KR102375710B1 - Artificial intellignece-based optimal distribution system and method thereof - Google Patents

Abstract

A method for an artificial intelligence-based optimal dispatch system according to an embodiment includes: generating reference dispatch information for planned dispatch based on dispatch reference information and learning data; determining whether the expected arrival time of the bus operating the first operating round exceeds the reference dispatch time according to the reference dispatch schedule information; When it is determined that the expected arrival time of the first operating round exceeds the standard dispatch time for the second operating round according to the reference schedule information, the dispatch interval of the next operating round is adjusted based on the adjustment time according to the automatic dispatch adjustment calculation method to do; updating reference dispatch information based on the adjusted dispatch time; and transmitting the updated reference dispatch information to the driver's smartphone.

Description

AI-based optimal dispatch system and method for the same

The present specification relates to an artificial intelligence-based optimal dispatch system and a method therefor, and more particularly, to an artificial intelligence-based optimal dispatch system capable of adjusting dispatch intervals in real time and a method therefor.

In the case of buses, there is a problem that not only the dispatch interval is generally longer than that of the subway, etc., but also the bus dispatch interval may be delayed compared to the schedule in case of traffic congestion. In particular, there was a problem in that the schedule of the bus schedule and the actual road conditions and passenger demand did not match due to the bus dispatch schedule according to the arbitrariness of the conventional bus dispatch scheduler. In addition, there was an inconvenience in that the dispatcher of the bus company had to manually fill out the schedule every day.

On the other hand, a recurrent neural network (　RNN) is a type of artificial neural network, and has a characteristic that the connection between units has a cyclic structure. Recurrent artificial neural networks (RNNs) are widely used in data processing with time-varying characteristics, such as time series analysis or natural language processing, because they can process a sequence-type input using an internal memory.

An object of the present specification is to provide an artificial intelligence-based optimal dispatch system capable of adjusting the bus dispatch interval in consideration of road conditions and bus operation conditions, and a method therefor.

A method for an artificial intelligence-based optimal dispatch system according to an embodiment includes: generating reference dispatch information for planned dispatch based on dispatch reference information and learning data; determining whether the expected arrival time of the bus operating the first operating round exceeds the reference dispatch time according to the reference dispatch schedule information; When it is determined that the expected arrival time of the first operating round exceeds the standard dispatch time for the second operating round according to the reference schedule information, the dispatch interval of the next operating round is adjusted based on the adjustment time according to the automatic dispatch adjustment calculation method to do; updating reference dispatch information based on the adjusted dispatch time; and transmitting the updated reference dispatch information to the driver's smartphone.

According to the present embodiment, an artificial intelligence-based optimal dispatch system capable of adjusting the bus dispatch interval in consideration of road conditions and bus operation conditions and a method therefor may be provided.

1 is an overall configuration diagram of a bus operation information guide system including an artificial intelligence-based optimal dispatch system according to an embodiment of the present invention.
2 is a diagram for explaining a basic concept of an RNN.
3 is a diagram showing various implementation methods of an RNN.
4 is a block diagram for explaining Long Short Term Memory networks (LSTMs).
5 is a diagram for explaining a method of performing learning and prediction based on LSTM according to an embodiment of the present invention.
6 is a view for explaining the optimal dispatch by the artificial intelligence-based optimal dispatch system according to the present embodiment.
7 is a view for explaining a process in which the AI-based optimal dispatch system according to the present embodiment is applied to one bus route.
8 is a diagram for explaining a process of planning dispatch and dispatch time adjustment by the AI-based optimal dispatch system according to the present embodiment.
9 is a flowchart illustrating a method for an AI-based optimal dispatch system according to an embodiment of the present invention.
10 is a flowchart for explaining a method for an AI-based optimal dispatch system according to another embodiment of the present invention.

All of the foregoing characteristics and the following detailed description are exemplary matters for helping the description and understanding of the present specification. That is, the present specification is not limited to such an embodiment and may be embodied in other forms. The following embodiments are merely examples for fully disclosing the present specification, and are descriptions for conveying the present specification to those skilled in the art to which this specification belongs. Therefore, when there are several methods for implementing the elements of the present specification, it is necessary to make it clear that the implementation of the present specification is possible in any one of these methods or an equivalent thereto.

In the present specification, when it is stated that a configuration includes specific elements, or when a process includes specific steps, it means that other elements or other steps may be further included. That is, the terms used in this specification are only for describing specific embodiments, and are not intended to limit the concepts of the present specification. Furthermore, the examples described to help the understanding of the invention also include complementary embodiments thereof.

Terms used in this specification have meanings commonly understood by those of ordinary skill in the art to which this specification belongs. Commonly used terms should be interpreted in a consistent meaning according to the context of the present specification. In addition, the terms used in this specification should not be construed in an overly idealistic or formal meaning unless the meaning is clearly defined. Hereinafter, embodiments of the present specification will be described with reference to the accompanying drawings.

1 is an overall configuration diagram of a bus operation information guide system including an artificial intelligence-based optimal dispatch system according to an embodiment of the present invention.

Referring to FIG. 1 , the bus operation information guide system 10 according to the present embodiment includes a bus 11 , a Global Positioning System (GPS) satellite 12 , a base station 12a, a mobile communication service system 13 , It may include a Bus Information System (hereinafter, BIS) center 14 , a bus control system 15 , and an AI-based optimal dispatch system 100 .

For example, the bus 11 is generally a transportation means for transporting passengers to a destination, and a bus operation guide device (not shown) may be provided.

The GPS satellite 12 is a device for measuring the position of a moving object on the earth by calculating the distance and distance change speed. Location information of a moving object (eg, the bus 11) may be acquired by four or more GPS satellites 12 .

The base station 12a and the mobile communication system 13 include voice data, text data, and image data received from a bus operation guide device (not shown) provided in the bus 11 or a driver's cell phone (eg, APP) through a communication network. , the location data may be transmitted to the BIS center 14 .

On the other hand, the base station 12a and the mobile communication system 13 transmit the voice data, text data, and optimal dispatch schedule information received from the BIS center 14 to the bus operation guide device of the bus 11 or the driver's cell phone (eg, APP). ) can be sent.

The BIS center 14 transmits voice data, text data, image data, and location data received from a bus operation guide device (not shown) provided in the bus 11 or the driver's cell phone (eg, APP) to the bus 11 . It can transmit to the managed bus control center 15.

In addition, the estimated arrival time data of the corresponding bus 11 according to the optimal dispatch schedule information generated based on the received location data and the arrival time calculation algorithm is transmitted through a communication network to a stop terminal installed at the expected arrival stop of the corresponding bus 11 ( 16) can be transmitted.

The bus control center 15 receives voice data, text data, and location data from the BIS center 14, and uses the received related data in a predetermined way (eg, displaying the current location of the bus on a map image). The current location of the bus, traffic conditions, etc. can be displayed on the monitor.

The AI-based optimal dispatch system 100 may receive public data related to bus operation based on an API (Application Programming Interface) from the BIS center 14 . For example, the public data may include information related to departure times and arrival times of bus stops for each bus route.

The artificial intelligence-based optimal dispatch system 100 may adjust the bus dispatch interval in consideration of road conditions and bus operation conditions based on artificial intelligence learned based on previously acquired public data.

For example, the previously obtained public data may be learned by the artificial intelligence-based optimal dispatch system 100 using a recurrent artificial neural network (RNN) technique, which will be described later.

The optimal dispatch arrangement information or alarm information generated by the artificial intelligence-based optimal dispatch system 100 may be transmitted to a bus operation guide device (not shown) provided in the bus 11 or a mobile phone (eg, APP) of the driver. .

Hereinafter, an artificial intelligence-based optimal dispatch system according to this embodiment will be described in more detail with reference to related drawings to be described later.

2 is a diagram for explaining a basic concept of an RNN.

Referring to FIG. 2 , the hidden state (t) of the current state may be updated based on the hidden state (t-1) of the previous time point. Meanwhile, the ouput of the current state may be updated based on the hidden state(t) of the current state. For example, the activation function for the hidden state may be a hyperbolic tangent (tanh) that is a nonlinear function.

3 is a diagram illustrating various implementation methods of an RNN.

For example, the One to One method is a method of generating output data based on data received at time t. An example of the One to One method may be image classification.

For example, the one-to-many method is a method of generating a plurality of output data based on one data received at time t. An example of the one-to-many method may be a case of reading an image and checking whether various symbols (person, dog, etc.) are included in the image.

For example, the Many to One method is a method of generating one output data based on a plurality of data received at a plurality of points in time. An example of the many-to-one method may be a case of performing sentiment analysis based on sequence data such as text.

For example, the Many to Many method is a method applied to the present embodiment and is a method of generating a plurality of data based on a plurality of data received at a plurality of times.

The implementation method according to the present specification may include unidirectional, bidirectional, many-to-one/many-to-many, and multi-layered LSTM (stacked LSTM). They can be configured in various ways by combining them with each other.

4 is a block diagram for explaining Long Short Term Memory networks (LSTMs).

In this specification, LSTM (Long Short Term Memory networks) may be understood as a special kind of RNN.

Referring to FIG. 4 , the LSTM may be implemented as a structure in which four interactive layers are repeated instead of a single neural network structure.

For example, each line in FIG. 4 means a vector, and each output value may be transmitted as an input value of another node.

For example, a square symbol in FIG. 4 may indicate a unit of a neural network, and a circle symbol in FIG. 4 may indicate a configuration in which a point-by-point operation (eg, a vector sum operation) is performed.

Meanwhile, the horizontal line at the top of the block diagram of FIG. 4 is the cell state of the LSTM, and information output from each step may pass through the entire chain and be continuously transmitted to the next step.

For example, the LSTM may selectively add or remove output information based on a gate added to the cell state. As an example, the gates may be implemented using a sigmoid neural net layer and a point-by-point product operation. A sigmoid neural net can output a value of 0 or 1, and the value can be used to determine how much influence each component has on future results.

5 is a diagram for explaining a method of performing learning and prediction based on LSTM according to an embodiment of the present invention.

1 to 5 , LSTM may be applied to data of the previous day to generate data of the next day. In this case, back propagation of the LSTM may be applied to the data of the next day.

Meanwhile, LSTM may be applied to the previous day's data to generate predicted data for the next day.

6 is a view for explaining the optimal dispatch by the artificial intelligence-based optimal dispatch system according to the present embodiment.

For a clear and concise understanding of the present embodiment, it may be assumed that 40 buses belonging to a bus company operating one route operate five times a day and operate a total of 200 times.

Referring to Fig. 6 (a), the horizontal axis of Fig. 6 (a) is the operating cycle of one route, and the vertical axis of Fig. 6 (a) is the operating time from the garage to the last stop corresponding to the operating cycle. can be understood

In this case, the P1 driving cycle, which takes the longest operating time, may be associated with the work time, and the P2 driving cycle may be associated with the leaving work time.

Referring to (b) of FIG. 6 , it may be understood that the horizontal axis of FIG. 6(b) is the operation cycle of one route, and the vertical axis of FIG. 6(b) is the dispatch interval according to the operation cycle.

In this case, it may be desirable to set the shortest dispatch interval to the P1 driving cycle and the P2 driving cycle in which the driving time is the longest. That is, the ideal dispatch interval graph of FIG. 6(b) may appear in the opposite direction to the running time graph of FIG. 6(a).

On the other hand, it is common that the actual dispatch interval of the bus company shows the same trend as the operation time graph of FIG. 6B .

The graph according to the artificial intelligence optimal dispatch of FIG. 6B according to the present embodiment may be generated based on a value between the actual dispatch and the ideal dispatch.

7 is a view for explaining a process in which the AI-based optimal dispatch system according to the present embodiment is applied to one bus route.

1 to 7 , it may be assumed that one bus route includes first to eighth stops TM_1 to TM_8. In addition, each of the buses B#1, ..., B#6 may be operated according to a predetermined dispatch interval.

The first stop TM_1 of FIG. 7 may be understood as a garage, which is the starting point of the corresponding bus route. Also, the first stop TM_1 of FIG. 7 may be understood as a reference point of dispatch time information according to a planned dispatch.

The second to seventh stops TM_2 to TM_7 may be understood as intermediate stops at which buses operate. Meanwhile, the eighth stop TM_8 may be understood as an end point of a corresponding bus route. Each of the buses B#1, ..., B#6 may return to the first stop TM_1, which is the garage, through the eighth stop TM_8, which is an end point.

8 is a diagram for explaining a process of planning dispatch and dispatch time adjustment by the AI-based optimal dispatch system according to the present embodiment.

1 to 8 , a plurality of stops included in a specific route may be associated with row information of FIG. 8 , and a bus running on a specific route may be associated with column information of FIG. 8 .

The AI-based optimal dispatch system according to the present embodiment may generate departure time information TB_start for planned dispatch.

That is, the column information corresponding to the first stop TM_1 of FIG. 8 may be understood as departure time information TB_start for scheduled dispatch. For example, the departure time table information (TB_start) may be transmitted through the APP of the assigned driver.

For example, the departure time information (TB_start) for scheduled dispatch may be generated to have a predetermined time interval (eg, an 8-minute interval).

For example, '4:10' included in the departure time table information TB_start of FIG. 8 is to be understood as the time when the first bus B#1 arrives at the first stop TM_1, picks up passengers, and departs. can

Also, '4:18' included in the departure time table information TB_start of FIG. 8 may be understood as the time when the second bus B#2 arrives at the first stop TM_1, picks up passengers, and departs. .

For example, '4:26' included in the departure time table information TB_start of FIG. 8 is to be understood as the time when the third bus B#3 arrives at the first stop TM_1, picks up passengers, and departs. can

Also, '4:34' included in the departure time table information TB_start of FIG. 8 may be understood as the time when the fourth bus B#4 arrives at the first stop TM_1, picks up passengers, and departs. .

For example, '4:42' included in the departure time table information TB_start of FIG. 8 is to be understood as the time when the fifth bus B#5 arrives at the first stop TM_1, picks up passengers, and departs. can

Also, '4:50' included in the departure time table information TB_start of FIG. 8 may be understood as the time when the sixth bus B#6 arrives at the first stop TM_1, picks up passengers, and departs. .

On the other hand, '4:23' corresponding to the column information of the eighth stop TM_8 of FIG. 8 is the starting point after the first bus B#1 arrives at the eighth stop TM_8, which is the last stop, and gets off all passengers. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a first return time T_1 for the first bus B#1 to return to the starting point (ie, TM_1).

In addition, '4:32' corresponding to the column information of the eighth stop TM_8 of FIG. 8 is the starting point after the second bus B#2 arrives at the eighth stop TM_8, which is the last stop, and all passengers are disembarked. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a second return time T_2 for the second bus B#2 to return to the starting point (ie, TM_1).

In addition, '4:40' corresponding to the column information of the eighth stop (TM_8) of FIG. 8 is the starting point after the third bus (B#3) arrives at the eighth stop (TM_8) and disembarks all passengers. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a third recovery time T_3 for the third bus B#3 to return to the starting point (ie, TM_1).

In addition, '4:49' corresponding to the column information of the eighth stop TM_8 of FIG. 8 is the starting point after the fourth bus B#4 arrives at the eighth stop TM_8, which is the last point, and all passengers are disembarked. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a fourth return time T_4 for the fourth bus B#4 to return to the starting point (ie, TM_1).

In addition, '4:57' corresponding to the column information of the eighth stop (TM_5) of FIG. 8 is the starting point after the fifth bus (B#5) arrives at the eighth stop (TM_8), which is the last point, and disembarks all passengers. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a fifth return time T_5 for the fifth bus B#5 to return to the starting point (ie, TM_1).

In addition, '5:05' corresponding to the column information of the sixth stop (TM_6) in FIG. 8 is the starting point after arriving at the eighth stop (TM_8) where the sixth bus (B#6) is the last and disembarking all passengers. (ie, TM_1) can be understood as a departure time to return.

Here, it may take a sixth return time T_6 for the sixth bus B#6 to return to the starting point (ie, TM_1).

The first to sixth return times T_1 to T_6 may be understood as information derived from learning data acquired through artificial intelligence learning based on previously secured data and driving data.

9 is a flowchart illustrating a method for an AI-based optimal dispatch system according to an embodiment of the present invention.

1 to 9 , in step S910, the artificial intelligence-based optimal dispatch system according to this embodiment generates reference dispatch table information 800 for planned dispatch based on dispatch reference information and learning data. can

For example, the dispatch reference information may include vehicle information related to a bus, personnel information related to a bus driver, and operation information. Here, the operation information may include the number of operations of the corresponding bus route, a basic dispatch interval, and a driver's break time.

For example, the learning data may be understood as data generated by performing RNN (or LSTM) learning based on bus operation data related to the stop arrival time through the bus arrival information inquiry service API provided by the public data portal.

For example, the reference schedule information 800 may include departure time information TB_start for the scheduled dispatch of FIG. 8 .

In step S920, the AI-based optimal dispatch system according to the present embodiment may derive the average arrival time required to move from the last stop to the starting point of the garage.

For example, the average arrival time is the starting point (ie, TM_1) at the sixth stop (TM_6), which is the end point of a specific bus (eg, B#1) route in a specific time zone for the operation of a specific bus (eg, B#1) It may mean a return time (eg, T_1) to return to .

Meanwhile, the average arrival time may be information predicted based on operation data accumulated through a bus arrival information inquiry service API provided by a public data portal for a certain period of time.

Furthermore, in order to derive a more accurate average arrival time, APP-based GPS location information installed in the driver's smartphone may be used together. For example, the GPS location information may be used to determine whether a specific bus in a specific time zone running a specific bus route arrives at the garage and the arrival time of the garage.

In other words, more accurate data can be derived by comparing the arrival time data predicted based on the accumulated driving data and the GPS location information based on the APP.

For example, when a data communication failure occurs on the BIS center 14 side, the average arrival time may be derived based on the driver's APP-based GPS location information.

As another example, a case in which a data communication failure occurs on the driver's smartphone side will be described. APP-based GPS information installed in the driver's smartphone may be set to be transmitted at 1-minute intervals by default.

If there is no GPS information received even after the estimated arrival time has elapsed, the dispatch interval of the next running round may be adjusted based on the pre-generated departure time information TB_start for the planned dispatch.

As another example, in the case of a smartphone that does not support transmission and reception of GPS information from the driver's mobile phone, the arrival time of the bus operated by the driver may be set as departure time information TB_start for scheduled dispatch.

It will be understood that step S920 may be selectively performed to improve the accuracy of the AI-based optimal dispatch system according to the present embodiment.

In step S930, the artificial intelligence-based optimal dispatch system according to the present embodiment may determine whether the expected arrival time of the corresponding operation round exceeds a reference dispatch time for scheduled dispatch.

Here, the expected arrival time of the corresponding operating round may mean a time at which the bus operating the corresponding operating round arrives at the garage TM_1 from the end point TM_8.

For example, when the 4th bus (B#4) running the 4th time of FIG. 8 is delayed by 20 minutes from 4:36, which is the standard dispatch time for the second stop (TM_2) due to traffic conditions, etc. The estimated arrival time of bus No. 4 (B#4) may be understood as the sum of 5:09 and the fourth return time (T_4).

In this case, referring to the departure time information TB_start of the reference dispatch table information 800 of FIG. 8 , the expected arrival time of the fourth bus B#4 that has completed the fourth operation is the reference dispatch time for the eighth operation. It may be judged to exceed the time (5:08).

If it is determined that the scheduled arrival time of the corresponding operation round exceeds the reference dispatch time for scheduled dispatch, the procedure proceeds to step S940.

If it is determined that the scheduled arrival time of the corresponding operation round does not exceed the reference dispatch time for scheduled dispatch, the procedure proceeds to step S970.

In step S940, the AI-based optimal dispatch system according to the present embodiment may adjust the dispatch interval of the next running round based on the automatic dispatch adjustment calculation method.

For example, the adjustment time α may be added to the departure time information TB_start from the fifth time of the reference schedule information 800 .

Here, the adjustment time (α) is set to a time value so that the estimated arrival time of the 4th bus (B#4), which has completed the 4th operation, does not exceed the standard dispatch time (5:08) for the 8th operation can be

Further, the adjustment time (α) is the actual arrival time (4:56) at the 2nd stop (TM_2) of the 4th bus (B#4) running the 4th time (4:56) and the 3rd bus running the 3rd time ( It may be set as a time value for adjusting a time difference (28 minutes) between the actual arrival time (4:28) at the second stop TM_2 of B#3).

In step S950, the artificial intelligence-based optimal dispatch system according to the present embodiment may update the pre-generated reference dispatch table information 800 based on the adjusted dispatch time. For example, since the adjustment time α is applied from the fifth cycle of the reference schedule information 800 of FIG. 8 , the entire reference schedule information 800 may be updated. In this case, the updated reference dispatch table information 800 may be referred to as optimal dispatch information in which the operating conditions of the corresponding route are reflected.

In step S960, the AI-based optimal dispatch system according to the present embodiment may transmit optimal dispatch information to the driver's smartphone.

In step S970, if it is determined that the expected arrival time of the corresponding operating round does not exceed the reference dispatch time for the planned dispatch, the artificial intelligence-based optimal dispatch system according to the present embodiment may maintain the pre-generated reference dispatch information. there is.

10 is a flowchart for explaining a method for an AI-based optimal dispatch system according to another embodiment of the present invention.

1 to 10 , it will be understood that the description of step S1010 of FIG. 10 may be replaced with the description of step S910.

In step S1020, the AI-based optimal dispatch system according to another embodiment may determine whether a time difference between reference dispatch information for planned dispatch and GPS location information according to actual operation exceeds a threshold value.

In other words, it may be determined whether a time difference between the time information for each stop according to the reference dispatch information 800 and the actual operation information collected from the bus exceeds a threshold value.

If the time difference between the time information for each stop according to the reference dispatch information 800 and the actual operation information collected from the bus exceeds a threshold value, the procedure proceeds to step S1030.

In step S1030, the AI-based optimal dispatch system according to another embodiment may provide real-time notification information through the driver's APP.

In the detailed description of the present specification, specific embodiments have been described, but various modifications are possible without departing from the scope of the present specification. Therefore, the scope of the present specification should not be limited to the above-described embodiments, but should be defined by the claims and equivalents of the claims of the present invention as well as the claims to be described later.

11: Bus 12: GPS Satellite
12a: base station 13: mobile communication service system
14: BIS sensor 15: bus control center
100: AI-based optimal dispatch system

Claims (5)

In the method performed by the optimal dispatch system based on artificial intelligence,
generating reference schedule information for planned dispatch based on the dispatch reference information and the learning data;
Derive the average arrival time in a specific time period required to return to the garage from the last stop of the bus route operated by the bus operating the first operating round,
The average arrival time is information predicted based on operation data accumulated through a bus arrival information inquiry service API provided by a public data portal; Determining whether or not the predetermined reference dispatch time is exceeded for the second operation round,
the estimated arrival time is obtained by adding up the average arrival time to the time when the bus actually arrived at the last stop;
When it is determined that the expected arrival time of the first operating round exceeds the reference dispatch time for the second operating round, a dispatch interval preset for the reference dispatch schedule information based on an adjustment time according to an automatic dispatch adjustment calculation method Adjust the
The adjustment time is determined in a range in which the estimated arrival time does not exceed the reference dispatch time;
updating the reference dispatch information based on the adjusted dispatch time; and
and transmitting the updated reference dispatch information to the driver's smartphone. According to claim 1,
The dispatch reference information includes vehicle information associated with the bus, the number of times the bus is operated, and a basic dispatch interval of the bus. According to claim 1,
operating the bus route and acquiring GPS location information associated with a plurality of stops corresponding to the bus route;
determining whether a time difference between the GPS location information and the reference timetable information exceeds a preset threshold value; and
If it is determined that the time difference exceeds the threshold, sending an alarm to the smartphone of the driver. delete According to claim 1,
The learning data is data generated by performing RNN learning based on bus operation data associated with a stop arrival time.

Images