Signal lamp scheduling system based on machine learning

Abstract

The invention discloses a signal lamp scheduling system based on machine learning, and relates to the technical field of traffic scheduling, the signal lamp scheduling system comprises a signal lamp management and control service platform, the signal lamp management and control service platform is connected with a vehicle information acquisition module, a traffic road condition monitoring module, a machine iterative learning module, a signal lamp scheduling module and a real-time monitoring circulation module; the vehicle information acquisition module is used for acquiring related information of the vehicle-mounted navigation terminal and calibrating vehicle-mounted positioning; the traffic road condition monitoring module is used for monitoring lane information of a traffic road and analyzing main influence factors of the monitored lane information; the machine iterative learning module is used for carrying out iterative learning based on the main influence factors causing the road congestion so as to generate a tree diagram structure model; the signal lamp scheduling module is used for scheduling traffic signal lamps according to the result of the tree diagram structure model; and the real-time monitoring circulation module is used for carrying out real-time scheduling on the signal lamps through circulation among the modules.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic scheduling, and specifically to a traffic signal scheduling system based on machine learning. Background Art

[0002] At present, traffic congestion has become a serious problem in our daily life. As the main carrier of urban traffic, the number of cars has been increasing year by year, and the resulting road congestion has become one of the problems faced in urban traffic. With the rapid development of technology and the continuous improvement and popularization of artificial intelligence and 5G technology, intelligent transportation will have a broader development prospect and will achieve more intelligent and precise management and control. Through the 5G network and intelligent transportation, data sharing and information interaction will be realized, and cooperation between vehicles and between vehicles and road infrastructure will be achieved.

[0003] In the prior art, the control of traffic lights at intersections in most cities generally adopts a control method with a fixed conversion time interval. This method conducts a prior investigation of the passing vehicle flow and uses statistical methods to preset the delay of the traffic lights. However, due to the complexity, randomness, and uncertainty of the vehicle flow at intersections at different times, the use of a fixed-time control method often results in a waste of the effective utilization time of the road. To solve these problems, a traffic signal scheduling system based on machine learning is provided herein. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a traffic signal scheduling system based on machine learning.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A traffic signal scheduling system based on machine learning includes a traffic signal control service platform, which is connected to a vehicle information collection module, a traffic condition monitoring module, a machine iterative learning module, a traffic signal scheduling module, and a real-time monitoring loop module;

[0006] The vehicle information collection module is used to collect relevant information of the in-vehicle navigation terminal and calibrate the vehicle location, and at the same time transmit the collected relevant information of the in-vehicle navigation terminal and the location information after vehicle location calibration to the traffic signal control service platform;

[0007] The vehicle information collection module is provided with an in-vehicle navigation terminal unit and a vehicle location module unit;

[0008] The in-vehicle navigation terminal unit is used to collect relevant information of the in-vehicle navigation terminal;

[0009] The relevant information of the in-vehicle navigation terminal includes: departure place, destination, and route;

[0010] The process of collecting information related to the in-vehicle navigation terminal includes:

[0011] The user determines the departure location and the destination through the in-vehicle navigation terminal;

[0012] The in-vehicle navigation terminal provides at least one route from the departure location to the destination for the user to select;

[0013] The in-vehicle navigation terminal determines the route selected by the user and uploads the route to the signal light control service platform;

[0014] The in-vehicle navigation terminal provides navigation services based on the route selected by the user;

[0015] The in-vehicle positioning module unit is used to collect in-vehicle positioning information and calibrate in-vehicle positioning;

[0016] The in-vehicle positioning module unit installs vehicle sensors;

[0017] The vehicle sensors calibrate vehicle positioning by measuring the distance between the vehicle and surrounding buildings during vehicle travel;

[0018] Establish a two-dimensional plane coordinate system O-XY based on the vehicle;

[0019] Set the vehicle as the coordinate origin, the forward direction of the vehicle as the X-axis, and the right side direction of the vehicle as the Y-axis;

[0020] At the same time, set the forward direction of the vehicle to always be forward;

[0021] Set the range within which the vehicle sensors identify buildings during vehicle travel;

[0022] The process by which the vehicle sensors achieve vehicle positioning calibration includes;

[0023] On the one hand, during vehicle travel, if the vehicle sensors identify the presence of a building in any direction, the vehicle sensors continue to identify the building in a direction perpendicular to the identified building;

[0024] Based on the direction and the buildings identified perpendicular to this direction, compare with the map provided by the in-vehicle navigation terminal unit, and then calibrate in-vehicle positioning;

[0025] On the other hand, during vehicle travel, based on the map provided by the in-vehicle navigation terminal unit, determine the current position information of the vehicle, and obtain the building information in the Y-axis direction during vehicle travel;

[0026] During vehicle travel, the vehicle sensors detect whether there is such a building in the Y-axis direction based on the position information of the vehicle;

[0027] If the building exists, it indicates that the vehicle positioning is relatively accurate and no calibration of in-vehicle positioning is required;

[0028] If the building does not exist, the in-vehicle positioning module unit tracks the vehicle's historical driving route and detects whether the building exists in the vehicle's historical driving route;

[0029] If the building exists, the vehicle sensor calibrates the in-vehicle positioning based on the position where the building exists and the vehicle's position information;

[0030] If the building does not exist, the in-vehicle positioning module unit continuously tracks the vehicle's current driving route and calibrates the in-vehicle positioning based on the position where the building exists and the vehicle's position information;

[0031] Meanwhile, the vehicle information collection module determines whether the vehicle's driving trajectory is consistent with the route selected by the user on the in-vehicle navigation terminal unit based on the vehicle position information provided by the in-vehicle positioning module unit after in-vehicle positioning calibration;

[0032] If the routes are consistent, the in-vehicle information collection module continues to collect the vehicle's position information and in-vehicle navigation information and transmits them to the signal light control service platform;

[0033] If the routes are inconsistent, the in-vehicle navigation terminal unit re-plans at least one route that can reach the destination based on the vehicle's current position information and the destination for the user to re-select, and continuously collects the vehicle position information to determine whether the vehicle's current position information is consistent with the route re-selected by the user;

[0034] The traffic condition monitoring module monitors the lane information of the traffic roads, and the process of analyzing the main influencing factors of the monitored lane information includes:

[0035] The traffic condition monitoring module is set at each intersection, monitors the lane information of the vehicles coming from each direction at the intersection, and sends the monitored lane information to the signal light control service platform;

[0036] Analyze the monitored lane information;

[0037] Eliminate the abnormal lane information after analysis;

[0038] Analyze the lane information after eliminating the abnormal lane information to obtain the main influencing factors affecting the lane information;

[0039] The main influencing factors are divided into two aspects: direct influencing factors and indirect influencing factors;

[0040] The machine iterative learning module conducts iterative learning based on the main influencing factors causing road congestion, and the process of generating a tree diagram structure model includes:

[0041] Feature selection for iterative learning and generation of an iterative learning model;

[0042] The process of iterative learning feature selection includes:

[0043] Set each indirect influencing factor that causes road congestion as a feature for iterative learning;

[0044] The process of generating a tree diagram structure model includes:

[0045] Take the training dataset D and feature a as the input of the iterative learning model, and the output is the information gain Gain(D,a) of feature a for the training dataset D;

[0046] Specify that the training dataset D is the main influencing factor causing road congestion and feature a is each indirect influencing factor causing road congestion;

[0047] Calculate the information entropy of the training dataset D;

[0048] Set the proportion of the i-th type of feature in the training dataset D as p i (i = 1, 2, ……, |n|);

[0049]

[0050] where Ent(D) is the information entropy;

[0051] Calculate the conditional information entropy of feature a for the training dataset D;

[0052]

[0053] where Ent(D|a) is the conditional information entropy;

[0054] Calculate the information gain;

[0055] Gain(D,a) = Ent(D) - Ent(D|a)

[0056] where Gain(D,a) is the information gain;

[0057] Obtain the maximum value of the information gain and denote it as the maximum information gain;

[0058] Present the iterative learning model as a tree diagram structure model and perform node division on the tree diagram structure model according to the obtained maximum information gain;

[0059] The tree - like structure model generated by the iterative learning model calculates the information gain for each feature, takes the feature corresponding to the maximum information gain as the first node, and establishes the second node based on the feature corresponding to the maximum information gain among the remaining features, and so on, to complete the generation of the tree - like structure model. If there are cases where the information gains corresponding to features are the same, the features with the same information gain are taken as nodes at the same level;

[0060] The signal light scheduling module is used to schedule traffic signal lights according to the results of the tree - like structure model;

[0061] The completed tree - like structure model after learning is used as the machine iterative learning model for automatic traffic signal light scheduling;

[0062] Set the adjustment state of the traffic signal lights according to the results output by the machine iterative learning model;

[0063] The real - time monitoring loop module is used to perform real - time scheduling of signal lights through a loop;

[0064] When the signal light scheduling provided by the machine iterative learning model cannot meet the requirements of signal light scheduling in actual traffic, the machine iterative learning module regenerates the iterative learning model based on the traffic conditions monitored by the traffic condition monitoring module, and controls the signal light scheduling module according to the results output by the generated tree - like structure model.

[0065] Compared with the prior art, the beneficial effects of the present invention are as follows: The vehicle positioning module uses the method of installing vehicle sensors to measure the distance between the vehicle and surrounding buildings during vehicle driving, and then calibrates the vehicle positioning, improving the accuracy of vehicle positioning information; on the other hand, the machine iterative learning module realizes the prediction of lane information through feature selection in iterative learning and the generation of the iterative learning model, and then completes the scheduling of traffic signal lights. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is the schematic diagram of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] As Figure 1 shown, a signal light scheduling system based on machine learning, which relates to the technical field of traffic scheduling, includes a signal light control service platform. The signal light control service platform is connected to a vehicle information collection module, a traffic condition monitoring module, a machine iterative learning module, a signal light scheduling module, and a real - time monitoring loop module;

[0068] The vehicle information collection module is used to collect relevant information of the in - vehicle navigation terminal and calibrate vehicle positioning, and at the same time transmit the collected relevant information of the in - vehicle navigation terminal and the position information after vehicle positioning calibration to the signal light control service platform;

[0069] The vehicle information collection module is provided with an in-vehicle navigation terminal unit and an in-vehicle positioning module unit;

[0070] The in-vehicle navigation terminal unit is used to collect relevant information of the in-vehicle navigation terminal;

[0071] The relevant information of the in-vehicle navigation terminal includes: departure place, destination, and route;

[0072] The process of collecting relevant information of the in-vehicle navigation terminal includes:

[0073] The user determines the departure place and destination through the in-vehicle navigation terminal;

[0074] The in-vehicle navigation terminal provides at least one route from the departure place to the destination for the user to select;

[0075] The in-vehicle navigation terminal determines the route selected by the user and uploads the route to the signal light control service platform;

[0076] The in-vehicle navigation terminal provides navigation services based on the route selected by the user;

[0077] When the vehicle passes through an intersection during the driving process on this route, the in-vehicle navigation terminal unit prompts the vehicle's driving route, and the prompt content includes at least the driving direction and the selection of lanes;

[0078] The in-vehicle positioning module unit is used to collect in-vehicle positioning information and calibrate the in-vehicle positioning;

[0079] Currently, common in-vehicle positioning modules all use GPS for positioning, but when the buildings around the vehicle's location are relatively complex, the positioning accuracy of GPS is poor;

[0080] The in-vehicle positioning module unit installs vehicle sensors;

[0081] The vehicle sensors calibrate the vehicle positioning by measuring the distance between the vehicle and the surrounding buildings during the vehicle's driving process;

[0082] Establish a two-dimensional plane coordinate system O-XY based on the vehicle;

[0083] Set the vehicle as the coordinate origin, the forward direction of the vehicle as the X-axis, and the right side direction of the vehicle as the Y-axis;

[0084] At the same time, set the forward direction of the vehicle to always be forward. When the vehicle turns left, turns right, or makes a U-turn, the established two-dimensional plane coordinate system turns left, turns right, or makes a U-turn with the vehicle, so that the forward direction of the vehicle is always forward in the two-dimensional plane coordinate system;

[0085] Set the range within which the vehicle sensors identify buildings during the vehicle's driving process;

[0086] The process of a vehicle sensor realizing vehicle positioning calibration includes:

[0087] On the one hand, when the vehicle sensor detects the presence of a building in any direction during the vehicle's driving process, the vehicle sensor continues to identify the building in the direction perpendicular to the identified building;

[0088] Based on the direction and the buildings identified perpendicular to this direction, a comparison is made with the map provided by the in-vehicle navigation terminal unit, and then the in-vehicle positioning is calibrated;

[0089] On the other hand, during the vehicle's driving process, the vehicle determines its current position information based on the map provided by the in-vehicle navigation terminal unit, and obtains the building information in the Y-axis direction during the vehicle's driving process;

[0090] During the vehicle's driving process, the vehicle sensor detects whether there is such a building in the Y-axis direction based on the vehicle's position information;

[0091] If there is such a building, it indicates that the vehicle positioning is relatively accurate and no in-vehicle positioning calibration is required;

[0092] If there is no such building, the in-vehicle positioning module unit tracks the vehicle's historical driving route and detects whether there is such a building in the vehicle's historical driving route;

[0093] If there is such a building, the vehicle sensor calibrates the in-vehicle positioning based on the position where the building exists and the vehicle's position information;

[0094] If there is no such building, the in-vehicle positioning module unit continuously tracks the vehicle's current driving route and calibrates the in-vehicle positioning based on the position where the building exists and the vehicle's position information;

[0095] Meanwhile, the vehicle information collection module determines whether the vehicle's driving trajectory is consistent with the route selected by the user on the in-vehicle navigation terminal unit based on the vehicle position information after in-vehicle positioning calibration provided by the in-vehicle positioning module unit;

[0096] If the routes are consistent, the vehicle information collection module continues to collect the vehicle's position information and in-vehicle navigation information and transmits them to the signal light control service platform;

[0097] If the routes are inconsistent, the in-vehicle navigation terminal unit re-plans at least one route that can reach the destination based on the vehicle's current position information and the destination for the user to re-select, and continuously collects the vehicle position information to determine whether the vehicle's current position information is consistent with the route re-selected by the user;

[0098] The in-vehicle navigation terminal unit uses vehicle sensors to collect vehicle position information and calibrate in-vehicle positioning, improving the accuracy of vehicle position information and providing reliable vehicle position information for the in-vehicle navigation terminal unit;

[0099] The traffic condition monitoring module is used to monitor the lane information of traffic roads and analyze the main influencing factors of the monitored lane information;

[0100] The traffic condition monitoring module is set at each intersection to monitor the lane information of vehicles coming from each direction at the intersection and send the monitored lane information to the signal light control service platform;

[0101] The lane information includes: vehicle passing information, lane queuing situation;

[0102] Analyze and eliminate the monitored lane information;

[0103] Grant the traffic condition monitoring staff the permission to eliminate abnormal lane information;

[0104] The abnormal lane information includes: road congestion caused by traffic accidents, road congestion caused by vehicle violations;

[0105] Analyze the lane information after eliminating abnormal lane information to obtain the main influencing factors causing road congestion;

[0106] The main influencing factors are divided into two aspects: direct influencing factors and indirect influencing factors;

[0107] The direct influencing factors include: traffic flow influencing factors;

[0108] The traffic flow influencing factor refers to the increase in traffic flow caused by the increase in the number of vehicles entering the intersection;

[0109] The indirect influencing factors include: weather influencing factors, date influencing factors, time influencing factors, etc.;

[0110] For the same intersection, the lane information during a day will show certain periodic changes according to different time periods; the lane information within a week will also show certain periodic changes according to the week numbers; the lane information within a certain time will also show certain periodic changes according to the weather;

[0111] The certain periodic changes presented by the indirect influencing factors to the lane information bring convenience to our understanding of the conditions and laws of modern urban roads, improving the accuracy and reliability of collecting and statistical urban traffic data;

[0112] The machine iterative learning module is used to perform iterative learning based on the main influencing factors causing road congestion, and then generate a tree diagram structure model;

[0113] The process of iterative learning by the machine iterative learning module includes:

[0114] Feature selection for iterative learning and generation of an iterative learning model;

[0115] The process of feature selection for iterative learning includes:

[0116] Set each indirect influencing factor that causes road congestion as a feature for iterative learning;

[0117] Assume that the features for iterative learning include: whether the weather condition is normal, whether the traffic flow is normal, whether it is in the morning rush hour, whether it is in the evening rush hour, whether it is a working day, whether it is a holiday, whether the vehicle navigation terminal passes through this intersection, whether it is congested within the rated time, etc.;

[0118] The rated time is set by the staff who manage the machine iterative learning module, such as 30 minutes or 60 minutes;

[0119] The process of generating a tree diagram structure model includes:

[0120] Take the training data set D and the feature a as the input of the iterative learning model, and the output is the information gain Gain(D,a) of the feature a for the training data set D;

[0121] Specify that the training data set D is the main influencing factor causing road congestion, and the feature a is each indirect influencing factor causing road congestion;

[0122] Calculate the information entropy of the training data set D;

[0123] Set the proportion of the i-th type of feature in the training data set D as p i (i = 1, 2, ……, |n|);

[0124] Among them, Ent(D) is the information entropy;

[0125] Calculate the conditional information entropy of the feature a for the training data set D;

[0126]

[0127] Among them, Ent(D|a) is the conditional information entropy;

[0128] Calculate the information gain;

[0129] Gain(D,a) = Ent(D) - Ent(D|a)

[0130] Among them, Gain(D,a) is the information gain;

[0131] Obtain the maximum value of the information gain and record it as the maximum information gain;

[0132] The iterative learning model is presented in a tree - diagram structure model, and the nodes of the tree - diagram structure model are divided according to the maximum information gain obtained.

[0133] The tree - diagram structure model generated by the iterative learning model calculates the information gain for each feature. The feature corresponding to the maximum information gain is used as the first node, and the feature corresponding to the maximum information gain among the remaining features is used to establish the second node, and so on, to complete the generation of the tree - diagram structure model. If there are cases where the information gains corresponding to features are the same, the features with the same information gain are used as nodes at the same level.

[0134] The tree - diagram structure model generated by the iterative learning model is a non - parametric model and does not require prior assumptions about the training data set. Therefore, it can handle complex samples, has a fast calculation speed, strong result interpretability, is not sensitive to missing values, and can also handle classification and prediction problems.

[0135] The machine iterative learning module generates an effective tree - diagram structure model for the main influencing factors causing road traffic jams, and then schedules the time of traffic lights, which not only saves police force but also improves efficiency.

[0136] The signal light scheduling module is used to schedule traffic lights according to the results of the tree - diagram structure model.

[0137] The tree - diagram structure model after learning is used as the machine iterative learning model for automatic traffic light scheduling.

[0138] Set the adjustment status of traffic lights according to the results output by the machine iterative learning model.

[0139] The adjustment status of traffic lights includes: need to increase, need to maintain, need to decrease.

[0140] The real - time monitoring loop module is used to combine the traffic condition monitoring module, the machine iterative learning module and the signal light scheduling module, and perform real - time scheduling of signal lights through a loop.

[0141] When the signal light scheduling provided by the machine iterative learning model cannot meet the requirements of signal light scheduling in actual traffic, such as the following situations:

[0142] In the case where the machine iterative learning model believes that the green light time should be increased, but there is no vehicle passing through.

[0143] In the case where the machine iterative learning model believes that the green light time should be reduced, but the vehicle queue length is continuously increasing.

[0144] At this time, the machine iterative learning module regenerates the iterative learning model based on the traffic conditions monitored by the traffic condition monitoring module, and controls the signal lamp scheduling module according to the output result of the generated tree diagram structure model, so as to realize the signal lamp scheduling based on machine learning according to the changes in real-time traffic conditions.

[0145] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A traffic light dispatching system based on machine learning, including a traffic light management and control service platform, characterized in that: The traffic light control service platform is connected to a vehicle information collection module, a traffic condition monitoring module, a machine iterative learning module, a traffic light scheduling module and a real-time monitoring cycle module; The vehicle information collection module is provided with a vehicle navigation terminal unit and a vehicle positioning module unit; The vehicle navigation terminal unit is used to collect relevant information of the vehicle navigation terminal; The vehicle positioning module unit is used to collect vehicle positioning information and perform vehicle positioning calibration; The traffic condition monitoring module is used to monitor the lane information of the traffic road and analyze the main influencing factors of the monitored lane information; The machine iterative learning module is used to perform iterative learning based on the main influencing factors causing road congestion, and then generate a tree diagram structure model; The traffic light scheduling module is used to schedule traffic lights according to the tree diagram structure model results; The real-time monitoring cycle module is used to schedule the traffic lights in real time through cycles between modules.

2. The signal light dispatching system based on machine learning according to claim 1, characterized in that: The process of the vehicle navigation terminal unit collecting relevant information of the vehicle navigation terminal includes: The relevant information of the vehicle navigation terminal includes: departure place, destination, and route; The user determines the departure and destination through the in-vehicle navigation terminal; The in-vehicle navigation terminal provides at least one route from a departure point to a destination for the user to choose; The in-vehicle navigation terminal determines the route selected by the user and uploads the route to the traffic light control service platform; The in-vehicle navigation terminal provides navigation services based on the route selected by the user.

3. The traffic light dispatching system based on machine learning according to claim 1, characterized in that: The process of the vehicle positioning module collecting vehicle positioning information and performing vehicle positioning calibration includes: Install vehicle sensors; The vehicle sensor calibrates the vehicle positioning by measuring the distance between the vehicle and surrounding buildings during driving; Establish a two-dimensional plane coordinate system O-XY based on the vehicle; Set the vehicle as the coordinate origin, the vehicle's forward direction as the X-axis, and the vehicle's right direction as the Y-axis; At the same time, the vehicle's forward direction is always set to forward; Setting the range of buildings that the vehicle sensor can identify during vehicle driving; If the vehicle sensor recognizes the existence of a building in any direction during driving, the vehicle sensor continues to recognize the building in a direction perpendicular to the recognized building, and compares the building recognized in this direction and perpendicular to this direction with the map provided by the vehicle navigation terminal unit, and then calibrates the vehicle positioning; During driving, the current location of the vehicle is determined based on the map provided by the vehicle navigation terminal unit, and the building information in the Y-axis direction is obtained during the driving process. The vehicle sensor detects whether there is a building in the Y-axis direction based on the vehicle's location information during driving, and then calibrates the vehicle positioning.

4. The traffic light dispatching system based on machine learning according to claim 1, characterized in that: The traffic condition monitoring module monitors the lane information of the traffic road and analyzes the main influencing factors of the monitored lane information. The process includes: The traffic condition monitoring module is set up at each intersection to monitor the lane information of vehicles coming from each direction at the intersection, and send the monitored lane information to the traffic light control service platform; Analyze the monitored lane information and remove abnormal lane information; Analyze the lane information after removing abnormal lane information to obtain the main influencing factors causing road congestion.

5. The traffic light dispatching system based on machine learning according to claim 1, characterized in that: The process of iterative learning by the machine iterative learning module based on the main factors causing road congestion includes: The feature selection of iterative learning sets each indirect influencing factor causing road congestion as a feature of iterative learning, constructs an iterative learning model according to the selected features, and generates a corresponding tree diagram structure model according to the constructed iterative learning model.

6. The signal light dispatching system based on machine learning according to claim 5, characterized in that: The generation process of the tree structure model includes: The training data set D and feature a are used as the input of the iterative learning model, and the output is the information gain Gain(D,a) of feature a to the training data set D; It is stipulated that the training data set D is the main influencing factor causing road congestion, and the feature a is each indirect influencing factor causing road congestion; Calculate the information entropy of the training data set D; Set the proportion of the i-th feature in the training data set D to p i (i=1,2,……,|n|); Among them, Ent(D) is information entropy; Calculate the conditional information entropy of feature a for training data set D; Among them, Ent(D|a) is the conditional information entropy; Calculate information gain; Gain(D,a)=Ent(D)-Ent(D|a); Among them, Gain(D,a) is the information gain; Get the maximum value of information gain and record it as the maximum information gain; The iterative learning model is presented in a tree diagram structure model, and the tree diagram structure model is node-divided according to the maximum information gain obtained; The feature corresponding to the maximum information gain is used as the first node, and the second node is established according to the feature corresponding to the maximum information gain among the remaining features, and so on, to complete the generation of the tree diagram structure model.

7. The traffic light dispatching system based on machine learning according to claim 1, characterized in that: The process of the traffic light scheduling module scheduling traffic lights according to the results of the tree diagram structure model includes: The tree graph structure model after learning is used as a machine iterative learning model for automatic scheduling of traffic lights; The adjustment state of the traffic light is set according to the output of the machine iterative learning model.

8. The traffic light dispatching system based on machine learning according to claim 1, characterized in that: The process of real-time monitoring and learning module to schedule traffic lights in real time through the cycle between modules includes: When the traffic light scheduling provided by the machine iterative learning model cannot meet the needs of traffic light scheduling in actual traffic, the machine iterative learning module regenerates the iterative learning model based on the traffic conditions monitored by the traffic condition monitoring module, and controls the traffic light scheduling module according to the output results of the generated tree graph structure model.

Images