converges, the call assignment function at the same level as the teacher signal is automatically embedded in the neural network, and the learning input The same level of call assignment as the teacher signal can be performed not only for patterns but also for unknown input patterns.

[0012]

The learning of the neural network for assignment is performed before the elevator is installed using a learning sample created in advance by simulation or the like. However, in practice, the learning of the building in which the elevator is installed is performed. It is extremely difficult to accurately grasp the nature and traffic conditions in advance. In addition, since the traffic conditions of the building may change after the elevator is installed, the training of this neural network is continuously performed after the elevator is installed, and the connection weights are adjusted appropriately to match the traffic conditions of the building. It is necessary to go.

For this reason, if a learning sample is created during the operation of the elevator and learning of the neural network is to be performed periodically, the assignment processing during that period becomes impossible.
There is a problem that the responsiveness of the assignment process is deteriorated. For this reason, it is conceivable that the learning process is performed during nighttime or other off-peak hours.However, it is not possible to perform sufficient learning due to time restrictions only at night, and the learning result is sent to the neural network for assignment. There was a problem that it could not be reflected quickly.

A first object of the present invention is to perform a neural network assignment process and a neural network learning process.
An object of the present invention is to provide a group management device that can perform parallel processing at the same time without affecting each other.

Further, a second object of the present invention is to take into account that even if parallel processing can be performed at the same time, if a failure occurs in a neuro assignment device that performs assignment by a neural network, assignment cannot be continued. An object of the present invention is to provide a group management device that is also excellent in fault tolerance.

[0016]

According to a first aspect of the present invention, there is provided a neural network learning apparatus for creating a learning sample and performing a neural network learning process during operation of an elevator. Is provided independently of the neuro-allocating device that performs the allocation process by using the neural network, so that the allocation process by the neural network and the learning process of the neural network can be simultaneously performed in parallel.

Further, in the second invention of the present application, a neural network learning device that prepares a learning sample and performs a neural network learning process while an elevator is operating is a neural allocation device that performs an allocation process using a neural network. Independently provided, an assignment monitoring device that monitors that the neuro assignment device is operating normally and downloads an assignment program in the neuro assignment device to the neural network learning device when an abnormality occurs. Features.

[0018]

According to the first aspect of the present invention, the assignment processing by the neural network and the learning processing of the neural network are executed independently, and are performed in parallel without affecting each other.

In the second invention of the present application, in addition to the above, when the assignment monitoring device detects a failure of the neuro assignment device, the assignment program and network data are downloaded from the neuro assignment device to the neural network learning device, The assignment is continued by the learning device.

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the overall configuration of the first invention of the present application. In the drawing, reference numeral 1 denotes a hall call button provided on each floor (only one floor is shown, and other floors are shown). Is omitted), 2 is a hall call signal, D1 is an operation control device for managing the operation of the first unit, and similarly D2 to Dn are the second to
An operation control device for the n-th car, 3 is a car information signal indicating the state of each car (car position and driving direction, whether running or stopped, door open / closed state, car call, load, service floor, abnormality, etc.), 10 Is a group management device for allocating a hall call to an optimal car based on the elevator system state data comprising the hall call signal 2 and the car information signal 3 and outputting it as an allocation signal 4. Each of the operation control devices D1 to Dn Controls the operation of the car so as to sequentially respond to the hall call assigned by the assignment signal 4 and the car call registered in the car of the own car.

The group management device 10 includes a neural network learning device 20 and a neuro assignment device 30.
CPU 21 (31), ROM 22 (32), R
It comprises an AM 23 (33), an I / O port 24 (34), and a communication IC 25 (35).

The neuro assignment device 30 stores an assignment program in the ROM 32 and an assignment neural net in the RAM 33, and converts an elevator system state data at the time of a landing call into an input pattern of the neural net. An input process is executed to input the neural network for allocation, select an optimum car from the output pattern, and output the selected car as an allocation signal 4.

In the neural network learning device 20, a learning program is provided in the ROM 22 and a learning neural network is provided in the RAM 23. A learning sample is prepared as needed during the learning of the elevator, and the learning of the neural network is continuously performed. Then, a learning process for updating the neural network every time learning is completed is executed. The exchange of necessary data between the neural network learning device and the neuro assignment device is performed via the communication ICs 25 and 35.

With the above configuration, the neural assignment processing and the neural network learning processing are executed independently without affecting each other, and the responsiveness of the assignment processing deteriorates due to the neural network learning. Such a situation can be prevented.

Although the above configuration can solve the problem of the responsiveness of the assignment process, if the neuro assignment device fails for some reason, the assignment becomes impossible immediately and the elevator operation cannot be continued. The problem remains. The second invention of the present application has solved this point.

FIG. 2 is a block diagram showing the overall configuration of the second invention in the present application. The configuration is exactly the same as that of FIG. 1 except that an assignment monitoring device 40 and signals A to C are added, and the same components as those in FIG. 1 are denoted by the same reference numerals.

The allocation monitoring device 40 also has a CPU 41, a ROM 42, a RAM 43, and an I / O like the neural network learning device 20 or the neuro allocation device 30.
A monitoring function for monitoring the operation of the neuro-allocating device, which is composed of a port 44 and a communication IC 45, and upon detecting a failure of the neuro-allocating device 30, downloads an allocation program from the neuro-allocating device 30 to the neural network learning device 20. And a program management function for allowing the neural network learning device to continue assignment.

The processing procedure in the assignment monitoring device 40 when an abnormality occurs in the neuro assignment device will be described with reference to the flowchart of FIG. First, when an abnormality of the neuro assignment device 30 is detected, the signals A and B are validated in step S1. This signal A is connected to the communication IC 35 of the neuro-allocating device, and by making the signal A valid, the communication IC 35 changes from the normal mode (mode operating like a normal communication IC) to the bus access mode (serial mode). (A mode for accessing the bus in accordance with command data from the line).

The signal A is also connected to the reset input of the CPU 31, and when the signal A becomes valid,
31 is forcibly reset, and the bus is no longer accessed. As a result, the right to use the bus can be reliably transferred from the CPU 31 to the communication IC
35, and the communication IC 35 can read the ROM 32 or R
It becomes possible to access the AM33 and the like.

On the other hand, the signal B is similarly connected to the communication IC 25 and the CPU 21 of the neural network learning device 20, respectively, and the signal B is made valid to switch the communication IC 25 to the bus access mode. To access the ROM 22 and the RAM 23, and reset the CPU 21 to stop the learning process.

As described above, when the communication IC can access the ROM or the RAM by the command from the assignment monitoring device, the assignment monitoring device 4
The assignment program is read from the ROM 32 of the neuro assignment device through the communication IC 35 in response to a command from 0.

Next, in step S3, the signal C is made valid. The signal C is connected to the control circuit of the neural network learning device, and when it becomes valid, plays a role of switching the ROM area to the RAM area. As a result, the ROM 22 of the neural network learning device becomes writable, and the allocation program is downloaded to the RAM (ROM 22) in step S4.

Similarly, in step S5, network data (connection configuration of the neural network for allocation and data of each connection weight) is read out from the RAM 33 of the neural allocating device, and in step S6, the network data is stored in the RAM 23 of the neural network learning device. to download. Then, in step S7, the signal B is invalidated, and C
Release the reset of the PU 21.

As a result, the neural network learning device is restarted by releasing the reset. At this time, the assignment program downloaded to the ROM 22 is read in place of the learning program, and thereafter, the neural network learning device uses the assignment program. The assignment will continue. When making the assignment by the neural network learning device 20, the hall call signal 2 and the car information signal 3 are required, which can be read from the I / O port 34 via the communication ICs 25 and 35. However, the landing call signal 2 and the car information signal 3 may be connected directly to the I / O port 24 in parallel with the I / O port 34.

[0035]

According to the present invention, the assignment process is not affected by the learning process at all, so that the responsiveness of the assignment process is improved, the learning efficiency is improved, and even if an abnormality occurs in the assignment device, The assignment can be continued by the learning device.

Further, according to the present invention, the data stored in the RAM of the neuro-allocating device can be directly obtained by a command from the allocation monitoring device, and the latest data is used even when the neuro-allocation is restarted. The excellent effect that can be performed can be exhibited.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of a first invention in the present application.

FIG. 2 is a block diagram showing an overall configuration of a second invention in the present application.

FIG. 3 is a flowchart illustrating an example of a processing procedure according to the second invention of the present application.

FIG. 4 is a diagram showing an example of a neural network for call assignment.

[Explanation of symbols]

 D1 to Dn Operation control devices of Nos. 1 to 1 1 hall call button 2 hall call signal 3 car information signal 4 assignment signal 10 group management device 20 neural network learning device 30 neuro assignment device 40 assignment monitoring device

Continued on the front page (72) Inventor Mami Nakagawa 1-28-10 Sho, Ibaraki City, Osaka Prefecture Inside Fujitec R & D Co., Ltd. (72) Inventor Tomoaki Tanabe 1 28-10 Sho, Ibaraki City, Osaka Stock Company In Fujitec R & D Co., Ltd. (72) Naoki Ota 1-28-10 Sho, Ibaraki-shi, Osaka Investigator in Fujitec R & D Co., Ltd. Yuji Shimizu (56) References JP-A-3-31173 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) B66B 1/18-1/20

Claims (2)

(57) [Claims] When a plurality of elevators are put into service on a plurality of floors and a hall call occurs, various data representing the state of the elevator system at that time are converted into a form that can be used as an input pattern of a neural network. In the elevator group management control device, which selects the most suitable car from the output pattern and assigns it to the hall call, a learning sample is created during operation of the elevator. A neural network learning device for performing a net learning process is provided independently of a neural allocation device for performing an allocation process using a neural network, and the normal operation of the neural allocation device is monitored, and an abnormality is detected. Sometimes the assignment program in the neural assignment device is Tsu DOO learning device group management control device for an elevator, characterized in that it includes an allocation monitoring device for downloaded. 2. The elevator group management control device according to claim 1, wherein the allocation program can be downloaded by a command from the allocation monitoring device.