Traffic congestion prediction device and traffic congestion prediction method

Abstract

To predict the degree of congestion after the lapse of a fixed time from a specific period based on the degree of congestion in a fixed time basis in the specific period regarding the same specific area.SOLUTION: A data center as a congestion prediction device comprises an execution unit and a storage unit. In the storage unit, a plurality of input parameters are inputted, thereby storing mapping data specifying a map which outputs an output parameter indicating the degree of congestion in a predetermined specific area. The map is learnt by machine learning beforehand. The plurality of input parameters includes a plurality of congestion parameters indicating the degrees of congestion in the specific area on a fixed time basis within a specific period. The output parameter is a parameter indicating the degree of congestion after the lapse of the fixed time with respect to the end of the specific period. The execution unit executes the acquisition of the plurality of input parameters. The execution unit executes output of the output parameter by inputting the acquired input parameters to the map.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a traffic congestion prediction device and a traffic congestion prediction method. [Background technology]

[0002] The traffic congestion prediction device of Patent Document 1 predicts vehicle congestion on an expressway. Specifically, the traffic congestion prediction device of Patent Document 1 divides the expressway into multiple sections in advance along the direction of travel of the expressway. Furthermore, when predicting congestion in a second section adjacent to the first section and located upstream of the first section in the direction of travel, the traffic congestion prediction device of Patent Document 1 acquires the traffic volume, vehicle density, etc. of the first section. The traffic congestion prediction device of Patent Document 1 then predicts the traffic volume of the second section based on the traffic volume, vehicle density, etc. of the first section. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2021-189734 Summary of the Invention [Problem to be solved by the invention]

[0004] The causes of congestion occurring in a particular area include not only congestion in areas adjacent to the particular area, but also past congestion in the same particular area. The congestion prediction device in Patent Document 1 does not pay attention to the above-mentioned relationship, and there is room for improvement. [Means for solving the problem]

[0005] A traffic congestion prediction device for solving the above problem comprises an execution unit and a memory unit, and the memory unit stores mapping data that specifies a mapping that outputs an output variable indicating the degree of traffic congestion within a predetermined specific area when multiple input variables are input, the mapping being learned in advance by machine learning, the multiple input variables including multiple traffic congestion variables that indicate the degree of traffic congestion within the specific area at each fixed time within a specific period, and the output variable being a variable that indicates the degree of traffic congestion after the fixed time has elapsed relative to the end of the specific period, and the execution unit acquires the multiple input variables and outputs the output variable by inputting the acquired input variables into the mapping.

[0006] According to the above configuration, for the same specific area, the degree of congestion after a certain time has elapsed from the end of a specific period is predicted based on the degree of congestion at each fixed time within the specific period. In this way, when predicting the degree of future congestion, the accuracy of the prediction can be improved by referring to the past degree of congestion within the specific period. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram of a traffic congestion prediction system. [Figure 2] FIG. 2 is a sequence diagram showing a process executed by the traffic congestion prediction system. [Figure 3] Schematic diagram of an LSTM neural network. [Figure 4] FIG. 10 is an explanatory diagram relating to date information. [Figure 5] FIG. 10 is an explanatory diagram showing the transition of average vehicle speed predicted by the congestion prediction system. [Figure 6] FIG. 10 is an explanatory diagram relating to the display of a personal terminal. DETAILED DESCRIPTION OF THE INVENTION

[0008] <Outline of the congestion prediction system> An embodiment of the present invention will be described below with reference to Figures 1 to 6. First, a schematic configuration of a traffic congestion prediction system 100 will be described.

[0009] As shown in Fig. 1, the traffic congestion prediction system 100 includes a plurality of vehicles 10. The vehicles 10 are, for example, automobiles owned by users. Note that Fig. 1 shows only one vehicle 10 as a representative.

[0010] The vehicle 10 is equipped with a vehicle speed sensor 31 and a GNSS receiver 34. The vehicle speed sensor 31 detects a vehicle speed SP, which is the speed of the vehicle 10. The GNSS receiver 34 detects position coordinates PC, which are the coordinates of the point where the vehicle 10 is located, by communicating with a GNSS satellite (not shown). Note that GNSS is a so-called global positioning satellite system.

[0011] The vehicle 10 includes a control device 20. The control device 20 acquires signals indicating various values ​​from a vehicle speed sensor 31 and a GNSS receiver 34. The control device 20 includes an execution unit 21, a storage unit 22, and a communication unit 23. The communication unit 23 is capable of wireless communication with devices external to the vehicle 10 via a communication network 200. The storage unit 22 stores information acquired by the control device 20. The storage unit 22 also stores various programs in advance. The execution unit 21 executes various processes by reading the programs from the storage unit 22. An example of the execution unit 21 is a so-called CPU.

[0012] As shown in FIG. 1, the traffic congestion prediction system 100 includes a data center 50. The data center 50 includes an execution unit 51, a storage unit 52, and a communication unit 53. The communication unit 53 is capable of communicating with devices external to the data center 50 via a communication network 200. The storage unit 52 stores information acquired by the data center 50. The storage unit 52 also stores mapping data 52A that defines multiple mappings M in advance. The mappings M are learned in advance through machine learning. The mappings M receive multiple input variables and output output variables that indicate the degree of traffic congestion within a predetermined specific area. The number of mappings M is the same as the number of specific areas. Note that FIG. 1 illustrates only three representative mappings M. A detailed description of the mappings M will be provided later. The storage unit 52 stores various programs in advance. The execution unit 51 executes various processes by reading the programs from the storage unit 52. An example of the execution unit 51 is a so-called CPU. An example of the data center 50 is a so-called server. In this embodiment, the data center 50 functions as a traffic congestion prediction device. The execution unit 51 executes a traffic congestion prediction program stored in the storage unit 52, and the data center 50 executes each process of the traffic congestion prediction method.

[0013] As shown in FIG. 1, the traffic congestion prediction system 100 includes a weather information center 70. The weather information center 70 includes an execution unit 71, a memory unit 72, and a communication unit 73. The communication unit 73 is capable of communicating with devices external to the weather information center 70 via a communication network 200. The memory unit 72 stores information acquired by the weather information center 70. The memory unit 72 also stores various programs in advance. An example of the execution unit 71 is a so-called CPU. An example of the weather information center 70 is a so-called server.

[0014] The execution unit 71 executes various processes by loading programs from the storage unit 72. Specifically, the execution unit 71 acquires weather information by communicating with multiple observation sensors (not shown) at predetermined control intervals. Based on the weather information acquired from the observation sensors, the execution unit 71 acquires current weather information IW and current precipitation information IP for multiple predetermined specified locations. Here, the weather information IW is a numerical value indicating the weather, such as sunny, cloudy, rainy, or snowy. The precipitation information IP is a numerical value indicating the amount of precipitation per unit time. The storage unit 72 stores the current weather information IW and precipitation information IP acquired by the execution unit 71 for each specified location. By repeatedly storing the weather information IW and precipitation information IP, the storage unit 72 stores the weather information IW and precipitation information IP from the present to a specified period prior. The execution unit 71 also acquires weather information by communicating with a weather forecasting satellite (not shown) at predetermined control intervals. The execution unit 71 then predicts future weather information IW and future precipitation information IP for a plurality of predetermined specified locations based on weather information acquired from weather forecasting satellites, etc. The storage unit 72 stores the future weather information IW and precipitation information IP predicted by the execution unit 71 for each specified location.

[0015] As shown in Fig. 1, the traffic congestion prediction system 100 includes a plurality of personal terminals 90. The personal terminals 90 are, for example, smartphones owned by users. Note that Fig. 1 shows only one representative personal terminal 90.

[0016] The personal terminal 90 comprises an execution unit 91, a memory unit 92, a communication unit 93, and a display 94. The communication unit 93 is capable of communicating with devices external to the personal terminal 90 via the communication network 200. The memory unit 92 stores information acquired by the personal terminal 90. The memory unit 92 also stores various programs in advance. The execution unit 91 executes various processes by reading the programs from the memory unit 92. An example of the execution unit 91 is a so-called CPU. The display 94 is capable of displaying various pieces of information. The display 94 is also a so-called touch panel display. Therefore, the user can also input various pieces of information via the display 94.

[0017] <Acquisition Control> Next, the acquisition control executed by the data center 50, the control devices 20 of the multiple vehicles 10, and the weather information center 70 will be described.

[0018] As shown in FIG. 2, during acquisition control, the control device 20 of the vehicle 10 transmits a signal indicating the vehicle speed SP and the position coordinate PC to the data center 50 at each predetermined control cycle. At this time, the control device 20 transmits the vehicle speed SP and the position coordinate PC as a set of data. Note that while FIG. 2 illustrates one vehicle 10 transmitting the vehicle speed SP, etc. to the data center 50, in reality, multiple vehicles 10 transmit the vehicle speed SP, etc. to the data center 50. The data center 50 receives the signal indicating the vehicle speed SP and the position coordinate PC. Furthermore, the execution unit 51 of the data center 50 determines which of multiple predetermined specific areas the received position coordinate PC belongs to. When a position coordinate PC belonging to a specific area exists, the execution unit 51 of the data center 50 determines the vehicle speed SP paired with the position coordinate PC as the vehicle speed SP for the specific area. Then, the execution unit 51 of the data center 50 associates the vehicle speed SP and the signal indicating the vehicle speed SP with the date and time of reception, and stores them as information about the specific area. Here, an example of a specific area is a road in the same traveling direction, divided into areas of several tens to several hundreds of meters along the traveling direction.

[0019] Meanwhile, in the acquisition control, the weather information center 70 transmits signals indicating weather information IW and precipitation information IP for each specified point to the data center 50 at each predetermined control cycle. The data center 50 receives the signals indicating the weather information IW and precipitation information IP. The execution unit 51 of the data center 50 also selects, for each specific area, the specified point closest to the specific area. The memory unit 52 of the data center 50 then stores the weather information IW and precipitation information IP for the selected specified point in association with the specific area of ​​interest.

[0020] <Predictive Control> Next, the predictive control executed by the data center 50 will be described. The execution unit 51 of the data center 50 executes predictive control multiple times for each specific area at each predetermined timing. By executing predictive control once for the specific area that is the target of prediction, the execution unit 51 can predict one average vehicle speed SPA as a variable indicating the degree of congestion. Then, by repeatedly executing predictive control, the execution unit 51 can predict the average vehicle speed SPA for a predetermined period. An example of the predetermined timing is a specific time on a specific day of the week, such as "3:00 p.m. on Wednesday." An example of the predetermined period is two weeks. First, the first predictive control that the data center 50 executes for the specific area that is the target of prediction will be described.

[0021] As shown in Fig. 2, when the execution unit 51 of the data center 50 starts predictive control, it executes the process of step S11. In step S11, the execution unit 51 acquires the average vehicle speed SPA for each fixed time period within a specific period for the specific area that is the target of prediction. In this embodiment, the length of the specific period is 3 hours. The length of the fixed time period is 15 minutes.

[0022] Specifically, when the execution unit 51 executes predictive control at 3:00 PM, the execution unit 51 sets 3:15 PM, a fixed time after 3:00 PM, as the time to be predicted. In this case, the execution unit 51 accesses the storage unit 52 to acquire multiple vehicle speeds SP from after 12:00 PM until before 3:00 PM for the specific area to be predicted. The execution unit 51 then calculates an average vehicle speed SPA, which is the average value of the multiple vehicle speeds SP, at predetermined fixed time intervals. For example, the execution unit 51 sets the average value of the multiple vehicle speeds SP from after 12:00 PM until before 12:15 PM for the specific area as the average vehicle speed SPA at 12:15 PM. In a similar manner, the execution unit 51 calculates the average vehicle speed SPA up to 3:00 PM in 15-minute increments. Therefore, the execution unit 51 acquires a total of 12 average vehicle speeds SPA. The storage unit 52 stores the total of 12 average vehicle speeds SPA calculated by the execution unit 51.

[0023] In this embodiment, the start of the specific period is 12:00, which is 3 hours and 15 minutes before the target prediction time of 15:15. The end of the specific period is 15:00, which is 15 minutes before the target prediction time of 15:15. Therefore, the target prediction time is a time a certain time after the end of the specific period.

[0024] The execution unit 51 also acquires weather information IW and precipitation information IP for a specific area to be predicted at regular intervals within a specific period by accessing the storage unit 52. That is, the execution unit 51 acquires information for every 15 minutes, such as the weather information IW and precipitation information IP for 12:15, the weather information IW and precipitation information IP for 12:30, and so on.

[0025] As shown in FIG. 4, the execution unit 51 acquires a date information ID indicating the date corresponding to each time in 15-minute increments within the specific period. Specifically, the execution unit 51 first associates the date corresponding to each time with the map shown in FIG. 4. This map indicates which week and which day of the week each time corresponds to. For example, assume that the 1st day of month X is a Wednesday. In this case, the execution unit 51 acquires "4" as the date information ID corresponding to the 1st day of month X. Then, from the 2nd day of month X onwards, the execution unit 51 acquires date information IDs that are incremented by one, such as "5," "6," and "7," as the date information IDs for each day that passes. Note that in the above example, the period from 12:00 to 15:00 does not span multiple dates. Therefore, in the above example, the execution unit 51 acquires the same date information ID for all times.

[0026] The execution unit 51 acquires time information IT indicating the time corresponding to each time. Here, the time information IT is obtained by converting a time, such as 12:15, into a numerical value according to a predetermined rule.

[0027] The execution unit 51 acquires calendar information IC for the date corresponding to each time. Here, the calendar information IC is a numerical value indicating information such as weekdays, holidays, and consecutive holidays. In this way, the execution unit 51 acquires six types of information for each of the 12 times in 15-minute increments. In this embodiment, the process of step S11 is a process of acquiring multiple input variables. As shown in FIG. 2, after step S11, the execution unit 51 proceeds to step S12.

[0028] As shown in FIG. 2, in step S12, the execution unit 51 generates various values ​​acquired in the processing of step S11 as input data X(1) to input data X(N) for the mapping M. Here, each input data includes input variables corresponding to six types of information for the corresponding time. Furthermore, "N" is an integer equal to or greater than 2. "N" is a value that is incremented by "1" every time a certain time elapses, with the oldest data as a reference. Therefore, in the above example, input data X(1) is data at 12:15, and input data X(N) is data at 3:00 p.m. As described above, six types of information are acquired for each of the 12 times. Therefore, in this embodiment, N=12. Furthermore, "N" is the number of congestion variables, which will be described later.

[0029] Here, as a specific example of the processing of step S12, a case where the target predicted time is 15:15 will be described. In this case, the execution unit 51 assigns the average vehicle speed SPA, weather information IW, precipitation information IP, date information ID, time information IT, and calendar information IC at 12:15 as input variables of the input data X(1). Similarly, the execution unit 51 assigns information for each time as input variables of the input data X(2) to input data X(N-1). Then, the execution unit 51 assigns the average vehicle speed SPA, weather information IW, precipitation information IP, date information ID, time information IT, and calendar information IC at 15:00 as input variables of the input data X(N).

[0030] Each of the average vehicle speeds SPA substituted in step S12 is an example of a congestion variable indicating the degree of congestion in a specific area at each fixed time within a specific period. Each of the average vehicle speeds SPA substituted in step S12 is an example of a variable indicating the speed of the vehicle 10. Each of the date information ID, time information IT, and calendar information IC substituted in step S12 is an example of a variable indicating the date and time in a specific area within a specific period. Each of the weather information IW and precipitation information IP substituted in step S12 is an example of a variable indicating the weather in a specific area within a specific period. After step S12, the execution unit 51 proceeds to step S13.

[0031] In step S13, the execution unit 51 outputs an output variable y(N+1) by inputting the multiple input variables generated in the processing of step S12 into the mapping M. The output variable y(N+1) indicates a predicted value of the average vehicle speed SPA after a certain time has elapsed since the end of the specific period. As a specific example, if the end of the specific period is 3:00 p.m., the output variable y(N+1) indicates a predicted value of the average vehicle speed SPA at 3:15 p.m. The average vehicle speed SPA output in step S13 is an example of a variable indicating the degree of congestion. Furthermore, the average vehicle speed SPA output in step S13 is an example of a variable indicating the speed of the vehicle 10.

[0032] As shown in FIG. 3 , an example of the mapping M is a so-called LSTM neural network, which is a machine-learned model capable of handling time-series data while retaining information on its transition along the time axis. Therefore, in this embodiment, the mapping M includes a recurrent neural network that outputs an output variable in response to input of multiple congestion variables within a specific period. This LSTM neural network is a type of recurrent neural network in which a mechanism called an LSTM block is provided in each hidden layer of the recurrent neural network to adjust the propagation of time-series information. Specifically, as shown in FIG. 3 , the mapping M is configured such that the output of the hidden layer connected to the input layer to which the input data X(N-1) of the immediately preceding time is input is reflected in the hidden layer connected to the input layer to which the input data X(N-1) of the immediately preceding time is input. Similarly, the mapping M is configured such that the output of the hidden layer connected to the input layer to which the input data X(N-2) of the immediately preceding time is input is reflected in the hidden layer connected to the input layer to which the input data X(N-1) of the immediately preceding time is input. Note that LSTM is also known as Long Short-Term Memory.

[0033] In this embodiment, the mapping M is learned, for example, as follows. First, input data X(1) to input data X(N) are acquired in the manner of steps S11 and S12 described above. Furthermore, an actual average vehicle speed SPA is acquired in the manner of step S11 described above as an output variable y(N+1) corresponding to the input data X(1) to input data X(N). Furthermore, a set of training data is generated by combining the acquired input data X(1) to input data X(N) and the actual average vehicle speed SPA as a correct label. Then, the mapping M is learned by machine learning based on the plurality of training data.

[0034] The storage unit 52 of the data center 50 stores the predicted average vehicle speed SPA in association with the target prediction time and the target specific area. In this embodiment, the processes of steps S12 and S13 are processes for inputting the acquired input variables into the mapping M to output the output variables. After step S13, the execution unit 51 ends the current prediction control.

[0035] Furthermore, after the first predictive control described above, the execution unit 51 can execute second or subsequent predictive control for the same specific area. In step S11 of the second or subsequent predictive control, the execution unit 51 accesses the storage unit 52 to acquire the average vehicle speed SPA calculated by the execution unit 51 in step S11 of the first predictive control as the average vehicle speed SPA within the specific period. At this time, if there is no average vehicle speed SPA calculated by the execution unit 51 in a previous step S11 as the average vehicle speed SPA within the specific period, the execution unit 51 acquires the average vehicle speed SPA predicted in step S13. For example, if the prediction time targeted in the second predictive control is 15:30, the average vehicle speed SPA predicted in step S13 of the first predictive control is acquired as the average vehicle speed SPA at 15:15.

[0036] Then, in step S12 of the second predictive control, the execution unit 51 assigns the average vehicle speed SPA, weather information IW, precipitation information IP, date information ID, time information IT, and calendar information IC at 12:30 as input variables of the input data X(1). Similarly, the execution unit 51 assigns the average vehicle speed SPA, weather information IW, precipitation information IP, date information ID, time information IT, and calendar information IC at 15:15 as input variables of the input data X(N). As described above, the average vehicle speed SPA at 15:15 is the average vehicle speed SPA predicted in step S13 in the first predictive control. Furthermore, the weather information IW and precipitation information IP at 15:15 are the weather information IW and precipitation information IP predicted by the weather information center 70. After performing the second predictive control in this manner, the execution unit 51 can further perform third and fourth predictive control.

[0037] Here, the specific period corresponding to the first output variable is referred to as the first specific period, and the specific period corresponding to the second output variable is referred to as the second specific period. The start of the first specific period is referred to as the first start period, and the end of the first specific period is referred to as the first end period. In this case, the start of the second specific period is referred to as the second start period, which is a certain amount of time after the first start period. In the above example, the second start period is 12:15, which is 15 minutes after the first start period, 12:00. The end of the second specific period is referred to as the second end period, which is a certain amount of time after the first end period. In the above example, the second end period is 15:15, which is 15 minutes after the first end period, 15:00.

[0038] In this case, the execution unit 51 executes acquiring a plurality of input variables including N average vehicle speeds SPA from a first start time to a first end time within the first specific period. The execution unit 51 also executes inputting the acquired input variables into a mapping M to output the average vehicle speeds SPA, which is the first output variable. The execution unit 51 also executes acquiring a plurality of input variables including N average vehicle speeds SPA from a second start time to a second end time within the second specific period, with the average vehicle speeds SPA, which is the first output variable, being the average vehicle speed SPA for the second end time. The execution unit 51 then executes inputting the acquired input variables into the mapping M to output the average vehicle speeds SPA, which is the second output variable.

[0039] In this way, by using the average vehicle speed SPA output in the previous predictive control as the average vehicle speed SPA at the end of the specific period in the next predictive control, the execution unit 51 can repeatedly execute predictive control for a third time, a fourth time, etc. In other words, when "L" is an arbitrary positive integer, the execution unit 51 can repeatedly execute the Lth predictive control, the L+1th predictive control, and so on. Then, by the execution unit 51 repeatedly executing predictive control in this way, a transition in the predicted value of the average vehicle speed SPA as shown in FIG. 5 is obtained.

[0040] <Distribution Control> Next, the distribution control executed by the data center 50 and the personal terminal 90 will be described. The distribution control is executed between one data center 50 and multiple personal terminals 90. The personal terminal 90 executes the distribution control each time a user performs an operation to request distribution via the display 94 of the personal terminal 90. Here, the operation to request distribution includes an operation in which the user selects a specific area and a requested period. As shown in FIG. 2, in the distribution control, the personal terminal 90 transmits a signal requesting distribution to the data center 50 together with a signal indicating the specific area and requested period selected by the user.

[0041] The execution unit 51 of the data center 50 transmits a signal indicating the average vehicle speed SPA predicted by the predictive control to the personal terminal 90. Specifically, the execution unit 51 selects multiple specific areas located around a specific area selected by the user. Then, the execution unit 51 transmits a signal indicating the average vehicle speed SPA corresponding to the requested period selected by the user for the selected multiple specific areas to the personal terminal 90. As shown in FIG. 6, the execution unit 51 of the personal terminal 90 displays a map showing the specific area selected by the user and its surroundings on the display 94. At this time, the execution unit 51 displays the specific area selected by the user as, for example, a black circle. The execution unit 51 also displays information indicating a traffic congestion forecast for the requested period selected by the user on the display 94. At this time, if the average vehicle speed SPA for each time included in the requested period is less than a predetermined specified speed, the execution unit 51 displays, for example, a black square, indicating that there is traffic congestion. Here, an example of the specified speed is 10 km / h. Then, the personal terminal 90 terminates the current distribution control.

[0042] <Operation of this embodiment> In traffic congestion within a specific area, the degree of traffic congestion after a certain time tends to change depending on the degree of traffic congestion before that time. In response to this, the execution unit 51 of the data center 50 outputs an output variable by inputting input variables including a plurality of average vehicle speeds SPA for each fixed time period within a specific period before the prediction time into the mapping M. In other words, the data center 50 predicts the average vehicle speed SPA for the same area after a fixed time has elapsed from the end of the specific period based on the average vehicle speeds SPA for each fixed time period within the specific period.

[0043] <Effects of this embodiment> (1) In this embodiment, when predicting the degree of future congestion as described above, the degree of congestion within a specific period for the same specific area is referenced. This improves the accuracy of prediction of the degree of congestion compared to, for example, a configuration that does not reference the degree of congestion within a specific period.

[0044] (2) When the execution unit 51 executes predictive control for the same specific area for the second or subsequent time, the execution unit 51 accesses the storage unit 52 to acquire the average vehicle speed SPA predicted in step S13 in the past predictive control as the average vehicle speed SPA within the specific period. Then, the execution unit 51 outputs an output variable by inputting input variables including the average vehicle speed SPA predicted in step S13 in the past predictive control into the mapping M. That is, in this embodiment, the output variable output in the past predictive control is used as an input variable in the subsequent predictive control. Therefore, each time the predictive control is repeated, a predicted value of the average vehicle speed SPA after a certain time has elapsed is obtained. As a result, by repeating the prediction multiple times, it is possible to predict the extent of long-term congestion.

[0045] (3) In this embodiment, the variable indicating the degree of congestion, which is the output variable of the mapping M, is the average vehicle speed SPA, i.e., a variable indicating the speed of the vehicle 10. This makes it easier to grasp the degree of congestion in more detail than, for example, when a variable indicating the presence or absence of congestion is output as an output variable.

[0046] (4) In this embodiment, the map M includes a recurrent neural network that outputs an output variable in response to input of multiple congestion variables within a specific period. This allows the output variable of the map M to take into account the history of multiple congestion variables included within the specific period. As a result, the prediction accuracy can be further improved.

[0047] (5) Regarding traffic congestion within a specific area, the degree of traffic congestion at a predicted time after a specific period may vary depending on the date and time of the specific period. For example, the degree of traffic congestion at a predicted time may vary depending on the day of the week in the specific period. Also, for example, the degree of traffic congestion at a predicted time may vary depending on the time of day in the specific period. Furthermore, for example, the degree of traffic congestion at a predicted time may vary depending on whether the date in the specific period is a weekday or a public holiday.

[0048] In this embodiment, the input variables of the mapping M include date information ID, time information IT, and calendar information IC, which are variables indicating the date and time within a specific area within a specific period. This can improve the prediction accuracy compared to a configuration that does not take into account, for example, variables indicating the date and time within a specific area within a specific period.

[0049] (6) Regarding traffic congestion within a specific area, the degree of traffic congestion at a predicted time after a specific period may change depending on the weather during that period. For example, the degree of traffic congestion at a predicted time may change depending on whether the weather during the specific period is sunny, cloudy, rainy, or snowy. Also, for example, the degree of traffic congestion at a predicted time may change depending on the amount of precipitation during the specific period.

[0050] In this embodiment, the input variables of the mapping M include weather information IW, which is a variable indicating the weather in a specific area within a specific period, and precipitation information IP. This can improve the prediction accuracy compared to a configuration that does not take into account, for example, a variable indicating the weather in a specific area within a specific period.

[0051] <Example of change> This embodiment can be modified as follows: This embodiment and the following modifications can be combined and implemented within the scope of technical compatibility.

[0052] In the above embodiment, the input variables of the mapping M may be changed. For example, other variables may be used instead of or in addition to the average vehicle speed SPA as the congestion variable indicating the degree of congestion, which is an input variable of the mapping M. As a specific example, the traffic volume within a specific area for each fixed time period within a specific period may be used as the congestion variable, which is an input variable of the mapping M. Note that traffic volume is the number of vehicles 10 passing through the specific area within a fixed time period. Also, as a specific example, the vehicle density within a specific area for each fixed time period within a specific period may be used as the congestion variable, which is an input variable of the mapping M. Note that vehicle density is the average value of the number of vehicles 10 located within the specific area within a fixed time period.

[0053] For example, the variables indicating the date and time in a specific area within a specific period as input variables of the mapping M may be changed or omitted. As a specific example of change, one or two types of information from among date information ID, time information IT, and calendar information IC may be used as variables indicating the date and time in a specific area within a specific period.

[0054] For example, the variable indicating the weather in a specific area within a specific period as an input variable of the mapping M may be changed or omitted. As a specific example of a change, one of the weather information IW and precipitation information IP may be used as the variable indicating the weather in a specific area within a specific period.

[0055] In the above embodiment, the configuration regarding the specific period may be changed. For example, the length of the specific period may be changed. As an example, the length of the specific period may be changed within a range of several tens of minutes to several days.

[0056] For example, the length of the certain period of time may be changed. As an example, the length of the certain period of time may be changed within a range of several minutes to several tens of minutes. Here, the shorter the length of the certain period of time, the easier it is to reflect fluctuations in congestion variables, but on the other hand, the fluctuations in congestion variables are more likely to become excessively large. Therefore, it is preferable to determine an appropriate length of the certain period of time based on experiments, simulations, etc. Note that when the length of the certain period of time or the length of the specific period of time is changed, it is sufficient to change the number of congestion variables input to the mapping M.

[0057] In the above embodiment, the specific area may be changed. Specifically, even if a road has multiple lanes in the same traveling direction, the degree of congestion may differ for each lane. Therefore, the specific area may be, for example, a road in the same traveling direction that is divided into regions of several tens to several hundreds of meters along the traveling direction, and each region may be divided into regions for each lane.

[0058] In the above embodiment, the output variables of the mapping M may be changed. For example, other variables may be used instead of the average vehicle speed SPA as the variable indicating the degree of congestion, which is the output variable of the mapping M. As a specific example, the traffic volume within a specific area may be used as the output variable of the mapping M. Also, as a specific example, the vehicle density within a specific area may be used as the output variable of the mapping M.

[0059] In the above embodiment, the mapping M may be changed. For example, the mapping M may employ a so-called GRU neural network instead of the LSTM neural network. Also, for example, the mapping M may employ a general recurrent neural network that omits the LSTM block instead of the LSTM neural network. Furthermore, for example, the mapping M may include a so-called feedforward neural network instead of or in addition to the recurrent neural network.

[0060] <Other technical ideas> The technical ideas that can be understood from the above-described embodiment and modified examples will be described. (Appendix 1) An execution unit and a storage unit, The storage unit storing mapping data that defines a mapping that outputs an output variable indicating the degree of congestion in a predetermined specific area when a plurality of input variables are input; the mapping is learned in advance by machine learning, the plurality of input variables include a plurality of congestion variables indicating the degree of congestion in the specific area for each fixed time period within a specific period; the output variable is a variable indicating the degree of congestion after the certain time has elapsed from the end of the specific period, The execution unit: obtaining a plurality of said input variables; and outputting the output variables by inputting the acquired input variables into the mapping. Traffic congestion prediction device.

[0061] (Appendix 2) Let "N", the number of congestion variables, be an integer equal to or greater than 2, and let "L" be any positive integer. In a case where the output variable is output a plurality of times, when the specific period corresponding to the Lth output variable is defined as a first specific period, the specific period corresponding to the L+1th output variable is defined as a second specific period, the start of the first specific period is defined as a first start period, and the end of the first specific period is defined as a first end period, the start date of the second specific period is a second start date that is a certain time period after the first start date, the end of the second specific period is a second end after the certain time has elapsed since the first end, The execution unit: acquiring a plurality of the input variables including N congestion variables from the first start time to the first end time within the first specific period; outputting the L-th output variable by inputting the acquired input variable into the mapping; acquiring a plurality of input variables including the N congestion variables from the second start time to the second end time within the second specific period, with the Lth output variable being the congestion variable of the second end time; and outputting the (L+1)th output variable by inputting the acquired input variable into the mapping. 2. The traffic congestion prediction device according to claim 1.

[0062] (Appendix 3) The congestion variable is a variable indicating a vehicle speed, The output variable is a variable indicating the speed of the vehicle. 3. The traffic congestion prediction device according to claim 1 or 2.

[0063] (Appendix 4) The mapping includes a recurrent neural network that outputs the output variable when a plurality of congestion variables within the specific period are input. 4. The traffic congestion prediction device according to claim 1.

[0064] (Appendix 5) The plurality of input variables includes a variable indicating a date and time in the specific area within the specific period as a variable separate from the congestion variable. 5. The traffic congestion prediction device according to any one of Supplementary notes 1 to 4.

[0065] (Appendix 6) The plurality of input variables includes a variable indicating the weather in the specific area within the specific period as a variable separate from the congestion variable. 6. The traffic congestion prediction device according to claim 1, wherein the traffic congestion prediction device is a device for predicting traffic congestion.

[0066] (Appendix 7) A traffic congestion prediction method in a traffic congestion prediction device including an execution unit and a storage unit, The storage unit storing mapping data that defines a mapping that outputs an output variable indicating the degree of congestion in a predetermined specific area when a plurality of input variables are input; the mapping is learned in advance by machine learning, the plurality of input variables include a plurality of congestion variables indicating the degree of congestion in the specific area for each fixed time period within a specific period; the output variable is a variable indicating the degree of congestion after the certain time has elapsed from the end of the specific period, The execution unit, obtaining a plurality of said input variables; and outputting the output variables by inputting the acquired input variables into the mapping. Traffic congestion prediction method. [Explanation of symbols]

[0067] M...Mapping 10...Vehicle 20...Control device 21...Executive Department 22...Storage section 23…Communications Department 31...Vehicle speed sensor 34...GNSS receiver 50...Data center 51...Executive Department 52...Storage section 52A...Mapping data 53…Communications Department 70...Weather Information Center 71...Executive Department 72...Storage section 73…Communications Department 90...Personal devices 91...Executive Department 92...Storage section 93…Communications Department 94...Display 100...Traffic congestion prediction system 200...Communication Network

Claims

1. An execution unit and a storage unit, The storage unit storing mapping data that defines a mapping that outputs an output variable indicating the degree of congestion in a predetermined specific area when a plurality of input variables are input; the mapping is learned in advance by machine learning, the plurality of input variables include a plurality of congestion variables indicating the degree of congestion in the specific area at each fixed time within a specific period, and a variable, separate from the congestion variables, indicating which week and which day of the week of each month a date and time in the specific area within the specific period corresponds to; the output variable is a variable indicating the degree of congestion after the certain time has elapsed from the end of the specific period, The execution unit: obtaining a plurality of said input variables; and outputting the output variables by inputting the acquired input variables into the mapping. Traffic congestion prediction device.

2. Let "N", the number of congestion variables, be an integer equal to or greater than 2, and let "L" be any positive integer. In a case where the output variable is output a plurality of times, when the specific period corresponding to the Lth output variable is defined as a first specific period, the specific period corresponding to the L+1th output variable is defined as a second specific period, the start of the first specific period is defined as a first start period, and the end of the first specific period is defined as a first end period, a start time of the second specific period is a second start time that is a time when the certain period has elapsed from the first start time, an end of the second specific period is a second end when the certain time has elapsed since the first end, The execution unit: acquiring a plurality of input variables including N congestion variables from the first start time to the first end time within the first specific period; outputting the L-th output variable by inputting the acquired input variable into the mapping; acquiring a plurality of input variables including the N congestion variables from the second start time to the second end time within the second specific period, with the Lth output variable being the congestion variable of the second end time; and outputting the (L+1)th output variable by inputting the acquired input variable into the mapping. The traffic congestion prediction device according to claim 1.

3. The congestion variable is a variable indicating a vehicle speed, The output variable is a variable indicating the speed of the vehicle. The traffic congestion prediction device according to claim 1 or 2.

4. The mapping includes a recurrent neural network that outputs the output variable when a plurality of congestion variables within the specific period are input. The traffic congestion prediction device according to claim 1 or 2.

5. The plurality of input variables includes a variable indicating the weather in the specific area within the specific period as a variable separate from the congestion variable. The traffic congestion prediction device according to claim 1 or 2.

6. A traffic congestion prediction method in a traffic congestion prediction device including an execution unit and a storage unit, The storage unit storing mapping data that defines a mapping that outputs an output variable indicating the degree of congestion in a predetermined specific area when a plurality of input variables are input; the mapping is learned in advance by machine learning, the plurality of input variables include a plurality of congestion variables indicating the degree of congestion in the specific area at each fixed time within a specific period, and a variable, separate from the congestion variables, indicating which week and which day of the week of each month a date and time in the specific area within the specific period corresponds to; the output variable is a variable indicating the degree of congestion after the certain time has elapsed from the end of the specific period, The execution unit, obtaining a plurality of said input variables; and outputting the output variables by inputting the acquired input variables into the mapping. Traffic congestion prediction method.

Images