Traffic control device and traffic control method

Abstract

To quickly detect an occurrence of a sudden event on a road.SOLUTION: A road traffic control device according to an embodiment comprises: a traffic congestion prediction model generating unit that generates a traffic congestion prediction model by machine-learning based on teacher data using, as input data, traffic condition data of when a sudden event occurred in the past on a road travelled by the vehicle, and element information being information on elements affecting a traffic condition, and using, as output data, traffic congestion information being information regarding traffic congestion on the road due to the sudden event; and a traffic congestion prediction processing unit that obtains the traffic congestion information of the road as output data based on the traffic congestion prediction model with the traffic condition data of when the sudden event occurred and the element information corresponding to the sudden event being used as input data.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] FIELD An embodiment of the present invention relates to a traffic control device and a traffic control method. [Background technology]

[0002] Forecasting future congestion on roads such as expressways has traditionally been important for vehicle drivers and road operators who perform traffic control operations. Furthermore, when unexpected events occur on roads, such as accidents (traffic accidents) or falling debris, lane closures and road closures may occur, resulting in congestion. In other words, traffic flow may change significantly when unexpected events occur.

[0003] The traffic control center also uses roadside sensor information and roadside CCTV (Closed Circuit Television) images to recognize the occurrence of an unexpected incident, identify the time and location of the incident, and take measures to deal with the accident, such as requesting the police, ambulances, tow trucks, etc., and closing / restricting lanes, as necessary. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2005-149052 [Patent Document 2] Japanese Patent Application Laid-Open No. 2019-200490 [Patent Document 3] Special Publication No. 2009-529186 [Patent Document 4] Japanese Patent Application Publication No. 2019-185232 Summary of the Invention [Problem to be solved by the invention]

[0005] In the above-mentioned conventional technology, it would be useful if the timing of detecting an unexpected incident on the road could be made earlier.

[0006] Therefore, an embodiment of the present invention provides a traffic control device and a traffic control method that can quickly detect the occurrence of an unexpected incident on a road. [Means for solving the problem]

[0007] The traffic control device of the embodiment includes a congestion prediction model generation unit that generates a congestion prediction model by performing machine learning based on teacher data, which takes as input data traffic condition data when an unexpected event occurred in the past for a road on which a vehicle is traveling and element information that is information on elements that the unexpected event affects the traffic condition of the road, and outputs congestion information that is information on congestion on the road due to the unexpected event; and a congestion prediction processing unit that takes as input data traffic condition data when an unexpected event occurred and the element information corresponding to the unexpected event, and obtains congestion information for the road as output data based on the congestion prediction model. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram schematically illustrating a road according to an embodiment. [Figure 2] FIG. 2 is a functional block diagram schematically illustrating the overall configuration of the traffic control system according to the embodiment. [Figure 3] FIG. 3 is an explanatory diagram of a process during learning of the accident detection and prediction model according to the embodiment. [Figure 4] FIG. 4 is an explanation of a part of the processing of the traffic congestion prediction model generating unit of the embodiment. [Figure 5] FIG. 5 is an explanatory diagram of a process during learning of the traffic congestion prediction model according to the embodiment. [Figure 6] FIG. 6 is an explanatory diagram of processing during operation of the accident detection and prediction model according to the embodiment. [Figure 7] FIG. 7 is an explanatory diagram of processing during operation of the traffic congestion prediction model according to the embodiment. [Figure 8] FIG. 8 is a flowchart showing the overall process during operation of the traffic control device of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the traffic control device and traffic control method of the present invention will be described with reference to the drawings. Note that the configurations of the embodiments described below, and the actions and results (effects) brought about by the configurations, are examples, and the present invention is not limited to the following descriptions. In addition, the following description will mainly focus on accidents as examples of unexpected events such as road accidents and falling objects.

[0010] Generally, when an unexpected incident (such as an accident or fallen objects) occurs on a road, some lanes are closed as necessary while the accident is dealt with. If the scale of the unexpected incident is large, the road itself may be closed (i.e., closed to traffic). In this case, if there is more than a certain amount of traffic upstream of the location of the unexpected incident, congestion will occur.

[0011] The congestion that occurs is greatest when roads are closed. Even partial lane closures due to accident clearance work reduce traffic capacity and cause congestion.

[0012] Generally, traffic congestion begins after an accident occurs and continues to grow until the accident response work at the site of the sudden incident is completed. After the accident response work is completed, traffic capacity returns to the level before the accident, and the traffic congestion gradually disappears.

[0013] Therefore, when an accident occurs, by grasping the circumstances of the accident more quickly and in detail, it is possible to "predict the time required to deal with the unexpected incident (accident)," and by properly "predicting the scale of congestion that will be expected once the unexpected incident (accident) handling is completed," it is possible to predict the time until the congestion is resolved, which in turn leads to the prediction of the travel time required for the relevant section. Therefore, below, we will explain the technology for quickly and in detail detecting the occurrence of unexpected incidents on roads.

[0014] FIG. 1 is a diagram schematically illustrating a road R in an embodiment. The road R is, for example, a highway. The road R is divided into sections 1, 2, 3, ... as management units. A vehicle detector 1 and a CCTV camera 2 are installed in each section. In the following, the reference symbol for road R will be omitted and the road will be referred to as "road." Furthermore, although unexpected events include various events such as accidents, fallen objects, and broken-down vehicles, in the following, they will be simply referred to as "accidents" for ease of explanation.

[0015] 2 is a functional block diagram showing a schematic overall configuration of a traffic control system S according to an embodiment. The traffic control system S includes a vehicle detector 1, a CCTV camera 2, a traffic control device 3, and an information output device 4.

[0016] The traffic control device 3 is a computer device, and includes a processing unit 31, a storage unit 32, an input unit 33, a display unit 34, a communication unit 35, and a traffic control information providing system 36.

[0017] The processing unit 31 includes, for example, a central processing unit (CPU), a read-only memory (ROM), and a random access memory (RAM), and executes various processes.

[0018] The CPU comprehensively controls the operation of the traffic control device 3. The COM is a storage medium for storing various programs and data. The RAM is a storage medium for temporarily storing various programs and rewriting various data.

[0019] The MPU uses the RAM as a work area to execute programs stored in the ROM, storage unit 32, etc. Details of the processing unit 31 will be described later.

[0020] The storage unit 32 is a storage device such as a hard disk drive (HDD) or a solid state drive (SSD), and stores various information. The storage unit 32 stores, for example, accident form data 321, an accident detection prediction model 322 (unexpected event detection prediction model), and a traffic congestion prediction model 323.

[0021] The accident report data 321 is report data that records details of accidents that have occurred on roads in the past. The accident report data 321 includes various information such as the location of the accident, the cause of the accident, accident detection, dispatch information of emergency vehicles (ambulances, accident response vehicles, police cars, etc.), the time until the accident response is completed, lane closure status, information on injured people, the presence or absence of fallen objects, and information on whether the accident vehicle can continue driving.

[0022] The accident detection and prediction model 322 detects the occurrence of an unexpected event (accident) from the characteristics of an image of a road. Furthermore, it is a learning model for predicting detailed information about the detected unexpected event (accident) from the image of the road. This accident detection and prediction model 322 can make these predictions by machine learning the records of past accidents recorded in the accident form data 321 (described later in Figures 3 and 6).

[0023] The congestion prediction model 323 is a learning model for predicting the scale of congestion that is expected to occur on roads. The congestion prediction model 323 is a model that predicts the scale of congestion caused by an accident that has occurred, the time until the congestion is resolved, etc., based on traffic condition data (for example, traffic volume, average speed, occupancy rate, etc.) when an accident occurred in the past and detailed information on other accidents (described later in the explanations of Figures 4, 5, and 7).

[0024] The memory unit 32 also stores various information such as road information such as road sections, number of lanes, interchanges, and locations of parking areas, sensor information obtained from the vehicle detector 1, video data obtained from the CCTV camera 2, and various calculation processing results by the processing unit 31.

[0025] The input unit 33 is an input device that accepts user operations on the traffic control device 3, and is, for example, a keyboard, a mouse, or the like.

[0026] The display unit 34 is a display unit located near the input unit 33 and is provided for each traffic controller. The display unit 34 is realized by, for example, a liquid crystal display device (LCD (Liquid Crystal Display)), an organic EL (Electro-Luminescence) display device, or the like.

[0027] The communication unit 35 is a communication interface for communicating with external devices (such as the vehicle detector 1, the CCTV camera 2, and the information output device 4).

[0028] The traffic control information providing system 36 is a system that controls the traffic of expressways and related roads, and displays traffic congestion-related information for target road sections on a large display device 36M.

[0029] The processing unit 31 has, as its functional configuration, an acquisition unit 311, an accident detection prediction model generation unit 312 (unexpected event detection prediction model generation unit), an accident detection prediction processing unit 313 (unexpected event detection prediction processing unit), a congestion prediction model generation unit 314, a congestion prediction processing unit 315, a congestion processing unit 316, and a control unit 317.

[0030] The acquisition unit 311 acquires various types of information from external devices. For example, the acquisition unit 311 acquires sensor information (traffic condition data) from the vehicle detector 1. Then, for example, the acquisition unit 311 creates information such as traffic volume, average speed, and occupancy rate in one-minute or five-minute increments based on the sensor information. The acquisition unit 311 also acquires video data of roads from the CCTV cameras 2.

[0031] The accident detection and prediction model generation unit 312 generates the accident detection and prediction model 322 by performing machine learning on the accident detection and prediction model 322 based on training data that are past accident cases (described later in FIG. 3). Here, the training data used for learning is training data in which road video data when an accident (unexpected event) occurred on the road in the past is used as input data, and accident detail information (element information) that indicates that an accident has occurred and is information on elements related to the accident that affect the traffic situation on the road is used as output data.

[0032] During operation, the accident detection prediction processing unit 313 inputs video data of a road in the event of an accident as input data to the accident detection prediction model 322, and obtains information that an accident has occurred and detailed accident information corresponding to the accident as output data, thereby detecting an accident and predicting detailed information about the accident that has occurred (described later in Figure 6).

[0033] The congestion prediction model generation unit 314 generates the congestion prediction model 323 by performing machine learning based on training data in which traffic congestion information, which is information about traffic congestion on a road due to an accident, is output data (described later in the explanation of FIGS. 4 and 5). Here, the training data is training data in which traffic condition data (for example, traffic volume, average speed, occupancy rate, etc.) when an accident occurred in the past and detailed information about the past accident are input data, and traffic congestion information, which is information about traffic congestion on a road due to an accident, is output data.

[0034] During operation, the traffic congestion prediction processing unit 315 predicts road congestion information using the traffic congestion prediction model 323. The traffic congestion prediction processing unit 315 receives traffic condition data at the time of an accident (sensor information from the vehicle detector 1) and detailed accident information acquired by the accident detection and prediction processing unit 313 as input data for the traffic congestion prediction model 323, and acquires road congestion information as output data (described later in the explanation of FIGS. 4 and 7).

[0035] The traffic congestion processing unit 316 executes traffic congestion processing such as calculation of the required travel time for a predetermined section based on the traffic congestion information predicted by the traffic congestion prediction processing unit 315, and creates traffic congestion related information.

[0036] The control unit 317 executes processes other than those executed by the units 311 to 316. The control unit 317 transmits the congestion-related information created by the congestion processing unit 316 to the information output device 4, for example.

[0037] The information output device 4 includes, for example, a road information board 42, a highway telephone 43, a highway information terminal 44, and a communication device 45. The control unit 317 of the traffic control device 3 provides (transmits) congestion-related information and the like to the traffic control information providing system 36, the information output device 4, the terminal device 51, and the car navigation system 52.

[0038] The road information board 42 is installed, for example, on the shoulder of a road including the target road section or above the road, straddling the road. The road information board 42 displays congestion-related information mainly in text.

[0039] The highway telephone 43 is a voice information provider that provides voice information via a public line. The highway telephone 43 provides congestion-related information as voice information by calling a mobile phone carried by a road user or a public or dedicated phone installed in, for example, a service area (SA) or a parking area (PA).

[0040] The highway information terminal 44 is an information providing device installed in service areas, parking areas, etc., and displays, for example, congestion-related information over a wide area centered on the service area or parking area in accordance with the operation of road users, and provides related audio information.

[0041] Furthermore, the information output device 4 transmits congestion-related information and the like to the terminal device 51 via the mobile communication network N. The terminal device 51 is a mobile phone, smartphone, tablet, or the like that can be used by road users, and displays congestion-related information and the like in accordance with the road user's operation and provides related audio information by accessing a predetermined web page, for example.

[0042] Furthermore, the information output device 4 transmits congestion-related information and the like to a car navigation system 52 via a communication device 45 such as a roadside antenna. The car navigation system 52 is installed in a vehicle in which a road user rides, and acquires congestion-related information and the like about surrounding roads based on the vehicle's position identified by a GPS (Global Positioning System), and provides the information to the road user.

[0043] This concludes the description of the configuration of each device, and next we will explain the operation of the traffic control system S of this embodiment. The operation of the traffic control system S can be broadly divided into a learning process for generating a model and an operational process for making actual predictions using the model. First, the model learning process will be described. First, the details of the processing of the accident detection and prediction model generation unit 312 will be described.

[0044] 3 is an explanatory diagram of the process during learning of the accident detection and prediction model 322 according to the embodiment. As shown in FIG. 3, explanatory variables (input values; training data of input data) given during learning are, for example, the following three (1) to (3). (1) Video data from the time of the accident (CCTV camera footage, dashcam footage, etc.) (2) Traffic condition data around the accident (traffic congestion at the time of the accident, etc.) (3) Vehicle CAN (Controller Area Network) data (accelerator operation information, brake operation information, etc. of the accident vehicle)

[0045] Specifically, the accident detection prediction model generation unit 312 uses the data (1) to (3) above to create the accident detection prediction model 322 through machine learning. Machine learning is a process of adjusting the parameters of a prediction model based on given data and some criterion (e.g., the error between a predicted value and a correct value) so as to obtain a desired output. For example, this is a method using a neural network, deep learning, random forest, SVM (support vector machine), etc.

[0046] Moreover, the objective variables (correct values; teacher data of output data) are, for example, the following three (11) to (13) in the accident form data 321. (11) Accident occurrence information (information that an accident has occurred) (12) Accident situation information (detailed information on the accident, such as whether the accident vehicle can move, information on injured people, whether there are any fallen objects, and lane closure status due to the sudden incident (accident)) (13) Emergency vehicle dispatch information (detailed accident information such as whether an ambulance vehicle will be dispatched, whether an accident response vehicle will be dispatched, and the scale of the on-site inspection)

[0047] Here, as much actual data as possible that represents combinations of objective variables and explanatory variables is prepared, and the relationship between these pairs is learned by machine learning, and the parameters in the accident detection and prediction model 322 are adjusted.

[0048] In this way, the accident detection prediction model 322 is constructed by learning the explanatory variables (input values) and objective variables (correct values). Therefore, the accident detection prediction model 322 can be learned so that not only the fact that an accident has occurred but also detailed information about the accident can be acquired from video data, etc. Note that the explanatory variables and objective variables are not limited to the combinations of data described above.

[0049] Next, the details of the learning process of the congestion prediction model 323 performed by the congestion prediction model generating unit 314 will be described. In order to improve the accuracy of the congestion prediction model 323, the congestion prediction model generation unit 314 generates multiple types of congestion prediction models 323 depending on whether the location is an accident-prone location (a location where the accident occurrence rate is equal to or greater than a predetermined value), road characteristics, traffic conditions when an accident occurs, etc. Specifically, the congestion prediction model generation unit 314 generates a location-specific congestion prediction model for each accident-prone location.

[0050] Fig. 4 illustrates a portion of the processing of the traffic congestion prediction model generation unit 314 according to the embodiment. In Fig. 4, the traffic congestion prediction model generation unit 314 performs clustering of the traffic congestion prediction model 323 for non-accident-prone points (points where the accident occurrence rate is less than a predetermined value) according to vehicle detector data (traffic condition data) at the time of the accident and road characteristic information (information on characteristics such as road structure). In other words, for the non-accident-prone points, a plurality of traffic congestion prediction models for each category (a category A traffic congestion prediction model, a category B traffic congestion prediction model, and a category C traffic congestion prediction model) are generated.

[0051] In this way, the congestion prediction model generating unit 314 generates a plurality of classification congestion prediction models for non-accident-prone points according to the road characteristics and the traffic conditions when an accident occurs.

[0052] 5 is an explanatory diagram of the process during learning of each congestion prediction model 323 of the embodiment. This learning process is performed for each of the congestion prediction models 323, namely, the point-by-point congestion prediction model and the classification-by-class congestion prediction model.

[0053] As shown in FIG. 5, explanatory variables (input values; training data of input data) given during learning are, for example, the following three (31) to (33). (31) Accident status information in accident report data 321 (32) Emergency vehicle dispatch information in accident report data 321 (33) Vehicle detector data (traffic condition data at the time of the accident and when the accident is resolved)

[0054] Moreover, the objective variables (correct values; teacher data of output data) are the following three (41) to (43). (41) Time until completion of accident processing in accident report data 321 (42) Vehicle closure status information in accident report data 321 (43) Vehicle detector data (traffic condition data at the time of the accident and when the accident is resolved)

[0055] Here, as much actual data as possible that represents combinations of objective variables and explanatory variables is prepared, and the relationship between these pairs is learned by machine learning, and the parameters in the congestion prediction model 323 are adjusted.

[0056] In this way, the traffic congestion prediction model 323 is constructed by learning the explanatory variables (input values) and the objective variables (correct values). Furthermore, among the traffic congestion prediction models 323, highly accurate learning can be performed for each of the location-specific traffic congestion prediction model and the classification-specific traffic congestion prediction model. That is, for the location-specific traffic congestion prediction model, highly accurate learning can be performed because there is a large amount of accident form data 321. Furthermore, for each of the non-accident-frequent locations, although there is little accident form data 321, by learning for each classification-specific traffic congestion prediction model after clustering rather than for each location, there is a large amount of usable accident form data 321, and therefore highly accurate learning can be performed. The explanatory variables and the response variables are not limited to the above-mentioned data combinations.

[0057] Next, the operation during operation will be described. First, before describing the overall operation during operation, the operation of each model during operation will be described. 6 is an explanatory diagram of processing during operation of the accident detection and prediction model 322 according to the embodiment. As shown in FIG. 6, input values ​​given during operation (prediction) include, for example, the following two values ​​(61) and (62). (61) Video data at the time of the accident (62) Traffic condition data around the accident (vehicle detector data)

[0058] Furthermore, if it is possible to acquire the information, the following information (63) may also be added. (In this case, the traffic control device 3 needs to have a separate component for acquiring vehicle CAN data.) (63) Vehicle CAN data (accelerator operation information, brake operation information, etc. of the accident vehicle)

[0059] Then, based on the input values ​​and the accident detection and prediction model 322, the following three predicted values ​​(71) to (73) are calculated. (71) Accident occurrence information (information that an accident has occurred) (72) Accident situation information (detailed information on the accident, such as whether the accident vehicle can move, information on injured people, whether there are any fallen objects, and lane closure status due to the sudden incident (accident)) (73) Emergency vehicle dispatch information (detailed accident information such as whether an ambulance will be dispatched, whether an accident response vehicle will be dispatched, and the scale of the on-site inspection)

[0060] In this way, using the accident detection and prediction model 322, not only information that an accident has occurred but also detailed accident information can be obtained from video data and the like.

[0061] 7 is an explanatory diagram of processing during operation of the traffic congestion prediction model 323 according to the embodiment. This processing during operation (prediction) is performed for each of the point traffic congestion prediction model and the classification traffic congestion prediction model among the traffic congestion prediction models 323.

[0062] If the location in the accident occurrence information (information that an accident has occurred) acquired by the accident detection and prediction processing unit 313 is an accident-prone location (a location where the accident occurrence rate is equal to or greater than a predetermined value), the congestion prediction processing unit 315 uses the location-specific congestion prediction model corresponding to that location. Also, if the location in the accident occurrence information acquired by the accident detection and prediction processing unit 313 is a non-accident-prone location (a location where the accident occurrence rate is less than a predetermined value), the congestion prediction processing unit 315 uses the location-specific congestion prediction model corresponding to the accident-prone location if there is an accident-prone location with similar road characteristics and traffic conditions at the time of the accident, and if there is no accident-prone location, it uses the corresponding classification-specific congestion prediction model based on the road characteristics and traffic conditions at the time of the accident.

[0063] As shown in FIG. 7, the input values ​​given during operation are, for example, the following three values ​​(81) to (83). (81) Accident situation information calculated by the accident detection and prediction processing unit 313 (information of (72)) (82) Emergency vehicle dispatch information calculated by the accident detection and prediction processing unit 313 (information of (73)) (83) Vehicle detector data (traffic condition data at the time of the accident)

[0064] Then, based on the input values ​​and the congestion prediction model 323, the following three predicted values ​​(91) to (93) are calculated. (91) Time until accident processing is completed (92) Accident Closure Information (93) Traffic condition data at the time of the accident and at the time of completion of the accident

[0065] In this way, the above-mentioned predicted values ​​can be calculated with high accuracy from the above-mentioned input values ​​using either the location-specific congestion prediction model or the category-specific congestion prediction model corresponding to the location where the accident occurred, among the congestion prediction models 323.

[0066] 8 is a flowchart showing the overall processing during operation of the traffic control device 3 of the embodiment. At this point, the accident detection and prediction model 322 and the congestion prediction model 323 have been trained and constructed.

[0067] First, in step S1, the accident detection prediction processing unit 313 determines whether or not an occurrence of an unexpected event (occurrence of an accident) has been detected based on road video data, etc., using the accident detection prediction model 322. If the accident detection prediction processing unit 313 has detected the occurrence of an accident (Yes), the process proceeds to step S2, and if the accident detection prediction processing unit 313 has not detected the occurrence of an accident (No), the process returns to step S1.

[0068] In step S2, the accident detection and prediction processing unit 313 acquires accident detail information using the accident detection and prediction model 322 based on road video data and the like. Here, the accident detail information is, as described above, accident situation information (accident detail information such as whether the accident vehicle can move, information on injured persons, whether there are any fallen objects, and whether lane closure status due to the sudden event (accident)) and emergency vehicle dispatch information (accident detail information such as whether an ambulance will be dispatched, whether an accident response vehicle will be dispatched, and information on the scale of the on-site inspection). If the accident detection and prediction processing unit 313 has acquired the accident detail information (Yes), the process proceeds to step S5, and if the accident detail information has not been acquired (No), the process proceeds to step S3. For example, if the video data does not show much video that indicates the characteristics of the accident, the result in step S2 may be No.

[0069] In step S3, when a traffic controller recognizes that an accident has occurred from the video data, the traffic controller manually inputs an event (detailed accident information) registration using the input unit 33. At this time, the control unit 317 accepts the event registration input from the input unit 33.

[0070] Next, in step S4, the control unit 317 acquires detailed accident information (accident situation information, emergency vehicle dispatch information, etc.) based on the input information from the traffic controller, and proceeds to step S5.

[0071] In step S5, the traffic congestion prediction processing unit 315 predicts traffic congestion that will occur due to the detected unexpected event. As described above, the traffic congestion prediction model 323 is divided into multiple models, so it is necessary to determine which traffic congestion prediction model 323 to apply. For this reason, the traffic congestion prediction processing unit 315 determines whether the identified accident occurrence point is an accident-prone point or not. If the answer is Yes in step S5, a traffic congestion prediction model for the occurrence point has already been constructed, so the process proceeds to step S11. On the other hand, if the answer is No in step S5, the process proceeds to step S6.

[0072] In step S6, the congestion prediction processing unit 315 determines whether there is a high-accident point that has similar road characteristics and traffic conditions at the time of the accident to the detected accident point. If the answer is Yes in step S6, the process proceeds to step S11 to make a prediction using the congestion prediction model 323 for the similar accident point. On the other hand, if the answer is No in step S6, the process proceeds to step S7.

[0073] Next, in step S7, the congestion prediction processing unit 315 selects the most appropriate classification congestion prediction model from the learned congestion prediction models based on the road characteristics and traffic conditions at the time of the accident. In other words, even if the accident occurrence point is not an accident-prone point, if there is an accident-prone point with similar road characteristics and traffic conditions at the time of the accident, the point-by-point congestion prediction model of the accident-prone point will be used.

[0074] Next, in step S8, the congestion prediction processing unit 315 predicts congestion information such as the time until the accident processing is completed and the traffic situation after the accident occurs, using the classification congestion prediction model selected in step S7.

[0075] Next, in step S9, the traffic congestion processing unit 316 performs traffic congestion processing such as calculating the required travel time for a predetermined section based on the traffic congestion information, and creates traffic congestion-related information.

[0076] Next, in step S10, the control unit 317 transmits the congestion-related information to the traffic control information providing system 36, the information output device 4, the terminal device 51, and the car navigation system 52. This enables the traffic control information providing system 36, the information output device 4, the terminal device 51, and the car navigation system 52 to display the congestion-related information.

[0077] In step S11, the traffic congestion prediction processing unit 315 selects the corresponding traffic congestion prediction model for each point.

[0078] Next, in step S12, the congestion prediction processing unit 315 uses the point-by-point congestion prediction model selected in step S9 to predict congestion information such as the time until the accident processing is completed and the traffic situation after the accident occurs.

[0079] Next, in step S13, the traffic congestion processing unit 316 performs traffic congestion processing such as calculating the required travel time for a predetermined section based on the traffic congestion information, and creates traffic congestion-related information.

[0080] Next, in step S14, the control unit 317 transmits the congestion-related information to the traffic control information providing system 36, the information output device 4, the terminal device 51, and the car navigation system 52. This enables the traffic control information providing system 36, the information output device 4, the terminal device 51, and the car navigation system 52 to display the congestion-related information.

[0081] In this way, the traffic control device 3 of this embodiment can quickly detect the occurrence of an accident on a road based on the accident detection and prediction model 322 that has been learned and generated in advance.

[0082] In addition, by using the accident detection and prediction model 322, highly accurate learning can be performed by the process shown in FIG. 3, and detailed information about the accident (accident situation information, emergency vehicle dispatch information, etc.) can be predicted highly accurately by the process shown in FIG.

[0083] In addition, the accident processing time and information about the congestion that will occur from the time the accident occurs until the time the accident processing is completed can be predicted with high accuracy using the accident detailed information predicted by the accident detection and prediction model 322 and the congestion prediction model 323. In other words, the impact of an accident on congestion can be reflected quickly and with high accuracy.

[0084] In this case, by using a location-specific congestion prediction model and a classification-specific congestion prediction model as the congestion prediction model 323, congestion information can be predicted with high accuracy regardless of whether the accident occurrence point is a high-accident point or not.

[0085] Furthermore, by calculating congestion-related information based on the congestion information and displaying it on the information output device 4, the terminal device 51, the car navigation system 52, etc., it is possible to, for example, disperse traffic flow and minimize congestion.

[0086] Furthermore, by displaying congestion-related information on the traffic control information providing system 36, traffic controllers can take into account the impact of accidents in their traffic control operations.

[0087] The program executed by the CPU of the traffic control device 3 of this embodiment may be configured to be provided by being recorded in an installable or executable file format on a computer-readable recording medium such as a CD-ROM, a flexible disk (FD), a CD-R, or a DVD (Digital Versatile Disk).

[0088] Furthermore, the program may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Also, the program executed in this embodiment may be provided or distributed via a network such as the Internet.

[0089] Although several embodiments of the present invention have been described above, the above-described embodiments and modifications are merely examples and are not intended to limit the scope of the invention. The above-described embodiments can be implemented in various forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. The above-described embodiments and modifications are included within the scope and spirit of the invention, and are also included in the inventions and their equivalents as set forth in the claims.

[0090] For example, the video data is not limited to video data from a fixed camera, and may be video data from a drive recorder mounted on a vehicle, video data from a drone, etc. When video data from a drive recorder is used, for example, video data from a drive recorder in a contracted taxi or bus may be used, but is not limited to this.

[0091] Furthermore, when predicting congestion, events that may affect congestion on the target road section, weather conditions, etc. may be added to the conditions.

[0092] Furthermore, the target roads are not limited to expressways, but may also be toll roads other than expressways or free general roads. [Explanation of symbols]

[0093] 1...vehicle detector, 2...CCTV camera, 3...traffic control device, 4...information output device, 31...processing unit, 32...storage unit, 33...input unit, 34...display unit, 35...communication unit, 36...traffic control information provision system, 42...road information board, 43...highway telephone, 44...highway information terminal, 51...terminal device, 52...car navigation system, 311...acquisition unit, 312...accident detection prediction model generation unit, 313...accident detection prediction processing unit, 314...congestion prediction model generation unit, 315...congestion prediction processing unit, 316...congestion processing unit, 317...control unit, 321...accident report data, 322...accident detection prediction model, 323...congestion prediction model, C...vehicle, R...road, S...traffic control system

Claims

1. a traffic congestion prediction model generation unit that generates a traffic congestion prediction model by performing machine learning based on training data in which input data includes traffic condition data when an unexpected event occurred in the past for a road on which a vehicle travels, and element information that is information about elements that cause the unexpected event to affect the traffic condition of the road, and output data includes congestion information that is information about congestion on the road due to the unexpected event; a congestion prediction processing unit that receives traffic condition data at the time when an unexpected event occurs and the element information corresponding to the unexpected event as input data, and acquires congestion information on the road as output data based on the congestion prediction model; Equipped with The traffic congestion prediction model generation unit A point-by-point congestion prediction model is generated for each point where the unexpected incident occurrence rate is equal to or greater than a predetermined value, generating a plurality of classification-specific congestion prediction models corresponding to the points where the unexpected incident occurrence rate is less than a predetermined value, the classification being based on road characteristics and traffic conditions when the unexpected incident occurs; Traffic control equipment.

2. The congestion prediction processing unit: The traffic control device according to claim 1 , wherein the traffic congestion prediction model used for prediction is changed depending on the probability of an unexpected incident occurring at a location indicated in the information indicating that an unexpected incident has occurred, to predict congestion.

3. The congestion prediction processing unit: When the occurrence rate of the unexpected incident at the point in the information indicating the occurrence of the unexpected incident is equal to or greater than a predetermined value, the point-by-point congestion prediction model corresponding to the point is used.

3. A traffic control device according to claim 2.

4. The congestion prediction processing unit: If the occurrence rate of the unexpected incident at the point in the information indicating the occurrence of the unexpected incident is less than a predetermined value, the congestion prediction model used for prediction is changed depending on the road characteristics and the presence or absence of points with similar traffic conditions at the time of the unexpected incident, and congestion is predicted.

3. A traffic control device according to claim 2.

5. The congestion prediction processing unit: If there is a point where the road characteristics and the traffic conditions at the time of the unexpected incident are similar and the unexpected incident occurrence rate is equal to or greater than a predetermined value, the point-by-point congestion prediction model corresponding to the point where the unexpected incident occurrence rate is equal to or greater than the predetermined value is used.

5. A traffic control device according to claim 4.

6. The congestion prediction processing unit: If there is no point where the road characteristics and the traffic conditions at the time of the unexpected incident are similar and the unexpected incident occurrence rate is equal to or greater than a predetermined value, the congestion prediction model for each category corresponding to the road characteristics and the traffic conditions at the time of the unexpected incident is used.

5. A traffic control device according to claim 4.

7. A congestion prediction model generation step for generating a congestion prediction model by performing machine learning based on training data in which input data is traffic condition data when an unexpected event occurred in the past for a road on which a vehicle is traveling, and element information which is information on elements that the unexpected event affects the traffic condition of the road, and output data is congestion information which is information on congestion on the road due to the unexpected event; a congestion prediction processing step of acquiring, as output data, congestion information on the road based on the congestion prediction model, using traffic condition data at the time when the unexpected event occurred and element information corresponding to the unexpected event as input data; Including, The traffic congestion prediction model generation step includes: generating a congestion prediction model for each location where the unexpected incident occurrence rate is equal to or greater than a predetermined value; generating a plurality of classification-specific congestion prediction models corresponding to points where the unexpected incident occurrence rate is less than a predetermined value, the classification being based on road characteristics and traffic conditions when the unexpected incident occurs; Including, Traffic control methods.

Images