JP2019079199A - Signal switching control device, signal switching control method and signal switching control program - Google Patents

Abstract

To smooth a flow of vehicles in relative to a case where the flow of vehicles is identified from traffic of a main road, when controlling switching of signals within a road network.SOLUTION: In accordance with a traffic light switching control device (12), in a signal control optimization unit (126), traffic information of a road network stored in a traffic information storage unit (123) is acquired via a related information extraction unit (124), switching timing of each of signals (SG1-SGN) installed in intersections is optimized while taking a main road connected to an intersection and a peripheral road connected to the other adjacent intersection into account by a computing element that is constructed by data from an optimized data storage unit (125), and information indicating the optimized switching timing is outputted to the signal. Thus, a vehicle traveling on a road included in the road network can be smoothly traveled, and a flow of vehicles in the road network is smoothed.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a traffic signal switching control device, a traffic signal switching control method, and a traffic signal switching control program.

  Conventionally, there is known a technique of adjusting parameters of traffic signals so as to maximize the number of vehicles passing through a main intersection for each intersection or each intersection row using information of traffic volume measured by a traffic survey or a vehicle sensor. ing. Further, for example, a technology has been proposed that takes into consideration the traffic volume around, such as the traffic volume entering and leaving the side road (see, for example, Patent Document 1). In this technology, road map storage means for storing road maps and road structures, passing vehicle number collecting means for collecting the number of passing vehicles passing through the road, and traffic between major points based on the number of passing vehicles. Traffic occurrence / disappearance amount calculating means for finding occurrence / disappearance amount, Traffic volume calculation means for finding traffic volume of main road from the traffic occurrence / disappearance amount, The traffic occurrence / disappearance from the traffic volume and the structure of the road It is characterized by having an area determination means for determining a non-congested traffic flow controllable area which can be passed through without causing congestion.

Japanese Patent Application Laid-Open No. 5-28395

By the way, the flow of the vehicle traveling on the road relates to the entry and exit of the vehicle to at least one of the other road connected to the intersection and the other road connected to the traveling road. However, when identifying the flow of a vehicle from the change in traffic volume of the main road and switching control of the traffic signal, it is difficult to determine which road affects the traffic volume, and the traffic condition of the road related to the flow of the vehicle Is not enough to reflect the
A traffic signal switching control device and a traffic signal switching device according to the present invention can smooth the flow of a vehicle compared to the case of specifying the flow of the vehicle from the traffic volume of the main road when switching control of the traffic signal in the road network. An object of the present invention is to provide a switching control method and a traffic signal switching control program.

In order to achieve the above object, the traffic signal switching control device according to claim 1 is:
For each of the plurality of intersections of the road network provided with a plurality of intersections in which each traffic light is installed, the plurality of intersections are installed at each of the plurality of intersections based on traffic information of a vehicle traveling on a main road connected to each intersection. An operation unit that calculates switching timing of each of the traffic signals;
For each of the traffic lights installed at each of the plurality of intersections, each of the first traffic flows when the traffic lights are switched at the switching timing calculated by the calculation unit based on traffic information of the vehicle traveling on the main road A first estimation unit to estimate;
The traffic information of the vehicle traveling on the main road of the specific intersection and the traffic information of the vehicle traveling on the surrounding road linked to the main road ahead of the specific intersection, for any one specific intersection of the plurality of intersections Second estimation based on the second traffic flow when switching the traffic light installed at the specific intersection and the traffic light installed at the previous intersection at the switching timing calculated by the calculation unit Department,
Optimal to optimize the switching timing of the traffic signal installed at the specific intersection based on the first traffic flow estimated by the first estimation unit and the second traffic flow estimated by the second estimation unit Department,
Equipped with

The invention according to claim 2 is the signal switching control device according to claim 1;
The arithmetic unit is
The traffic information is used as an input, the switching timing of the traffic light is used as an output, and the intersection model of the main road learned so as to be in the predetermined traffic flow state is included.

The invention according to claim 3 is the signal switching control device according to claim 1 or 2.
The second estimation unit is
The traffic information and the switching timing of the traffic signal are input, and the traffic flow state when the traffic signal is switched at the input switching timing is output, including the intersection model including the learned surrounding road.

The invention according to claim 4 relates to the signal switching control device according to any one of claims 1 to 3.
The arithmetic unit is
It includes a storage unit that stores the switching timing of each of the traffic signals corresponding to the calculated traffic information.

The invention according to claim 5 relates to the traffic signal switching control device according to any one of claims 1 to 4.
The optimization unit
A surrounding road connected to the previous intersection is added until the difference between the first traffic flow estimated by the first estimation unit and the second traffic flow estimated by the second estimation unit is less than a predetermined threshold. And optimize the switching timing of the traffic lights installed at the specific intersection.

The traffic signal switching control method of the invention according to claim 6 is
On the computer
The processing of each part of the traffic signal switching control device according to any one of claims 1 to 5 is performed.

The traffic signal switching control program of the invention according to claim 7 is
A computer is made to function as each part of the signal apparatus switching control apparatus described in any one of Claims 1-5.

According to the inventions of claims 1, 6, and 7, when the switching control of the traffic signal in the road network is performed, the flow of the vehicle is smoothed as compared with the case where the flow of the vehicle is specified from the traffic volume of the main road. Have the effect of being able to
According to the second and third aspects of the present invention, the structure of the traffic signal switching control device can be simplified as compared with the case where the switching timings of the individual traffic signals are individually simulated.
According to the fourth aspect of the present invention, the calculation load of the traffic signal switching control device can be suppressed as compared with the case where the switching timing is sequentially calculated.
According to the fifth aspect of the present invention, it is possible to consider only the roads that affect the switching timing of individual traffic signals, and to promote optimization of the switching timing, as compared to the case where the roads of the entire road network are considered simultaneously. Have.

It is a block diagram showing an example of composition of a signal change control device concerning this embodiment. It is a schematic diagram which shows an example of the arrangement | positioning of the road in the road network which concerns on this embodiment, an intersection, and a signal apparatus. It is a schematic diagram which shows an example of arrangement | positioning of the main road linked to the intersection which concerns on this embodiment. It is a schematic diagram which shows an example of arrangement | positioning of the surrounding road connected with the other intersection adjacent to the intersection which concerns on embodiment. It is a block diagram showing an example which realized a signal change control device concerning this embodiment by composition including a computer. It is a flowchart which shows an example of the flow of the process by the traffic volume predictor learning process performed by the computer operate | moving as a signal apparatus switching control apparatus which concerns on this embodiment. It is a flowchart which shows an example of the flow of the process by the optimization process performed by the computer operate | moving as a signal apparatus switching control apparatus which concerns on this embodiment. FIG. 2 is a schematic view showing an example of a road network including roads where a plurality of vehicles are expected to be concentrated according to the present embodiment.

  Hereinafter, an example of the signal switching control system according to the present embodiment will be described in detail with reference to the drawings. In addition, the same code | symbol may be provided to the component and process in which an effect | action and a function bear the same function through all the drawings, and the overlapping description may be abbreviate | omitted suitably.

FIG. 1 shows an example of the configuration of a traffic signal switching control system 10 according to the present embodiment.
As shown in FIG. 1, the traffic signal switching control system 10 includes a traffic signal switching control device 12 and signal control units 14-SG1 to 14-SGN mounted on each of a plurality of traffic signals SG1 to SGN (N> 1). It is. The traffic signal switching control system 10 controls switching timing of traffic signals included in the road network AR such that a vehicle traveling on a road included in the predetermined road network AR travels smoothly. For example, the traffic signal switching control device 12 is provided in an integrated center that collectively manages traffic signals installed in the road network AR. In the present embodiment, control of the switching timing of the traffic light will be described by way of an example where the traffic light switching control device 12 is installed at the integrated center. The traffic signal switching control device 12 is not limited to the installation in the integrated center. For example, any base, place, or device capable of outputting a control signal to the signal control units 14-SG1 to 14-SGN may be used.

  Each of the plurality of traffic lights SG1 to SGN controls each of the plurality of signal lights 16-SG1 to 16-SGN and each of the plurality of signal lights 16-SG1 to 16-SGN, for example, signal control for controlling lighting, blinking and extinguishing Each of the units 14-SG1 to 14-SGN is provided. Each of the signal control units 14-SG1 to 14-SGN controls each of the plurality of signal lights 16-SG1 to 16-SGN according to a control signal from the signal switching control device 12 installed at the integrated center. One example of the control of each of the plurality of signal lights 16-SG1 to 16-SGN includes switching control of lighting, blinking and extinguishing of the signal lights indicating an instruction related to traveling of the vehicle. In addition, when each signal light includes a plurality of signal lights of different signal colors indicating different instructions for traveling of the vehicle, switching control of the signal color may be mentioned.

  In addition, since each of the plurality of traffic signals SG1 to SGN has substantially the same configuration, when it is not necessary to distinguish each of them, the traffic signals SG will be collectively described. On the other hand, when each of the plurality of traffic signals SG is described separately, any one of the symbols of the traffic signals SG1 to SGN is attached. In addition, since each of the signal control units 14-SG1 to 14-SGN mounted on each of the plurality of traffic signals SG has substantially the same configuration, when it is not necessary to distinguish between them, the signal control units 14 are collectively referred to. Explain. On the other hand, when each of the plurality of signal control units 14 is described separately, any one of the signal control units 14-SG1 to 14-SGN is attached. Furthermore, since each of the plurality of signal lights 16-SG1 to 16-SGN has substantially the same configuration, when it is not necessary to distinguish each of them, the plurality of signal lights 16-SG1 to 16-SGN will be collectively described as the signal lights 16. On the other hand, when each of the plurality of signal lights 16 is described separately, any one of the signs of the signal lights 16-SG1 to 16-SGN is attached.

  The traffic signal switching control device 12 is a functional unit that controls each of the switching timings of the signal lights 16 in each of the plurality of traffic signals SG included in the road network AR such that a vehicle traveling on a road included in the road network AR travels smoothly. is there. Specifically, the traffic signal switching control device 12 controls the vehicle-side information collection unit 121, the road-side information collection unit 122, the traffic information storage unit 123, the related information extraction unit 124, the optimization data storage unit 125, and signal control. And an optimization unit 126. The traffic signal switching control device 12 determines the switching timing of each of the plurality of traffic signals SG using traffic information related to traffic in the road network AR, and a plurality of signal control units mounted on each of the plurality of traffic signals SG Send to each of the fourteen. Next, each functional unit included in the traffic signal switching control device 12 will be described in detail.

  The traffic information storage unit 123 is a functional unit that stores traffic information that changes every moment. One example of the traffic information is information indicating traffic volume on the road. The state in which the vehicle is traveling with the traffic volume indicated by the traffic information on each of the roads in the road network AR is the traffic condition of the road network AR. The traffic information storage unit 123 acquires traffic information and stores the acquired traffic information in a database. For example, as traffic information that changes from moment to moment, a road ID for identifying a road by a road section in a predetermined distance (for example, several hundred meters), a time indicating a detection time in the road section, and traveling the road section A data set in which the vehicle speed indicating the average vehicle speed of the vehicle and the number of vehicles traveling the road section are associated is stored as a database record. Thus, a traffic database is constructed in the traffic information storage unit 123.

  In the present embodiment, in order to obtain the traffic information stored in the traffic information storage unit 123, the traffic information storage unit 123 is connected with the car-side information collection unit 121 and the roadside information collection unit 122. That is, the traffic information storage unit 123 stores information including the car side information collected by the car side information collection unit 121 and the roadside information collected by the roadside information collection unit 122 as traffic information.

  The vehicle-side information collection unit 121 is a functional unit that collects vehicle-side information transmitted from a vehicle traveling in the road network AR. An example of the vehicle side information includes a road on which a vehicle travels in a predetermined unit time, and road / speed information indicating an average vehicle speed.

  The roadside information collection unit 122 is a functional unit that collects roadside information detected by a sensor installed in the road infrastructure. Examples of roadside information include information on traffic volume including the number of vehicles passing a predetermined position on the road where the sensor is installed, the vehicle speed at the time of passing, and the detection time. In the present embodiment, the roadside information includes map information of the road network AR.

  The roadside information collection unit 122 may acquire roadside information by receiving information transmitted from the roadside, and may acquire roadside information by receiving information input by an input device.

  The related information extraction unit 124 is connected to the traffic information storage unit 123, the optimization data storage unit 125, and the signal control optimization unit 126. The related information extraction unit 124 uses the traffic information extracted from the traffic information storage unit 123 for the related information used for deriving the switching timing of each of the plurality of signal lights 16 optimized by the signal control optimization unit 126 described in detail later. Is a functional unit that derives In the present embodiment, as the related information used to derive the switching timing of the signal light 16, a case is described in which information indicating the number of vehicles traveling on the road and the traffic volume such as the vehicle speed is used.

  The related information extraction unit 124 includes a traffic flow predictor in order to derive information indicating traffic as the related information using the traffic information stored in the traffic information storage unit 123. In the present embodiment, a traffic volume predictor for deriving information indicating traffic volume is constructed in advance, and the parameters of the constructed traffic volume predictor are stored in the optimization data storage unit 125. When deriving the switching timing of the traffic light 16, the parameter indicating the constructed traffic predictor is acquired from the optimization data storage unit 125, and the traffic predictor is reconstructed to construct the traffic predictor. The information indicating traffic volume is output to the signal control optimization unit 126 using That is, the related information extraction unit 124 constructs in advance a traffic flow predictor according to the purpose, extracts information according to the set purpose, and outputs the information to the signal control optimization unit 126.

  In detail, the related information extraction unit 124 is a machine learning model such as a neural network system (hereinafter referred to as a traffic volume model) that learns so that output information in accordance with input information is output when certain input information is input. Use.). In this embodiment, the timing to specify the switching timing of the signal light at the traffic light around the road of the road ID and the average vehicle speed on the road of a certain road ID at a certain time t, the road around the road ID road Parameters and input information. On the other hand, the output information is information indicating the number of vehicles after a predetermined time (for example, several minutes) on the road of a certain road ID and the traffic volume by the vehicle speed and the like.

  Therefore, in the related information extraction unit 124, when the average vehicle speed on the road of the road ID, the average vehicle speed on the surrounding road, and the switching timing of the surrounding traffic light are input using the traffic information of the traffic information storage unit 123, A traffic volume predictor is built that predicts traffic volume several minutes after the road with the road ID. Then, parameters defining the constructed traffic predictor, ie, parameters in the traffic model, are stored in the optimization data storage unit 125.

  The related information extraction unit 124 also has a function of extracting and outputting, for example, traffic information at a designated time from the traffic information stored in the traffic information storage unit 123.

  The optimization data storage unit 125 is a functional unit that stores information used to control the switching timing of the signal light 16 of the traffic light SG. As described above, parameters of the traffic volume model defining the traffic volume predictor constructed by the related information extraction unit 124 are stored. Further, optimization data used when optimizing the switching timing of the signal light 16 of the traffic light SG is also stored in the signal control optimization unit 126 (details will be described later).

  The signal control optimization unit 126 is a functional unit that optimizes the switching timing of the signal lights 16 of the individual traffic lights SGi (1 ≦ i ≦ N) in the road network AR. The signal control optimization unit 126 controls the switching timing of the signal light 16 of each traffic light SG in the road network AR, thereby optimizing the traffic flow in the road network AR.

  FIG. 2 is a schematic view showing an example of the arrangement of roads by road ID in the road network AR, intersections CR, and traffic lights SG. Further, FIG. 2 also shows a plurality of vehicles C traveling in the road network AR. In addition, in FIG. 2, the case where the vehicle C drive | works in the direction of arrow FR is shown. Further, as the traffic light SG, only the traffic light related to the travel instruction when the vehicle C travels in the direction of the arrow FR is described. FIG. 2 is an example of the road network AR in which nine intersections CR1 to CR9 exist in the road network AR. Further, for example, roads of road ID 01 to road ID 04 are connected to the intersection CR 1.

  The signal control optimization unit 126 includes a computing unit that uses the information from the related information extraction unit 124 to optimize the switching timing of the traffic light SG in the road network AR. In the present embodiment, an arithmetic unit for optimizing the switching timing of the traffic light SG in the road network AR is constructed in advance by learning, and the parameters of the constructed arithmetic unit are stored in the optimization data storage unit 125. . When deriving the switching timing of the signal light 16, the parameter indicating the already constructed computing element is acquired from the optimization data storage unit 125, the computing element is reconstructed, and the traffic signal SG is generated using the constructed computing element. The information indicating the switching timing of is output to each of the traffic signals SG. That is, the signal control optimization unit 126 constructs in advance an arithmetic unit that optimizes the switching timing of each of the traffic signals SG in the road network AR, and switches the switching timing of each of the traffic signals SG optimized by the arithmetic unit. Information indicating each of the traffic signals to each of the traffic signals SG.

  In detail, the arithmetic unit included in the signal control optimization unit 126 is a machine learning model such as a neural network system which learns that output information in accordance with input information is output when certain input information is input. It can be used. In this embodiment, a traffic condition at a certain time t is used as input information, and a model using output parameters as a timing parameter that defines switching timing of a traffic light in a traffic light is used. That is, a model (hereinafter referred to as an intersection model) that outputs the switching timing of the traffic light SG installed at the intersection CR according to the traffic condition of the road related to the intersection CR is used.

In the present embodiment, the computing unit included in the signal control optimization unit 126 is constructed by two-stage learning processing.
The learning process of the first stage uses the traffic information of the road (hereinafter referred to as a main road) connected to each intersection CR to the switching control timing of the signal light at the traffic light SG installed at each intersection CR in the road network AR. To learn. In the learning process of the second stage, the switching control timing of the signal light in the traffic light SG installed at each intersection CR in the road network AR is connected to another intersection adjacent to each intersection CR (hereinafter, referred to as a surrounding road. To learn including traffic information). That is, the surrounding road is a road ahead of the main road connected to a specific intersection CR. The other intersections adjacent to the intersection CR include other intersections existing around the intersection CR.

  FIG. 3 is a schematic view showing an example of the arrangement of the main roads connected to the intersection, and shows the roads of road ID 01 to ID 04 as oblique lines as the main roads connected to the intersection CR 1. Further, FIG. 4 is a schematic view showing an example of the arrangement of the peripheral road connected to another intersection adjacent to the intersection, and as an example of the peripheral road, a road having a road ID06 connected to another intersection CR2 adjacent to the intersection CR1. Are indicated by diagonal lines.

  When the switching timing of the traffic signal SG05 installed at the intersection CR1 is optimized (for example, FIG. 2), in the learning process of the first step, learning is performed using traffic information of roads of road ID 01 to road ID 04 (for example, FIG. 3). The learning process of the second stage learns including traffic information of a road (for example, a road with road ID 06) connected to at least one of the other intersections CR2 to CR9 adjacent to the intersection CR1 (for example, the intersection CR2). (For example, FIG. 4).

Specifically, in the first stage of learning processing, the following processing is performed on each of the traffic signals SG at each intersection.
In the learning process of the first stage, a model is used as the intersection model, with the traffic situation at the intersection CR at a certain time t as input information and the timing parameter defining the switching timing of the signal light 16 of the traffic light SG as output information. That is, a model (hereinafter, referred to as a first intersection model) that outputs the switching timing of the traffic signal SG installed at the intersection CR according to the traffic situation of the main road of the intersection CR is used.

First, save the road ID indicating all major roads are connected to the intersection CRi (1 ≦ i ≦ 9) , the road set _R i utilized in situations _{s i (s i).} The situation s _i is the traffic condition of the main road in which the vehicle is traveling with the traffic volume indicated by the traffic information on each of the main roads.

Next, the traffic information of the road set R _i (s _i ) is used to separately learn the switching timing of the traffic light SG using Q learning, which is a type of reinforcement learning, to obtain the policy π _i and the state value function V _i save.

  Specifically, in the present embodiment, the computing units included in the signal control optimization unit 126 are modeled using Markov decision processes (MDPs) suitable for an environment in which a state transitions stochastically. That is, this model assumes that the state transition at a certain time depends on the history of states and actions up to a predetermined time before, observes the environmental state at each time, and determines the action based on the observed states. , Reward corresponding to the state transition when transitioning to a new state. Optimize this model. That is, when the state s, which is a traffic condition, is given, the policy π that defines the action a, which is the switching timing of the traffic light SG, is optimized. An example of the objective function in this case is shown below.

E is an expected value, and t is a unit time. γ is a discount rate of reward, and the closer to 1, the longer-term policy can be obtained, and the closer to 0, the shorter-term policy can be acquired.

Q learning is to learn a state action value function Q that minimizes an error TD shown below.

  In addition, since Q learning is a known technique, detailed explanation is omitted (Watkins, Christopher JCH, and Peter Dayan. "Q-learning." Machine learning 8.3-4 (1992): 279-292. ).

  Next, in the learning process of the second stage, after the learning process of the first stage is completed for each of the traffic lights SG of each intersection, the following process is performed for each of the traffic lights SG of each intersection.

  In the learning process of the second stage, a model is used in which traffic conditions at an intersection CR at a certain time t and adjacent intersections are used as input information and timing parameters defining the switching timing of the signal light 16 of the traffic light SG as output information. That is, a model (hereinafter referred to as a second intersection model) which outputs the switching timing of the traffic light SG installed at the intersection CR according to the traffic condition of the traffic condition of the road including the main road of the intersection CR and the surrounding road of the traffic light SG of the adjacent intersection. Use.).

First, the road set R _i (s _i ) used in the situation s _i at time t of the intersection CR i (1 ≦ i ≦ 9) is connected to the adjacent intersections CR j (1 ≦ j ≦ 9, i ≠ j) A road set R ' _i (s _i ) is constructed by adding the road IDs of the surrounding roads.

Next, the state value function V ' _i (s _i ) at the next time when the traffic lights SG of all intersections CR included in the road set R' _i (s _i ) are switched by the policy learned in advance is observed .

Here, the road set _{R i (s} i) of the traffic information 'and _(i, state value function V next time the state value function _V i s)' that was experienced in the case of learning by using the _i (s' _i) If the bets are significantly different, i.e., if the difference condition value exceeds a predetermined threshold value, the intersection CRi and intersections CRj, road aggregate utilized in situations s _i as related to traffic flow control R _i ( Update s _i ) to the road set R ' _i (s _i ). On the other hand, when not changing below the threshold, the road set R _i (s _i ) is maintained.

Next, the policy π _i and the state value function V _i are updated based on Q learning using traffic information of the road set R _i (s _i ).

  There are no more roads to add the above processing, and the process is repeated until the values of the policy and state value functions converge. The parameter of the computing unit for optimizing the switching timing of the traffic signal SG in the road network AR, which is constructed in the signal control optimization unit 126 whose learning is finished in this way, is stored in the optimization data storage unit 125 Ru.

  As described above, when the related information extraction unit 124 and the signal control optimization unit 126 finish learning, the traffic light switch control device 12 controls the vehicle C traveling on the road included in the road network AR based on the traffic information. Control each of the switching timing of the signal light 16 in each of the plurality of traffic signals SG included in the road network AR so that the vehicle travels smoothly.

  That is, when deriving the switching timing of the signal light 16, the related information extraction unit 124 acquires the parameter indicating the traffic volume predictor that has already been constructed from the optimization data storage unit 125, and reconstructs the traffic volume predictor. And outputs the information indicating traffic volume to the signal control optimization unit 126 using the constructed traffic volume predictor. In addition, the signal control optimization unit 126 acquires a parameter indicating a constructed computing unit from the optimization data storage unit 125, reconstructs the computing unit, and uses the constructed computing unit to switch the traffic light SG. Information indicating timing is output to each of the traffic lights SG. Each of the signal control units 14 of the traffic lighter SG performs switching timing of each of the signal lights 16 by a control signal from the signal control optimization unit 126. This optimizes the traffic flow in the road network AR.

  The traffic signal switching control device 12 is an example of the traffic signal switching control device of the disclosed technology. The arithmetic unit included in the signal control optimization unit 126 and constructed by the learning process of the first stage is an example of the arithmetic unit of the disclosed technology. The computing unit constructed by the learning process of the second stage is an example of the first estimation unit and the optimization unit of the disclosed technology. The traffic volume predictor included in the related information extraction unit 124 and using traffic information to derive information indicating traffic volume is an example of a first estimation unit of the disclosed technology. Further, the optimization data storage unit 125 is an example of a storage unit of the disclosed technology. Furthermore, the first intersection model is an example of the intersection model of the main road of the disclosed technology, and the second intersection model is an example of the intersection model including the surrounding roads of the disclosed technology.

  The traffic signal switching control device 12 shown in FIG. 1 can be realized by a configuration including a computer that executes processing for controlling each of the switching timings of the signal lights 16 in each of the plurality of traffic signals SG included in the road network AR.

FIG. 5 shows an example in which the signal switching control device 12 is realized by a configuration including a computer 12X.
As shown in FIG. 5, the computer 12X includes a central processing unit (CPU) 12A, a random access memory (RAM) 12B, a read only memory (ROM) 12C, and an input / output interface (I / O) 12D via a bus 12E. The vehicle-side communication unit 12F, the roadside communication unit 12G, the traffic signal communication unit 12H, and the storage unit 12M such as a hard disk drive are connected to the I / O 12D.

  The ROM 12 </ b> C stores an information processing program 12 </ b> P related to a switching control process of the traffic light SG to be executed by the computer 12 </ b> X. The CPU 12A reads the information processing program 12P from the ROM 12C and develops the information processing program 12P on the RAM 12B, and executes processing by the information processing program 12P. When the CPU 12A executes the processing by the information processing program 12P, the computer 12X operates as an execution device that executes the switching control processing of the traffic light SG in the traffic light switching control device 12. The information processing program 12P includes a process for realizing various functions for controlling switching of the traffic light SG (details will be described later). The information processing program 12P may be provided by a recording medium such as a CD-ROM.

  Next, the operation of the computer 12X operating as the traffic signal switching control device 12 will be described. In the computer 12X, each of the processes shown in FIG. 6 and FIG. 7 included in the information processing program 12P is executed.

  FIG. 6 shows an example of the flow of processing by the traffic volume predictor learning process included in the information processing program 12P executed by the computer 12X. In the computer 12X, the traffic volume predictor learning process shown in FIG. 6 included in the information processing program 12P is read from the ROM 12C and expanded in the RAM 12B, and the CPU 12A executes the traffic volume predictor learning process expanded in the RAM 12B.

  In step S100, traffic information acquisition processing is executed. In step S100, information including car side information and road side information is acquired as traffic information. The process of step S100 is an example of a process of acquiring information including the vehicle-side information collected by the vehicle-side information collection unit 121 shown in FIG. 1 and the roadside information collected by the roadside information collection unit 122. In the next step S102, processing for storing the traffic information acquired in step S100 in the storage unit 12M as a database is executed.

  The processes in steps S100 and S102 shown in FIG. 6 are an example of the operation of the traffic information storage unit 123 shown in FIG. That is, when the computer 12X executes the processing of step S100 and step S102 shown in FIG. 6, the computer 12X operates as the traffic information storage unit 123 shown in FIG.

  In the process of the next step S104, a learning process of the traffic volume predictor is performed. In step S104, using traffic information stored in the traffic information storage unit 123, a traffic volume predictor that derives information indicating traffic volume as related information is constructed by learning.

  That is, as described above, the traffic volume model is learned so that the output information along the input information is output when a certain input information is input. Thereby, when the average vehicle speed on the road of the road ID, the average vehicle speed of the surrounding road, and the switching timing of the surrounding traffic light are input using the traffic information of the traffic information storage unit 123, A traffic volume predictor is built to predict traffic volume after a few minutes. In the next step S106, the parameters defining the constructed traffic predictor, ie, the parameters in the traffic model, are stored in the storage unit 12M.

  The process of step S104 shown in FIG. 6 is an example of the operation of the related information extraction unit 124 shown in FIG. That is, as the computer 12X executes the process of step S104 shown in FIG. 6, the computer 12X operates as the related information extraction unit 124 shown in FIG. Also, the process of step S106 shown in FIG. 6 is an example of the operation of the optimization data storage unit 125 shown in FIG. That is, as the computer 12X executes the process of step S106 shown in FIG. 6, the computer 12X operates as the optimization data storage unit 125 shown in FIG.

  FIG. 7 shows an example of the process flow of the optimization process included in the information processing program 12P executed by the computer 12X. In the computer 12X, the optimization process shown in FIG. 7 is read from the ROM 12C and expanded on the RAM 12B, and the CPU 12A executes the optimization process expanded on the RAM 12B.

In step S110, the intersection CR in the road network AR is extracted, and in the next step S112, a first intersection model is constructed by a set Ri of roads used at the intersection CRi. That is, to save the road ID indicating all major roads are connected to the intersection CRi (1 ≦ i ≦ 9) , the road set _R i utilized in situations _{s i (s i),} constructing a first intersection model Do.

In the next step S114, traffic information corresponding to the road set R _i (s _i ) is acquired, and in the next step S116, after learning the traffic signal switching timing of the intersection CRi, the learning result is stored in the storage unit 12M. That is, the switching timing of the traffic signal SG is individually learned (Q learning) using the traffic information of the road set R _i (s _i ), and the policy π _i and the state value function V _i are stored.

  In the next step S120, it is determined whether or not an unprocessed intersection CR remains in the road network AR. If a negative determination is made, the process returns to step S112. If an intersection is determined, the process proceeds to step S122. .

  The processes of steps S110 to S120 illustrated in FIG. 7 are an example of the operation of the learning process of the first stage in the signal control optimization unit 126 illustrated in FIG. That is, when the computer 12X executes the processes of steps S110 to S120 shown in FIG. 7, the computer 12X operates as an execution device that executes the learning process of the first step in the signal control optimization unit 126 shown in FIG. 1. . Also, the process of step S118 shown in FIG. 7 is an example of the operation of the optimization data storage unit 125 shown in FIG.

  In the learning process of the second stage, a model is used in which traffic conditions at an intersection CR at a certain time t and adjacent intersections are used as input information and timing parameters defining the switching timing of the signal light 16 of the traffic light SG as output information. That is, a model (hereinafter referred to as a second intersection model) which outputs the switching timing of the traffic light SG installed at the intersection CR according to the traffic condition of the traffic condition of the road including the main road of the intersection CR and the surrounding road of the traffic light SG of the adjacent intersection. Use.).

In step S122, a road to be used at the intersection CRj adjacent to the intersection CRi, that is, a surrounding road is extracted. In the next step S124, a second intersection model is constructed by a set R'i in which the road of the road ID used at the intersection CRj is added to the road set Ri. That is, the road set R _i (s _i ) used in the situation s _i at time t of the intersection CR i (1 ≦ i ≦ 9) is connected to the adjacent intersections CR j (1 ≦ j ≦ 9, i ≠ j) A road set R ' _i (s _i ) is constructed by adding the road IDs of the surrounding roads.

In the next step S126, traffic information corresponding to the road set R ' _i (s _i ) is acquired, and in the next step S128, using the traffic signal switching timing of the learned first intersection model in the second intersection model , Estimate the traffic situation after a predetermined time has passed. That is, the state value function V ' _i (s _i ) at the next time when the traffic lights SG of all intersections CR included in the road set R' _i (s _i ) are switched by the policy learned in advance is observed.

  In the next step S130, the intersection CRi and the intersection CRj are related to the traffic flow control by determining whether the difference between the traffic condition by the road set Ri and the traffic condition estimated in step S126 exceeds the threshold value. Determine if you want to When an affirmative determination is made in step 130, the process proceeds to step S132, and when a negative determination is made, the process proceeds to step S134.

In step S132, the road set Ri is changed to a road set R ′ and after learning, the process proceeds to step S136. In step S134, the process proceeds to step S136 without changing the road set Ri. Ie, as described above, if the difference of the state value exceeds a predetermined threshold value, the intersection CRi and intersections CRj, road aggregate utilized in situations s _i as related to traffic flow control R _i ( Update s _i ) to road set R ′ _i (s _i ), and maintain road set R _i (s _i ) if it does not change below the threshold.

In step S136, the negative judgment is repeated until there is no surrounding road to be added and the values of the policy and the state value function converge, and the process is returned to step S122. On the other hand, if an affirmative decision is made in step S136, the learning for each intersection CR, an optimized parameter policy [pi _i, stores each status value function V _i.

  The processes of steps S122 to S138 shown in FIG. 7 are an example of the operation of the learning process of the second stage in the signal control optimization unit 126 shown in FIG. That is, when the computer 12X executes the processing of steps S122 to S138 shown in FIG. 7, the computer 12X operates as an execution device that executes the learning processing of the second stage in the signal control optimization unit 126 shown in FIG. 1. . Also, the process of step S138 shown in FIG. 7 is an example of the operation of the optimization data storage unit 125 shown in FIG.

  Then, in the case of controlling each of the switching timings of the signal lights 16 of each of the plurality of traffic lights SG included in the road network AR, the traffic volume predictor learned by the processing shown in FIG. The arithmetic unit is reconstructed to control each of the plurality of traffic signals SG. Each of the signal control units 14 of the traffic lighter SG performs switching timing of each of the signal lights 16 by a control signal from the signal control optimization unit 126. This optimizes the traffic flow in the road network AR.

  As described above, according to the present embodiment, when switching control of the traffic signal in the road network, the flow of the vehicle is smoothed compared to the case where the flow of the vehicle is specified from the traffic volume of the main road. Can.

  Here, in the traffic signal switching control system 10 according to the present embodiment, the switching timing of each traffic signal SG of the road network AR is optimized. It is possible to estimate the traffic condition of the road network from the traffic flow when each of the traffic signals SG is switched and controlled by the optimized switching timing, that is, the input traffic information. The estimated traffic condition of the road network is information effective for a vehicle traveling on the road network AR. For example, the vehicle traveling in the road network AR or the vehicle entering the road network is confirmed by confirming the estimated traffic condition of the road network in the vehicle traveling in the road network AR or in the vehicle entering the road network The travel plan can be changed, for example, to reach the destination in a short time. The traffic signal switching control system 10 can also present an optimal travel route to a vehicle traveling on the road network AR or a vehicle entering the road network based on the estimated traffic condition of the road network.

  That is, the traffic signal switching control system 10 according to the present embodiment uses at least one of the vehicle traveling on the road network AR and the vehicle entering the road network using the traffic condition estimated by the traffic information. Can support driving.

  FIG. 8 is a schematic view showing an example of a road network AR including roads where a plurality of vehicles C are expected to be densely packed. In the example illustrated in FIG. 8, a road IDX (an area indicated by hatching) including the roads of road ID 01, ID 02, ID 06, and ID 12 is shown as a road on which a plurality of vehicles C are predicted to be densely packed.

  For a vehicle traveling on a road IDX, while traveling on the road IDX, by transmitting information prompting a change to a road other than the roads of the neighboring roads, that is, roads ID01, ID02, ID06, and ID12. The vehicle can be changed to a travel plan in which the vehicle travels avoiding a dense road. In addition, for vehicles scheduled to enter Road IDX, by changing the information to indicate that multiple vehicles C are concentrated on Road IDX, it is changed to a travel plan that travels by avoiding entry to dense roads. can do. Furthermore, for a vehicle scheduled to enter road network AR, information indicating that a plurality of vehicles C are concentrated on road IDX is transmitted, thereby avoiding traveling of dense roads in road network AR. It can be changed to a plan.

  As mentioned above, although it demonstrated using each embodiment, the technical scope of the technique of an indication is not limited to the range as described in the said embodiment. Various changes or modifications can be added to the above-described embodiment without departing from the scope of the invention, and forms to which the changes or improvements are added are also included in the technical scope of the disclosed technology.

10 traffic signal switching control system 12 traffic signal switching control device 12X computer 12 P information processing program 14 signal control unit 16 signal light 121 car side information collection unit 122 road side information collection unit 123 traffic information storage unit 124 related information extraction unit 125 optimization data storage unit 126 Signal control optimization unit AR Road network C Vehicle CR Intersection SG Traffic light

Claims (7)

For each of the plurality of intersections of the road network provided with a plurality of intersections in which each traffic light is installed, the plurality of intersections are installed at each of the plurality of intersections based on traffic information of a vehicle traveling on a main road connected to each intersection. An operation unit that calculates switching timing of each of the traffic signals;
For each of the traffic lights installed at each of the plurality of intersections, each of the first traffic flows when the traffic lights are switched at the switching timing calculated by the calculation unit based on traffic information of the vehicle traveling on the main road A first estimation unit to estimate;
The traffic information of the vehicle traveling on the main road of the specific intersection and the traffic information of the vehicle traveling on the surrounding road linked to the main road ahead of the specific intersection, for any one specific intersection of the plurality of intersections Second estimation based on the second traffic flow when switching the traffic light installed at the specific intersection and the traffic light installed at the previous intersection at the switching timing calculated by the calculation unit Department,
Optimal to optimize the switching timing of the traffic signal installed at the specific intersection based on the first traffic flow estimated by the first estimation unit and the second traffic flow estimated by the second estimation unit Department,
Signal switching control device including. The arithmetic unit is
The traffic signal switching control device according to claim 1, further comprising: an intersection model of a learned main road so that the traffic information is input and the switching timing of the traffic signal is output so as to be a predetermined traffic flow state. The second estimation unit is
The traffic information and the switching timing of the traffic light are input, and the traffic flow state when switching the traffic light at the input switching timing is output, including an intersection model including the learned surrounding road. The traffic signal switching control device according to 2. The arithmetic unit is
The traffic signal switching control device according to any one of claims 1 to 3, further comprising a storage unit that stores a switching timing of each traffic signal corresponding to the calculated traffic information. The optimization unit
A surrounding road connected to the previous intersection is added until the difference between the first traffic flow estimated by the first estimation unit and the second traffic flow estimated by the second estimation unit is less than a predetermined threshold. The traffic signal switching control device according to any one of claims 1 to 4, wherein the switching timing of the traffic signal installed at the specific intersection is optimized. On the computer
A signal apparatus switching control method comprising performing processing of each part of the signal apparatus switching control device according to any one of claims 1 to 5.   A traffic signal switching control program for causing a computer to function as each part of the traffic signal switching control device according to any one of claims 1 to 5.