Intelligent traffic signal control method and system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent traffic control processing, in particular to an intelligent traffic signal control method and system based on the Internet of Things, and the method comprises the steps: obtaining a flow propagation coefficient, achieving the dynamic flow distribution, obtaining an optimized flow balance factor in the current direction, and the like. According to the method, the problems that in the prior art, due to the fact that part of intersections cannot adapt to real-time flow changes, the queuing length is accumulated rapidly, the vehicle speed is reduced, and regional congestion is caused are solved, urban traffic congestion can be effectively relieved, the road passing efficiency is improved, and the traffic safety is improved. And an innovative solution is provided for smart city traffic management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent traffic control processing, and in particular to an intelligent traffic signal control method and system based on the Internet of Things. Background Art

[0002] The field of intelligent transportation is a core pillar of modern urban development and is directly related to the improvement of traffic efficiency, energy consumption and residents' quality of life. Traditional traffic signal control methods mostly rely on fixed durations or simple sensor-based adjustments. Although these solutions have alleviated congestion to a certain extent, they are unable to adapt to real-time traffic changes when faced with complex and changing traffic scenarios. The sensor control optimized at a single intersection lacks regional coordination, resulting in limited overall improvement in traffic conditions. In particular, during peak hours or emergencies, the efficiency of the road network decreases significantly. In summary, the existing technology has the problem that some intersections cannot adapt to real-time traffic changes, resulting in a rapid accumulation of queue lengths, a decrease in vehicle speeds, and causing regional congestion. Summary of the Invention

[0003] In view of the problems existing in the prior art, the present invention provides an intelligent traffic signal control method and system based on the Internet of Things.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, an intelligent traffic signal control method based on the Internet of Things comprises:

[0006] S102: Based on node traffic and historical data, the traffic density change trend is calculated using a deep learning prediction model. Key features are extracted from the traffic density change prediction value, and the current traffic density is adjusted through data comparison. The traffic propagation coefficient is calculated based on the corrected traffic density sequence to obtain the traffic propagation coefficient.

[0007] S103: Based on the traffic propagation coefficient, the traffic state variation range is determined through data fusion analysis. The intersection state is analyzed and judged to obtain a preliminary saturation determination result. The initial quantized value of the edge delay time is obtained by extracting the fluctuation characteristics. The quantized value is adjusted using a linear regression algorithm to obtain a corrected edge delay time. The presence of a local congestion point is determined based on the intersection state. The direction of flow adjustment is determined using a support vector machine algorithm to achieve dynamic flow distribution.

[0008] S104: Based on the quantized edge delay values, an adjacency weight matrix is calculated using a graph network algorithm. Traffic concentration areas within the region boundaries are divided using a clustering algorithm. The average delay value within each region boundary is calculated. Traffic fluctuation characteristics between intersections are obtained from the factor distribution. The traffic balancing factor is adjusted using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction.

[0009] S105: Based on the distribution of flow balancing factors and the trend of traffic density changes, a traffic density adjustment coefficient is calculated using a collaborative optimization algorithm. A baseline adjustment offset is calculated based on the trend data. The adjustment coefficient is integrated using a linear regression algorithm. Fluctuation characteristics are extracted from the traffic density. The green light duration distribution is compared with the duration plan based on a preset threshold to determine whether to adjust the green light duration distribution. The traffic density adjustment result after collaborative optimization is calculated by adjusting the green light duration distribution to obtain the final green light duration for the current direction.

[0010] S106: Based on the gain data obtained from real-time feedback, the verification process result is obtained by determining whether the green light duration is lower than the minimum duration baseline. The time window boundary is determined by adjusting the interval range. The smoothing rate distribution is calculated based on the time window boundary. The adjusted interval offset is obtained by adjusting the baseline through gain data fusion. The light position switching characteristics are updated by adjusting the interval offset. The smoothing rate change trend is compared with the minimum baseline to obtain the final green light switching interval for the current direction.

[0011] S107: Based on the smoothing rate fluctuation range, the smoothing rate allocation range is determined by the time window boundary. By determining the scope of the intersection relevance on the neighborhood influence radius, the real-time traffic data and vehicle speed change values are classified using the support vector machine algorithm. The signal light parameters are updated based on the parameter adjustment amount. The optimized sensor data stream is obtained by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval.

[0012] S108: Based on the full-time predicted intersection traffic density, the initial analysis results are obtained by extracting the change trend and deviation data, and the traffic propagation coefficient is calculated in combination with the dynamic update frequency. The traffic propagation model is adjusted by the deviation data, and the green light duration calculation logic is updated with the corrected coefficient. The baseline plan is generated with the adjusted duration value, and the plan is optimized in combination with the traffic density change trend. The green light duration allocation logic is updated according to the final adjustment result to output the optimized baseline plan.

[0013] At the same time, the present invention also provides an intelligent traffic signal control system based on the Internet of Things, including:

[0014] Based on the above intelligent traffic signal control method, the intelligent traffic signal control system includes:

[0015] Module 101: Based on the collected real-time traffic data of multiple groups of different types at intersections, a multi-dimensional parameter set for the current direction is obtained through multi-dimensional parameter combination, a smoothed parameter sequence is obtained through time series analysis, and traffic distribution characteristics are determined by calculating the node traffic values in each direction to achieve data preprocessing;

[0016] Module 102: Calculates the traffic density change trend based on node traffic and historical data using a deep learning prediction model, extracts key features from the traffic density change prediction value, adjusts the current traffic density through data comparison, and calculates the traffic propagation coefficient based on the corrected traffic density sequence to obtain the traffic propagation coefficient;

[0017] Module 103: Based on the traffic propagation coefficient, data fusion analysis is used to determine the range of traffic state changes. Analysis and judgment of the intersection state are used to obtain a preliminary saturation determination result. Initial quantitative values of the edge delay time are obtained by extracting fluctuation characteristics. The quantitative values are adjusted using a linear regression algorithm to obtain a corrected edge delay time. Based on the intersection state, the presence of local congestion points is determined. A support vector machine algorithm is used to determine the direction of flow adjustment to achieve dynamic flow distribution.

[0018] Module 104: Calculates an adjacency weight matrix based on the quantized edge delay time using a graph network algorithm, divides the traffic concentration areas within the area boundary using a clustering algorithm, calculates the average delay value within each area boundary, obtains traffic fluctuation characteristics between intersections from the factor distribution, and adjusts the traffic balancing factor using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction;

[0019] Module 105: Calculates the traffic density adjustment coefficient using a collaborative optimization algorithm based on the distribution of traffic balancing factors and the trend of traffic density changes. Calculates the baseline adjustment offset based on the trend data. Fusions the adjustment coefficient using a linear regression algorithm. Extracts fluctuation characteristics from traffic density. Determines whether to adjust the green light duration distribution based on a preset threshold and comparison with the duration scheme. Calculates the traffic density adjustment result after collaborative optimization by adjusting the green light duration distribution to obtain the final green light duration for the current direction.

[0020] Module 106: Based on the gain data obtained from real-time feedback, it determines whether the green light duration is lower than the minimum duration baseline to obtain the verification process result, determines the time window boundary by adjusting the interval range, calculates the smoothing rate distribution based on the time window boundary, obtains the adjusted interval offset by adjusting the baseline through gain data fusion, updates the light position switching characteristics based on the adjusted interval offset, and compares the smoothing rate change trend with the minimum baseline to obtain the final green light switching interval for the current direction;

[0021] Module 107: used to determine the smoothing rate allocation range based on the smoothing rate fluctuation range and the time window boundary, classify the real-time traffic data and vehicle speed change values by judging the scope of the intersection relevance on the neighborhood influence radius, update the traffic light parameters by parameter adjustment, and obtain the optimized sensor data stream by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval;

[0022] Module 108: It is used to obtain the initial analysis results based on the traffic density of the intersection predicted for the entire period by extracting the change trend and deviation data, calculate the traffic propagation coefficient in combination with the dynamic update frequency, adjust the traffic propagation model through the deviation data, update the green light duration calculation logic through the corrected coefficient, generate the baseline plan through the adjusted duration value, optimize the plan in combination with the traffic density change trend, and update the green light duration allocation logic according to the final adjustment result to output the optimized baseline plan.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The intelligent traffic signal control method includes seven main steps, wherein the seven main steps of the method include: obtaining a flow propagation coefficient, realizing dynamic flow distribution, obtaining an optimized current direction flow balancing factor, obtaining a final current direction green light duration, obtaining a final current direction green light switching interval, obtaining collaborative optimization of the green light duration and the green light switching interval, and outputting an optimized baseline solution. Based on the above seven main steps, the present invention provides an intelligent traffic signal control method based on the Internet of Things. The present invention discloses an intelligent traffic signal control method based on the Internet of Things and deep learning. The method collects flow, vehicle speed, and queue length data in real time through road sensors, and calculates the traffic density change trend using time series analysis and deep learning prediction models. According to the flow propagation coefficient, edge delay time, and flow balancing factor, the present invention adopts a collaborative optimization algorithm to dynamically adjust the green light duration, and optimizes the light position switching interval by real-time feedback gain data. When the degree of improvement in traffic conditions is lower than expected, the present invention can extract deviation data from the prediction model, recalculate the flow propagation coefficient, and realize adaptive adjustment of the green light duration. This method can effectively alleviate urban traffic congestion, improve road traffic efficiency, and provide an innovative solution for smart city traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a specific flow chart of Example 1 of the present invention;

[0026] Figure 2 It is a structural diagram of an intelligent traffic signal control system based on the Internet of Things provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] It is worth noting that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products, and their sources are not specifically limited unless otherwise specified.

[0028] The present invention will be further described below with reference to specific embodiments, but the protection scope of the present invention is not limited thereto.

[0029] Figure 1Flowchart according to an embodiment of the present invention, Figure 1 As shown, the embodiments of the present invention include claims:

[0030] like Figure 1 As shown, an intelligent traffic signal control method based on the Internet of Things is characterized by comprising:

[0031] S102: Based on node traffic and historical data, the traffic density change trend is calculated using a deep learning prediction model. Key features are extracted from the traffic density change prediction value, and the current traffic density is adjusted through data comparison. The traffic propagation coefficient is calculated based on the corrected traffic density sequence to obtain the traffic propagation coefficient.

[0032] S103: Based on the traffic propagation coefficient, the traffic state variation range is determined through data fusion analysis. The intersection state is analyzed and judged to obtain a preliminary saturation determination result. The initial quantized value of the edge delay time is obtained by extracting the fluctuation characteristics. The quantized value is adjusted using a linear regression algorithm to obtain a corrected edge delay time. The presence of a local congestion point is determined based on the intersection state. The direction of flow adjustment is determined using a support vector machine algorithm to achieve dynamic flow distribution.

[0033] S104: Based on the quantized edge delay values, an adjacency weight matrix is calculated using a graph network algorithm. Traffic concentration areas within the region boundaries are divided using a clustering algorithm. The average delay value within each region boundary is calculated. Traffic fluctuation characteristics between intersections are obtained from the factor distribution. The traffic balancing factor is adjusted using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction.

[0034] S105: Based on the distribution of flow balancing factors and the trend of traffic density changes, a traffic density adjustment coefficient is calculated using a collaborative optimization algorithm. A baseline adjustment offset is calculated based on the trend data. The adjustment coefficient is integrated using a linear regression algorithm. Fluctuation characteristics are extracted from the traffic density. The green light duration distribution is compared with the duration plan based on a preset threshold to determine whether to adjust the green light duration distribution. The traffic density adjustment result after collaborative optimization is calculated by adjusting the green light duration distribution to obtain the final green light duration for the current direction.

[0035] S106: Based on the gain data obtained from real-time feedback, the verification process result is obtained by determining whether the green light duration is lower than the minimum duration baseline. The time window boundary is determined by adjusting the interval range. The smoothing rate distribution is calculated based on the time window boundary. The adjusted interval offset is obtained by adjusting the baseline through gain data fusion. The light position switching characteristics are updated by adjusting the interval offset. The smoothing rate change trend is compared with the minimum baseline to obtain the final green light switching interval for the current direction.

[0036] S107: Based on the smoothing rate fluctuation range, the smoothing rate allocation range is determined by the time window boundary. By determining the scope of the intersection relevance on the neighborhood influence radius, the real-time traffic data and vehicle speed change values are classified using the support vector machine algorithm. The signal light parameters are updated based on the parameter adjustment amount. The optimized sensor data stream is obtained by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval.

[0037] S108: Based on the full-time predicted intersection traffic density, the initial analysis results are obtained by extracting the change trend and deviation data, and the traffic propagation coefficient is calculated in combination with the dynamic update frequency. The traffic propagation model is adjusted by the deviation data, and the green light duration calculation logic is updated with the corrected coefficient. The baseline plan is generated with the adjusted duration value, and the plan is optimized in combination with the traffic density change trend. The green light duration allocation logic is updated according to the final adjustment result to output the optimized baseline plan.

[0038] like Figure 1 As shown, the intelligent traffic signal control method further includes:

[0039] S101: Based on the collected real-time traffic data of multiple groups of different types at intersections, a multi-dimensional parameter set for the current direction is obtained through multi-dimensional parameter combination. A smoothed parameter sequence is obtained through time series analysis. The traffic distribution characteristics are determined by calculating the node traffic values in each direction to achieve data preprocessing.

[0040] The step S101 further includes:

[0041] S1011: Based on the collected real-time current direction traffic, current direction vehicle speed, and current direction length, a multi-dimensional parameter set of the current direction including a timestamp is obtained through multi-dimensional parameter aggregation;

[0042] S1012: Based on the current direction flow of the multi-dimensional parameter set of the current direction, obtain the average flow of the current sliding window by performing a segmented averaging process based on the current direction flow within the sliding window range; and generate a current direction average flow sequence by combining the average flows of the current sliding window based on multiple segmented averaging processes;

[0043] Based on the current direction vehicle speed of the multi-dimensional parameter set of the current direction, the average vehicle speed of the current sliding window is obtained by performing a segmented average processing of the current direction vehicle speed within the sliding window range; based on the average vehicle speed of the current sliding window processed multiple segments, a sequence of average vehicle speeds of the current direction is generated by combining the window average vehicle speeds;

[0044] Based on the current direction length of the multi-dimensional parameter set of the current direction, the average length of the current sliding window is obtained by segmented averaging of the lengths of the current direction within the sliding window; based on the average lengths of the current sliding window processed multiple times by segmented averaging, the average lengths of the current sliding window are combined to generate a sequence of average lengths of the current direction;

[0045] S1013: Based on the average flow sequence in the current direction, obtain the average flow value and the flow distribution in the current direction through multi-segment averaging processing; based on the average flow value in the current direction, obtain the congestion level of the main flow direction of the current intersection by comparing the average flow values in the current direction;

[0046] Based on the average speed sequence of the current direction, the average speed value and the speed distribution of the current direction are obtained through multi-segment averaging. Based on the average speed value of the current direction, the congestion level of the main speed direction of the current intersection is obtained by comparing the average speed values of the current direction.

[0047] Based on the average length sequence of the current direction, the average length value and the distribution of the length in the current direction are obtained through multi-segment averaging. Based on the average length value of the current direction, the congestion level of the main length direction of the current intersection is obtained by comparing the average length values of the current directions.

[0048] Based on the current direction flow distribution, current direction speed distribution, and current direction queue distribution, the accuracy of the current intersection parameter distribution and the congestion degree of the main direction is determined through a multi-parameter verification method of the congestion degree of the current main flow direction of the intersection, the congestion degree of the current main speed direction of the intersection, and the congestion degree of the current main queue direction;

[0049] Based on the parameter distribution of the current intersection and the parameter distribution of historical intersections in the same time period, the credibility of the parameter distribution and congestion level of the current intersection is determined through a multi-method verification method based on the congestion level of the main direction of the current intersection and the congestion level of the main direction of the historical intersection in the same time period;

[0050] Based on the parameter distribution of the current intersection and the parameter distribution of the historical intersection in different time periods, the regularity of the parameter distribution and the congestion degree of the current intersection is determined by verifying the regularity of the congestion degree of the main direction of the current intersection and the congestion degree of the main direction of the historical intersection in different time periods.

[0051] The step S102 further includes:

[0052] S1021: Based on the parameter distribution of the current intersection and the parameter distribution of historical intersections in the current time period, and according to the regularity of historical data and the collected real-time intersection traffic data, the traffic density in the direction of the current intersection in the next window is predicted using a deep learning prediction model;

[0053] S1022: Based on the traffic flow density in the current direction of the current segment window, the peaks and valleys of the traffic flow density in the current direction of each segment window are extracted to determine a predicted fluctuation range of the current window as a traffic flow density trough to a traffic flow density peak;

[0054] S1023: Based on the current direction traffic density of the current segment window, a comparison is made between the predicted fluctuation range of the current segment window and a preset window traffic density threshold range to determine whether the range is exceeded. If the range is exceeded, a correction value of the current direction traffic density is obtained based on the current direction traffic density of the current segment window by multi-segment averaging of the current segment window and other segment windows.

[0055] S1024: Based on the traffic propagation coefficient in the current direction, the traffic propagation coefficient in the other direction is obtained through the linear relationship between the traffic density and the traffic propagation coefficient.

[0056] The step S103 further includes:

[0057] S1031: If the traffic propagation coefficient in the current direction exceeds the traffic aggregation threshold, determine the traffic change state in the current direction based on the traffic propagation coefficient in the current direction, according to the change in vehicle speed in the current direction and the change in the length of the vehicle in the current direction;

[0058] S1032: Based on the traffic change state in the current direction, determine the saturation factor of the current direction using the linear relationship between the flow propagation coefficient in the current direction and the saturation factor in the current direction; determine whether the saturation factor in the current direction has reached a saturation suppression threshold; if so, determine that the preliminary saturation state in the current direction is a critical state close to full load;

[0059] S1033: If the initial saturation state in the current direction is a critical state close to full load, then based on the average vehicle speed sequence in the current direction, the fluctuation range of the average vehicle speed in the current direction is determined by extracting the peaks and valleys of the average vehicle speed sequence in the current direction; based on the fluctuation range of the average vehicle speed in the current direction, the initial side delay time is determined according to the increasing and decreasing relationship between the fluctuation range of the average vehicle speed in the current direction and the initial side delay time;

[0060] S1034: Based on the initial side delay time, a linear regression adjustment is performed between the initial side delay time and the length of the current direction length to obtain a corrected initial side delay time;

[0061] S1035: If the corrected initial side delay time exceeds the delay threshold, determine the local congestion state of the current direction based on the corrected initial side delay time and the current direction saturation factor;

[0062] S1036: Based on the local congestion status of the current direction, the direction of intersection adjustment is determined by the vector machine algorithm; based on the direction of intersection adjustment, according to the flow rate in the adjustment direction, the traffic change status in the adjustment direction, and the initial edge delay time of the adjustment direction, a dynamic flow distribution plan for the adjustment direction is obtained.

[0063] The step S104 further includes:

[0064] S1041: Based on the corrected edge delay time in each direction of the intersection, weights are assigned according to the traffic interaction frequency and the distance between intersections, and a graph network algorithm is used to establish a weight matrix for each direction of the intersection. Based on the weight matrix for each direction of the intersection, a flow balancing factor distribution for each direction is generated according to the linear relationship between the weight of each direction of the intersection and the flow balancing factor for each direction of the intersection.

[0065] S1042: Based on the distribution of flow balancing factors in each direction and the correlation of directions with similar weights, the flow correlation in each direction is determined; based on the intersection direction of the flow correlation in each direction, different flow concentration areas are generated through K-means clustering;

[0066] S1043: Based on each traffic concentration area, average the corrected edge delay times of each traffic concentration area to obtain an average delay value of each traffic concentration area;

[0067] S1044: If the average delay value of the current traffic concentration area exceeds the delay threshold, the fluctuation range of the current direction traffic balancing factor is determined based on the current direction traffic balancing factor sequence by extracting the peaks and valleys of the current direction traffic balancing factor sequence; if the fluctuation range of the current direction traffic balancing factor exceeds the factor fluctuation threshold, the load capacity of the current direction is determined to be high; otherwise, the load capacity of the current direction is determined to be low.

[0068] S1045: If the current direction load capacity is high load, based on the current direction flow balancing factor sequence, the current direction flow balancing factor sequence and the weight matrix are linearly regressed to obtain an optimized current direction flow balancing factor.

[0069] The step S105 further includes:

[0070] S1051: Based on the historical intersection traffic data and the real-time intersection traffic data, according to the optimized linear relationship between the current direction traffic flow balancing factor and the current direction traffic flow density adjustment coefficient, a current direction traffic flow density adjustment coefficient is obtained; based on the current direction traffic flow density adjustment coefficient, according to the linear relationship between the current direction traffic flow density adjustment coefficient and the current direction green light duration, a preliminary current direction green light duration is obtained;

[0071] S1052: Based on the traffic change state in the current direction, according to the offset relationship between the traffic density in the current direction and the preliminary green light duration in the current direction, determine the optimized green light duration in the current direction;

[0072] S1053: Based on the optimized green light duration in the current direction, an updated current direction traffic flow balancing factor is obtained by linear regression adjustment of the optimized green light duration in the current direction and the current direction traffic flow balancing factor; based on the updated current direction traffic flow balancing factor, an optimized current direction traffic flow density is obtained by linear regression adjustment of the intersection distance and the current direction vehicle speed and the current direction traffic flow density adjustment coefficient;

[0073] S1054: Based on the traffic density sequence in the current direction, the fluctuation range of the traffic density in the current direction is determined by extracting the peaks and valleys of the traffic density sequence in the current direction; if the fluctuation range of the traffic density in the current direction is greater than the traffic density fluctuation threshold, the traffic density area in the current direction is determined to be a high traffic density area;

[0074] S1055: If the traffic density area in the current direction is a high traffic density area, then based on the current direction traffic flow balancing factor, according to the optimized linear relationship between the traffic density in the current direction and the traffic flow balancing factor, an updated current direction traffic flow balancing factor is obtained; based on the updated current direction traffic flow balancing factor, according to the updated linear relationship between the traffic flow balancing factor in the current direction and the green light duration in the current direction, an optimized green light duration in the current direction is obtained; determine whether the optimized green light duration in the current direction exceeds a green light duration threshold; if so, adjust the current direction green light duration based on the linear relationship between the distance between intersections and the current direction vehicle speed and the green light duration in the current direction to obtain an adjusted green light duration in the current direction;

[0075] S1056: Based on the adjusted green light duration for the current direction, a final green light duration for the current direction is determined by a linear relationship between the adjusted green light duration for the current direction and the traffic density for the current direction.

[0076] The step S106 further includes:

[0077] S1061: Based on the final green light duration in the current direction and using real-time incremental traffic feedback in the current direction, determine whether the final green light duration in the current direction is lower than the minimum baseline for the green light duration. If so, obtain a green light duration adjustment plan.

[0078] S1062: Based on the green light duration adjustment scheme, extract the light position switching characteristics of the green light duration adjustment scheme to obtain the green light switching interval for the current direction. Based on the green light switching interval for the current direction, determine whether there are still vehicles stranded when the green light for the current direction ends. If there are still vehicles stranded, preliminarily adjust the green light switching interval for the current direction.

[0079] S1063: Based on the preliminary adjustment of the green light switching interval in the current direction, a time window boundary for the peak period is set, and the smoothing rate fluctuation range is obtained by extracting the peaks and valleys of the average traffic sequence in the current direction based on the time window boundary;

[0080] S1064: Based on the smoothing rate fluctuation range, if the smoothing rate fluctuation range exceeds the smoothing rate threshold, a green light interval offset is obtained based on the vehicle speed increment in the current direction and a linear relationship between the vehicle speed increment in the current direction and the green light duration increment;

[0081] S1065: Based on the green light switching interval of the current direction, the green light switching interval of the current direction is compensated by the green light interval offset to obtain a compensated green light switching interval of the current direction; based on the compensated green light switching interval of the current direction, the time window boundary of the peak period is set again, and the peak and valley of the average traffic sequence of the current direction based on the time window boundary are extracted again to obtain the verified smoothing rate fluctuation range;

[0082] S1066: Based on the verified smoothing rate fluctuation range, determine whether the verified smoothing rate fluctuation range meets the smoothing rate standard. If it does not meet the smoothing rate standard, adjust the green light switching interval for the current direction to generate a final green light switching interval for the current direction.

[0083] The step S107 further includes:

[0084] S1071: Based on the smoothing rate fluctuation range, the initial signal light parameter adjustment amount is obtained by extracting the current direction green light duration and the current direction green light switching interval;

[0085] S1072: Based on the current direction traffic flow, the current direction vehicle speed, and the correlation between the traffic flows in each direction, determine whether there is a vehicle backlog in the current direction. If there is a vehicle backlog, determine the scope of the neighborhood influence radius as the degree of vehicle overflow in the current direction.

[0086] S1073: Based on the current direction traffic flow, the current direction vehicle speed, and the current direction traffic overflow degree, the current direction intersection state value is quantized in the range of 0-1 using a support vector machine algorithm;

[0087] S1074: Based on the current direction intersection state value, if the current direction intersection state value is lower than the intersection state threshold, an intersection state error value between the current direction intersection state value and the intersection state threshold is obtained, and based on the intersection state error value, the current direction green light duration and the current direction green light switching interval are dynamically adjusted according to a linear relationship between the current direction green light duration and the current direction green light switching interval and the intersection state error value;

[0088] S1075: Based on the dynamically adjusted current direction green light duration and current direction green light switching interval, and according to the overflow degree of traffic in the current direction, the data collection frequency is adjusted to obtain optimized current direction traffic flow and current direction vehicle speed.

[0089] The step S108 further includes:

[0090] S1081: Based on the predicted traffic flow density at the intersection for all time periods, a change trend of the traffic flow density in the current direction of the current time period is obtained by comparing the predicted traffic flow density in the current direction of the current time period with the predicted traffic flow density in the current direction during the peak period; and a predicted deviation of the traffic flow density in the current direction is obtained by comparing the predicted traffic flow density in the current direction with the actual traffic flow density in the current direction.

[0091] S1082: Based on the predicted deviation of the traffic flow density in the current direction and the linear relationship between the traffic flow density change trend in the current direction and the traffic propagation coefficient, determine the traffic propagation coefficient in the current direction;

[0092] S1083: Based on the prediction deviation of the traffic flow density in the current direction, the optimized traffic propagation coefficient in the current direction is determined by correcting the coefficient of the traffic propagation coefficient in the current direction and the coefficient of the prediction deviation;

[0093] S1084: Based on the optimized traffic propagation coefficient of the current direction and a linear relationship between the optimized traffic propagation coefficient of the current direction and the green light duration of the current direction, a green light duration of the current direction is determined, and at the same time, a temporary baseline plan for the current direction is generated;

[0094] S1085: Based on the temporary baseline plan for the current direction, determine whether it meets the dynamic update requirements by comparing it with the actual traffic density feedback for the current direction;

[0095] S1086: If the dynamic update requirements are met, the temporary baseline solution for the current direction is converted into the official baseline solution for the current direction. Through repeated iterative prediction, the official baseline solution for the next direction is generated.

[0096] S1087: Based on the predicted traffic density in the current direction and the traffic propagation coefficient in the current direction, the green light duration in the other direction is predicted through repeated iterations to form the final baseline solution for the current intersection.

[0097] like Figure 2 As shown, at the same time, the present invention also provides an intelligent traffic signal control system based on the Internet of Things, including:

[0098] Based on the above intelligent traffic signal control method, the intelligent traffic signal control system includes:

[0099] Module 101: Based on the collected real-time traffic data of multiple groups of different types at intersections, a multi-dimensional parameter set for the current direction is obtained through multi-dimensional parameter combination, a smoothed parameter sequence is obtained through time series analysis, and traffic distribution characteristics are determined by calculating the node traffic values in each direction to achieve data preprocessing;

[0100] Module 102: Calculates the traffic density change trend based on node traffic and historical data using a deep learning prediction model, extracts key features from the traffic density change prediction value, adjusts the current traffic density through data comparison, and calculates the traffic propagation coefficient based on the corrected traffic density sequence to obtain the traffic propagation coefficient;

[0101] Module 103: Based on the traffic propagation coefficient, data fusion analysis is used to determine the range of traffic state changes. Analysis and judgment of the intersection state are used to obtain a preliminary saturation determination result. Initial quantitative values of the edge delay time are obtained by extracting fluctuation characteristics. The quantitative values are adjusted using a linear regression algorithm to obtain a corrected edge delay time. Based on the intersection state, the presence of local congestion points is determined. A support vector machine algorithm is used to determine the direction of flow adjustment to achieve dynamic flow distribution.

[0102] Module 104: Calculates an adjacency weight matrix based on the quantized edge delay time using a graph network algorithm, divides the traffic concentration areas within the area boundary using a clustering algorithm, calculates the average delay value within each area boundary, obtains traffic fluctuation characteristics between intersections from the factor distribution, and adjusts the traffic balancing factor using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction;

[0103] Module 105: Calculates the traffic density adjustment coefficient using a collaborative optimization algorithm based on the distribution of traffic balancing factors and the trend of traffic density changes. Calculates the baseline adjustment offset based on the trend data. Fusions the adjustment coefficient using a linear regression algorithm. Extracts fluctuation characteristics from traffic density. Determines whether to adjust the green light duration distribution based on a preset threshold and comparison with the duration scheme. Calculates the traffic density adjustment result after collaborative optimization by adjusting the green light duration distribution to obtain the final green light duration for the current direction.

[0104] Module 106: Based on the gain data obtained from real-time feedback, it determines whether the green light duration is lower than the minimum duration baseline to obtain the verification process result, determines the time window boundary by adjusting the interval range, calculates the smoothing rate distribution based on the time window boundary, obtains the adjusted interval offset by adjusting the baseline through gain data fusion, updates the light position switching characteristics based on the adjusted interval offset, and compares the smoothing rate change trend with the minimum baseline to obtain the final green light switching interval for the current direction;

[0105] Module 107: used to determine the smoothing rate allocation range based on the smoothing rate fluctuation range and the time window boundary, classify the real-time traffic data and vehicle speed change values by judging the scope of the intersection relevance on the neighborhood influence radius, update the traffic light parameters by parameter adjustment, and obtain the optimized sensor data stream by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval;

[0106] Module 108: It is used to obtain the initial analysis results based on the traffic density of the intersection predicted for the entire period by extracting the change trend and deviation data, calculate the traffic propagation coefficient in combination with the dynamic update frequency, adjust the traffic propagation model through the deviation data, update the green light duration calculation logic through the corrected coefficient, generate the baseline plan through the adjusted duration value, optimize the plan in combination with the traffic density change trend, and update the green light duration allocation logic according to the final adjustment result to output the optimized baseline plan.

[0107] Example 1:

[0108] The following combination Figure 1 The working principle of the intelligent traffic signal control method shown in the embodiment is described.

[0109] like Figure 1As shown, the intelligent traffic signal control method includes seven major steps, including the second step (step S102): obtaining the flow propagation coefficient; the third step (step S103): realizing dynamic flow distribution; the fourth step (step S104): obtaining the optimized current direction flow balancing factor; the fifth step (step S105): obtaining the final current direction green light duration; the sixth step (step S106): obtaining the final current direction green light switching interval; the seventh step (step S107): obtaining the green light duration and the green light switching interval collaborative optimization; the eighth step (step S108): outputting the optimized baseline solution. Through the above eight steps, the intelligent traffic signal control method is completed; wherein, the second step: according to the initial node flow capacity Based on the initial value and historical data, the deep learning prediction model is used to calculate the trend of traffic density changes in each direction, and the prediction results are compared with the current traffic density to determine the accurate distribution of the flow propagation coefficient; the third step: if the flow propagation coefficient exceeds the preset flow aggregation threshold, the queue length data and vehicle speed change data are fused and analyzed, combined with the path saturation factor, to determine whether the traffic state of the current intersection has reached the saturation suppression threshold, and obtain the quantitative value of the edge delay time; the fourth step: through the quantitative value of the edge delay time, combined with the data collection results of adjacent intersections, the graph network algorithm is used to calculate the adjacency weight matrix, and the flow balance factor distribution between each intersection within the regional boundary is determined; the fifth step: based on the flow balance factor distribution and the trend of traffic density changes, the coordination factor is used to calculate the flow balance factor distribution between each intersection within the regional boundary. Calculate the density adjustment coefficient with the same optimization algorithm to obtain a preliminary green light duration baseline adjustment plan; Step 6: If the green light time in a certain direction in the preliminary green light duration baseline adjustment plan is lower than the preset minimum green light duration baseline, verify it through real-time feedback gain data, adjust the phase switching interval, and obtain the optimized time window smoothing rate allocation result; Step 7: Dynamically adjust the signal light control parameters of adjacent intersections through the optimized time window smoothing rate allocation result, obtain the adjusted real-time traffic flow and speed change data from the road sensor, and judge the degree of improvement of the regional traffic status within the neighborhood influence radius; Step 8: When the degree of improvement is lower than expected, extract the deviation data of the traffic density change trend from the prediction model, and recalculate the traffic flow in combination with the dynamic update frequency. The traffic propagation coefficient is calculated to obtain an updated green light duration baseline adjustment plan; after the above steps, the present invention collects traffic flow, vehicle speed and queue length data in real time through road sensors, and uses time series analysis and deep learning prediction models to calculate the traffic density change trend; according to the traffic propagation coefficient, edge delay time and traffic balancing factor, the present invention adopts a collaborative optimization algorithm to dynamically adjust the green light duration, and optimizes the phase switching interval through real-time feedback gain data; when the degree of improvement in traffic conditions is lower than expected, the present invention can extract deviation data from the prediction model, recalculate the traffic propagation coefficient, and realize adaptive adjustment of the green light duration; this method can effectively alleviate urban traffic congestion, improve road traffic efficiency, and provide an innovative solution for smart city traffic management.

[0110] This step implements data preprocessing; the method can be summarized as follows: IoT devices acquire real-time traffic flow, speed changes, and queue length data from road sensors. Time series analysis is then used to preprocess these multi-dimensional parameters to obtain initial values for node traffic capacity in each direction. This collection method, leveraging the low latency and high-frequency sampling characteristics of IoT devices, accurately reflects road conditions and provides a reliable foundation for subsequent analysis. The beneficial effect of this approach is that real-time data provides a dynamic basis for traffic management, avoiding reliance on outdated, stereotyped statistics.

[0111] This step implements data preprocessing and includes three main steps: the first main step (step S1011) collects real-time traffic flow, vehicle speed change, and queue length data from road sensors through IoT devices to obtain a multidimensional parameter set; the second main step (step S1012) preprocesses the multidimensional parameter set using a time series analysis method to obtain a smoothed parameter sequence; the third main step (step S1013) calculates the node flow values in each direction based on the smoothed parameter sequence to determine the flow distribution characteristics; wherein:

[0112] Specifically, by collecting real-time traffic flow, vehicle speed change, and queue length data from road sensors through IoT devices, a multi-dimensional parameter set can be obtained;

[0113] For example, sensors deployed at intersections on urban arterial roads record the number of vehicles passing, their average speed, and the number of vehicles in queues every second, forming a multidimensional dataset with timestamps. Real-time traffic flow might show 200 vehicles passing per minute during peak hours, with speeds decreasing from 40 km / h to 20 km / h and queues increasing from 5 to 15 vehicles. This data collection method, relying on the low latency and high-frequency sampling characteristics of IoT devices, can truly reflect road conditions and provide a reliable foundation for subsequent analysis. The beneficial effect of this is that real-time data provides a dynamic basis for traffic management, avoiding reliance on outdated, stereotyped statistics.

[0114] In one possible implementation, a time series analysis method is used to preprocess the multidimensional parameter set to obtain a smoothed parameter sequence;

[0115] Specifically, the raw data can be processed using the moving average method;

[0116] For example, a 5-minute sliding window can be set to average vehicle speeds to eliminate short-term fluctuations. For example, the speeds within a 5-minute period can be smoothed from 40, 38, 20, 22, and 25 km / h to a trend value of approximately 29 km / h. The core principle of this method is to highlight trends by reducing noise, facilitating subsequent analysis of traffic distribution characteristics.

[0117] It should be noted that smoothing can also effectively reduce the interference of occasional sensor errors, improve data quality, and help to more accurately capture the long-term changes in traffic flow;

[0118] For the smoothed parameter sequence, calculating the node flow values in each direction and determining the flow distribution characteristics is a key step;

[0119] In one embodiment, assuming that the intersection has four directions: east, west, south, and north, hourly flow values are calculated based on smoothed data;

[0120] For example, the eastbound traffic volume is 1,200 vehicles / hour, the westbound traffic volume is 800 vehicles / hour, the southbound traffic volume is 600 vehicles / hour, and the northbound traffic volume is 1,000 vehicles / hour;

[0121] It is understandable that by comparing these values, it can be found that east and north are the main traffic directions, accounting for about 60%;

[0122] Preferably, further verification can be performed by combining queue length data. For example, a longer eastbound queue indicates concentrated traffic pressure. This multi-dimensional, mutually supportive analysis method can not only reveal the spatial characteristics of traffic distribution, but also provide a basis for traffic light optimization.

[0123] For example, consider the calculation of node flow value from multiple aspects;

[0124] In one embodiment, in addition to directly counting the number of passing vehicles, the congestion level can also be inferred based on changes in vehicle speed. For example, if the speed is less than 15 km / h, the flow rate value may be adjusted to 80% of the actual capacity to reflect the road bottleneck effect.

[0125] Specifically, if westbound vehicle speeds remain low and queue lengths increase, the traffic flow value might be adjusted from 800 vehicles per hour to 640 vehicles per hour. This approach has the advantage of not relying solely on a single indicator but also improving accuracy through multi-parameter verification, thereby providing more scientific support for traffic forecasting and diversion.

[0126] For example, when determining traffic distribution characteristics, we can further analyze the temporal variation patterns. For example, suppose that during the morning rush hour from 7:00 to 9:00, eastbound traffic increases from 800 vehicles / hour to 1500 vehicles / hour, while northbound traffic remains stable. This difference suggests temporal heterogeneity in traffic flow, which may be related to tidal flows during rush hour.

[0127] Preferably, historical data comparison is combined, e.g., if the traffic distribution consistency during the same time period in the previous week reaches 90%, this can enhance the credibility of the feature. The technical benefit of this analysis is that it can provide data support for the dynamic adjustment of traffic resources (such as increasing the eastbound green light time) and improve road utilization.

[0128] The entire process, from data collection to feature determination, forms a complete logical chain. Sensors provide raw input, time series analysis optimizes data quality, and node flow calculation and distribution feature analysis directly serve traffic management decisions. Each step focuses on optimizing road flow, with example data and methods mutually reinforcing to ensure the solution is practical and scalable. This multi-faceted approach not only ensures the rigor of the core solution, but also enriches application scenarios through optional analysis, significantly improving the intelligent level of the transportation system.

[0129] This step obtains the traffic propagation coefficient, which includes four main steps. The first main step (step S1021) uses a deep learning prediction model to calculate the traffic density change trend through historical data and node traffic to obtain the predicted value in each direction. The second main step (step S1022) extracts key features from the predicted value based on the change trend and determines the density fluctuation range within the time window. The third main step (step S1023) adjusts the current density through data comparison to obtain a corrected density sequence if the density fluctuation range exceeds the preset threshold. The fourth main step (step S1024) calculates the traffic propagation coefficient based on the corrected density sequence to determine the propagation distribution in each direction. Among them:

[0130] Specifically, by using historical data and node traffic, and using deep learning prediction models to calculate the changing trend of traffic density, forward-looking support can be provided for traffic management.

[0131] For example, deep learning models, such as long short-term memory networks (LSTMs), can capture long-term dependencies in time series. For example, given historical data on urban intersections, including hourly traffic density over the past 30 days, the model is trained to predict density changes over the next hour.

[0132] For example, the eastbound traffic density may increase from 20 vehicles / km to 35 vehicles / km, while the westbound traffic density remains at 15 vehicles / km. This prediction relies on the regularity of historical data and real-time input from nodes to highlight dynamic trends.

[0133] The core of the analysis is to extract key features from the predicted values and determine the density fluctuation range within the time window.

[0134] In a possible implementation, the time window is set to 15 minutes, and features such as density peaks and valleys are extracted.

[0135] For example, the eastbound predicted density fluctuates from 25 vehicles / km to 40 vehicles / km within 15 minutes, with a fluctuation range of 15 vehicles / km. If the preset threshold is 10 vehicles / km, it means that the fluctuation exceeds the limit.

[0136] It should be noted that feature extraction can reveal the severity of density changes, providing a basis for subsequent adjustments. If the density fluctuation range exceeds the preset threshold, the current density is adjusted through data comparison to obtain a corrected density sequence.

[0137] Specifically, by comparing real-time sensor data with predicted values, for example, if the predicted density in the eastbound direction is 35 vehicles / km but the sensor shows 30 vehicles / km, the predicted value can be corrected to 32 vehicles / km, forming a smoothed sequence. This method uses real-time data to verify prediction bias and ensure that the density sequence is closer to reality. Calculating the traffic propagation coefficient based on the corrected density sequence and determining the propagation distribution in each direction is an extension of this analysis.

[0138] In one embodiment, the propagation coefficient reflects the intensity of the traffic flow spreading from a certain direction to other directions.

[0139] For example, the corrected density in the east direction is 32 vehicles / km, and in the north direction it is 25 vehicles / km. The calculated propagation coefficient is 0.6 in the east direction and 0.4 in the north direction, indicating that the eastward traffic flow has a greater impact on the surrounding area.

[0140] Preferably, the queue length verification is combined with the increase in queues in the east direction and the rationality of the support coefficient.

[0141] It is understandable that the application of the prediction model starts from historical data and gradually refines to the propagation distribution to form a complete chain.

[0142] For example, the eastward density shows a clear upward trend during the morning peak, and the propagation coefficient is high, indicating that the flow pressure is diffusing.

[0143] In one embodiment, historical data can also reveal patterns in density fluctuations during holidays, such as the eastbound density dropping to 20 vehicles / km on weekends, which reduces the coefficient. This multi-faceted analysis ensures a comprehensive and practical analysis.

[0144] For example, for westbound traffic, the predicted density remains stable at 15 vehicles / km, fluctuating by only 5 vehicles / km, not exceeding the threshold and requiring no correction, resulting in a propagation coefficient of 0.3. Combined with speed data, such as a westbound speed of 40 km / h, this supports the conclusion of low propagation. This multi-parameter approach enhances prediction reliability and supports traffic management.

[0145] This step implements dynamic traffic allocation and includes six major steps. The first major step (step S1031) determines the range of traffic state changes by fusing queue length data with vehicle speed change data through data fusion analysis. The second major step (step S1032) determines whether the intersection state has reached the saturation suppression threshold based on the range of traffic state changes and the path saturation factor, thereby obtaining a preliminary saturation determination result. The third major step (step S1033) extracts fluctuation characteristics from the vehicle speed change data based on the preliminary saturation determination result to obtain an initial quantized value of the side delay time. The fourth major step (step S1034) uses the initial quantized value of the side delay time, combined with the queue length data, to adjust the quantized value using a linear regression algorithm to obtain a corrected side delay time. The fifth major step (step S1035) determines whether the corrected side delay time exceeds the preset delay threshold by obtaining the distribution characteristics of the path saturation factor based on the intersection state to determine whether there is a local congestion point. The sixth main step (step S1036) is to determine the direction of flow adjustment based on the judgment result of the local congestion point, integrate the traffic status and edge delay time, and obtain a dynamic flow distribution plan through the support vector machine algorithm.

[0146] Specifically, when the traffic propagation coefficient exceeds the traffic aggregation threshold, the range of traffic status changes can be analyzed by fusing queue length data with vehicle speed change data.

[0147] For example, suppose the eastbound traffic flow propagation coefficient at a city intersection reaches 0.8, exceeding the threshold of 0.5. At this point, the queue length increases from 50 meters to 80 meters, and the speed decreases from 30 km / h to 20 km / h. By integrating these two sets of data, it can be determined that the traffic state has changed from stable to congested, and the range of change is reflected in a significant increase in traffic density.

[0148] In one possible implementation, the path saturation factor is combined with traffic flow to determine whether the intersection has reached the saturation suppression threshold. The path saturation factor reflects the traffic carrying capacity. For example, if the eastbound saturation factor increases from 0.7 to 0.95, approaching the saturation suppression threshold of 1.0, it indicates that the intersection is nearing capacity. Preliminary results indicate that the intersection may be in a critical state.

[0149] It should be noted that this judgment relies on the real-time nature of the data and can provide a basis for subsequent analysis. Extracting fluctuation characteristics from vehicle speed change data and obtaining the initial quantitative value of the edge delay time is a key step.

[0150] Specifically, if the vehicle speed drops from 30 km / h to 20 km / h, with a fluctuation of 10 km / h, the initial side delay can be estimated to increase waiting time by 5 seconds per vehicle. Combined with queue length data, if the queue length increases to 80 meters, the corrected side delay might become 6 seconds after linear regression adjustment. This adjustment more accurately reflects the actual delay. If the corrected side delay exceeds the preset delay threshold, for example, 4 seconds, further analysis of the intersection status is required.

[0151] In one embodiment, the eastbound edge delay is 6 seconds. Combined with the saturation factor of 0.95, the distribution characteristics of the path saturation factor show that the pressure is concentrated in the east direction, and there may be a local congestion point.

[0152] Understandably, this distribution reveals the imbalance of traffic flow diffusion and provides a direction for optimization. Based on the results of the local congestion point judgment, the traffic status and edge delay time are integrated, and the support vector machine algorithm is used to determine the direction of traffic flow adjustment.

[0153] Preferably, if the eastbound traffic is significantly congested, the support vector machine can analyze that part of the traffic needs to be diverted to the northbound. The dynamic allocation solution may be to reduce the traffic flow in the eastbound direction by 20% and increase it in the northbound direction by 15%.

[0154] For example, the current saturation factor in the north direction is 0.6, and the vehicle speed is 35 km / h, which has the capacity to accept traffic and supports the rationality of diversion.

[0155] In one example, assume that the westbound traffic flow has a propagation coefficient of 0.3, which does not exceed the threshold, queue lengths are stable at 30 meters, and speeds are 40 km / h. Traffic conditions are stable and no adjustments are required. However, if the speed of westbound traffic drops to 30 km / h after eastbound diversion, the traffic flow status needs to be reassessed. This multi-directional verification ensures a comprehensive solution.

[0156] For example, during the morning rush hour, the eastbound delay increased to 8 seconds, while the northbound delay was reduced to 5 seconds after diversion, proving that the adjustment was effective, relieving pressure and improving traffic efficiency.

[0157] Specifically, when integrating queue length and vehicle speed change data, priority can be given to peak hour patterns.

[0158] For example, during Monday morning rush hour, eastbound queue lengths increased significantly, and vehicle speeds decreased more steeply. This analysis allowed for the early prediction of congestion points. After careful consideration, it was determined that if eastbound traffic decreased during holidays, the propagation coefficient dropped to 0.4, and the delay was only 3 seconds, then no diversion would be necessary, demonstrating the flexibility of the solution. This approach optimizes resource allocation and reduces congestion risks.

[0159] This step obtains the optimized current direction flow balancing factor, which includes five major steps, namely, the first major step (step S1041) fuses the data collected from adjacent intersections through the quantized value of the edge delay time, uses the graph network algorithm to calculate the adjacency weight matrix, and obtains the flow balancing factor distribution. The second major step (step S1042) extracts the correlation strength between intersections from the flow balancing factor distribution, uses the clustering algorithm to divide the flow aggregation area within the regional boundary, and obtains the aggregation area division result. The third major step (step S1043) fuses the edge delay time of adjacent intersections according to the aggregation area division result, calculates the average delay value within each regional boundary, and determines the delay distribution characteristics. The fourth major step (step S1044) If the average delay value exceeds the preset delay threshold, the flow fluctuation characteristics between intersections are obtained from the factor distribution to determine whether there is a high-load intersection. The fifth major step (step S1045) fuses the weight matrix data through the judgment result of the high-load intersection, uses the linear regression algorithm to adjust the flow balancing factor, and obtains the optimized factor distribution. Among them:

[0160] Specifically, by fusing data collected from adjacent intersections with the quantized values of edge delay time, the graph network algorithm can effectively capture the spatial correlation between intersections.

[0161] For example, consider five adjacent intersections in a city's core area. The eastbound intersection delay is 6 seconds, the southbound intersection delay is 4 seconds, and the westbound intersection delay is 3 seconds. When using a graph network algorithm to calculate the adjacency weight matrix, weights can be assigned based on the frequency and distance of traffic flow interactions between intersections. For example, a weight of 0.7 for eastbound and southbound intersections and 0.4 for westbound and southbound intersections can be assigned to reflect the strength of traffic flow connections. Once the weight matrix is generated, the distribution of traffic balancing factors reveals the distribution of traffic within the area.

[0162] Specifically, the factor is 0.85 in the east direction, 0.65 in the south direction, and 0.5 in the west direction, indicating that the flow pressure in the east direction is greater.

[0163] In one possible implementation, after extracting correlation strength from the distribution of traffic balancing factors, a clustering algorithm is used to divide traffic concentration areas. Assume that, through K-means clustering, eastbound and southbound traffic are classified as high-flow concentration areas, while westbound traffic is separately classified as a low-flow area. The clustering results show that eastbound and southbound traffic is dense, posing a potential congestion risk. When the edge delays of adjacent intersections are incorporated to calculate the average delay value for the clustered areas, the average delay is 5 seconds in the high-flow area and 3 seconds in the low-flow area.

[0164] It should be noted that the delay distribution characteristics reflect the concentration of pressure in high-traffic areas, which may be caused by the influx of traffic during the morning rush hour. If the average delay value exceeds the preset threshold, for example, set to 4 seconds, then 5 seconds in the high-traffic area indicates that further analysis is needed.

[0165] It is understandable that when extracting traffic fluctuation characteristics from the factor distribution, the eastbound factor fluctuates from 0.85 to 0.9, and the delay increases to 7 seconds, indicating a high-load intersection. The southbound factor remains stable at 0.65, with less fluctuation and normal load.

[0166] Preferably, after fusing the weight matrix data, a linear regression algorithm adjusts the flow balancing factor.

[0167] In one embodiment, the eastward factor is lowered to 0.75 and the southward factor is increased to 0.7, and the optimized distribution is more balanced.

[0168] Specifically, after the high-load intersection is determined, the adjustment plan can be verified in multiple directions.

[0169] For example, by diverting 10% of eastbound traffic to the southbound direction, the southbound speed dropped slightly from 30 km / h to 28 km / h, maintaining sufficient capacity. Westbound traffic remained unchanged, as it was delayed by only 3 seconds and remained stable.

[0170] In one embodiment, the eastbound delay during the morning rush hour rises to 8 seconds and drops to 5 seconds after diversion. The regional average delay tends to be below the threshold, and traffic efficiency is improved.

[0171] For example, if eastbound traffic drops to 0.6 during holidays and the delay is reduced to 3 seconds, no adjustment is required, demonstrating the flexibility of the solution. This approach ensures reasonable regional traffic distribution and alleviates local pressure through correlation analysis and dynamic adjustment.

[0172] This step obtains the final green light duration in the current direction, which includes six major steps. The first major step (step S1051) obtains traffic density data through the traffic balance factor distribution, calculates the density adjustment coefficient using a collaborative optimization algorithm, and obtains a preliminary green light duration adjustment plan. The second major step (step S1052) extracts the change trend characteristics from the preliminary green light duration adjustment plan, calculates the baseline adjustment offset based on the change trend data, and determines the adjusted duration distribution. The third major step (step S1053) obtains the factor distribution update value based on the adjusted duration distribution, and uses a linear regression algorithm to fuse the adjustment coefficient to obtain the optimized solution distribution. The fourth major step (step S1054) determines whether there is an overloaded area through the optimized solution distribution. If an overloaded area exists, the fluctuation characteristics are extracted from the traffic density to determine the range of the high-density area. The fifth major step (step S1055) obtains the traffic balance update value based on the high-density area range, uses a preset threshold to compare the duration plan, and determines whether to adjust the green light duration distribution. The sixth main step (step S1056) is to calculate the density adjustment result after collaborative optimization by adjusting the green light duration distribution to obtain the final duration plan distribution.

[0173] Specifically, obtaining traffic density data through the distribution of flow balancing factors is an important starting point for optimizing traffic management.

[0174] For example, consider three intersections within a city with factors of 0.8 for eastbound traffic, 0.65 for southbound traffic, and 0.5 for westbound traffic, reflecting differences in traffic density. Using a collaborative optimization algorithm to calculate density adjustment factors based on historical data and real-time traffic flow, we can arrive at coefficients of 1.2 for eastbound traffic, 1.0 for southbound traffic, and 0.9 for westbound traffic, preliminarily indicating that longer green light durations are needed for eastbound traffic.

[0175] In one possible implementation, when extracting trend characteristics from the preliminary green light duration adjustment plan, it can be observed that the density in the east direction continues to increase during the peak period, remains stable in the south direction, and slightly decreases in the west direction. Based on the trend, a baseline adjustment offset is calculated, for example, adding 5 seconds to the east direction, 2 seconds to the south direction, and keeping it unchanged in the west direction, to form the adjusted duration distribution.

[0176] It should be noted that this distribution takes into account the positive relationship between density and duration to ensure that the traffic capacity matches actual demand.

[0177] Specifically, after obtaining updated factor distribution values based on the adjusted duration distribution, the eastbound factor rose to 0.85, stabilized at 0.65 for southbound traffic, and slightly decreased to 0.45 for westbound traffic. Using a linear regression algorithm to incorporate the adjustment coefficients, which incorporates variables such as intersection distance and vehicle speed, the optimized distribution shows a green light duration of 35 seconds for eastbound traffic, 28 seconds for southbound traffic, and 20 seconds for westbound traffic.

[0178] Optimally, this solution balances traffic pressure within the region. When identifying overloaded areas through the optimized distribution of the solution, for example, if the eastbound duration is extended but the density still exceeds the standard, it indicates an overload risk. Extracting fluctuation characteristics from traffic density reveals that the eastbound density increases from 2,000 to 2,500 vehicles per hour during the morning rush hour, confirming this as a high-density area.

[0179] It is understandable that after the high-density area is combined with the flow balance update value, the eastward factor is adjusted to 0.9 and the southward factor is slightly adjusted to 0.7.

[0180] In one embodiment, a preset green light duration threshold of 30 seconds was used. After comparing the optimized solution, the eastbound traffic exceeded the threshold and required further adjustment. By adjusting the green light duration distribution, for example, increasing the eastbound traffic to 40 seconds and maintaining the southbound traffic at 28 seconds, the density adjustment results after collaborative optimization showed that the eastbound traffic density dropped to 2,200 vehicles per hour, making traffic in the area smoother.

[0181] In one embodiment, the final duration plan is distributed as 40 seconds to the east, 28 seconds to the south, and 20 seconds to the west, and the density distribution tends to be reasonable.

[0182] For example, during holidays, the eastbound density is reduced to 1,500 vehicles per hour, and the green light duration can be adjusted back to 30 seconds, reflecting the flexibility of the plan.

[0183] Specifically, when verifying traffic from multiple directions, after diverting traffic from the eastbound direction to the southbound direction, the southbound speed dropped from 35 km / h to 33 km / h, still being able to handle the additional traffic, while the westbound direction remained stable. This dynamic adjustment ensures the efficiency and practicality of the solution.

[0184] This step obtains the final green light switching interval for the current direction, which includes six major steps. The first major step (step S1061) obtains gain data through real-time feedback, determines whether the green light duration is lower than the minimum baseline, and obtains the verification process result. The second major step (step S1062) extracts the phase switching characteristics from the verification process result, adjusts the interval range, and determines the time window boundary. The third major step (step S1063) calculates the smoothing rate distribution according to the time window boundary to obtain the optimization scheme update value. The fourth major step (step S1064) If the optimization scheme update value exceeds the preset threshold, the gain data fusion baseline adjustment is performed to obtain the adjustment interval offset. The fifth major step (step S1065) updates the phase switching sequence according to the adjustment interval offset to obtain the smoothing rate change trend. The sixth major step (step S1066) compares the smoothing rate change trend with the lowest baseline to determine whether to adjust the green light duration and obtain the final allocation result. Among them:

[0185] Specifically, obtaining gain data through real-time feedback is an important step in optimizing traffic signal control.

[0186] For example, suppose that a sensor at a certain intersection monitors an increase in eastbound traffic flow in real time, and the gain data shows that the number of waiting vehicles increases from 50 to 70.

[0187] Understandably, this data reflects a potential shortfall in green light duration. When determining whether the green light duration is below the minimum baseline, assuming the baseline is 25 seconds and the current duration is 20 seconds, the verification results indicate that adjustment is necessary.

[0188] Specifically, by extracting phase switching features from the verification process results, it may be found that there are still vehicles stranded when the eastbound green light ends, and the switching interval is relatively short.

[0189] In one possible implementation, when adjusting the interval range, the original 15-second phase switching interval is extended to 20 seconds, and the time window boundary is determined to be between 8:00 and 9:00 during peak hours. When calculating the smoothing rate distribution based on the time window boundary, it can be observed that eastbound traffic fluctuates significantly within the time window, and the smoothing rate decreases from 0.7 to 0.6.

[0190] It should be noted that the smoothness rate reflects the smoothness of traffic flow. After obtaining the updated value of the optimization solution, if the updated value is 0.8 and exceeds the preset threshold of 0.75, further adjustment is required.

[0191] Preferably, the gain data fusion baseline adjustment, combined with the feedback of the real-time vehicle speed decreasing from 40 km / h to 35 km / h, results in an adjustment interval offset of 5 seconds. This offset updates the phase switching sequence, increasing the eastbound green light duration from 20 seconds to 25 seconds.

[0192] In one embodiment, the smoothing rate change trend shows an increase to 0.65 after adjustment, compared with the minimum baseline of 25 seconds, indicating that the duration has reached the target but can still be optimized.

[0193] Specifically, when determining whether to adjust the green light duration, if the eastbound waiting time during peak hours still exceeds 10 seconds, the final allocation result may increase the duration to 30 seconds.

[0194] For example, the southbound and westbound traffic flows were stable at 25 seconds and 20 seconds respectively. After the eastbound flow increased to 30 seconds, the overall traffic efficiency in the region was improved.

[0195] Understandably, this adjustment avoids the spread of congestion in a single direction.

[0196] In one possible implementation, when verified from multiple aspects, the eastbound traffic flow was reduced to 60 stranded vehicles, the southbound speed remained stable, and the westbound waiting time did not increase significantly, proving that the plan was reasonable.

[0197] In one embodiment, if traffic decreases during holidays, the eastbound duration can be adjusted back to 25 seconds to reflect flexibility.

[0198] Optimally, this dynamic adjustment ensures that signal control matches actual demand, alleviating pressure during peak hours. The entire process, from gain data to final allocation, is logically clear and consistent, effectively improving the adaptability of traffic management.

[0199] This step obtains the coordinated optimization of the green light duration and the green light switching interval, and includes five major steps, namely, the first major step (step S1071) determines the smoothing rate distribution range through the time window boundary, and obtains the initial signal light parameter adjustment amount. The second major step (step S1072) obtains real-time traffic data and vehicle speed change values from the sensor data stream, and determines the scope of influence of the intersection correlation on the neighborhood influence radius. The third major step (step S1073) uses the support vector machine algorithm to classify the real-time traffic data and vehicle speed change values to obtain the distribution characteristics of the traffic status value. The fourth major step (step S1074) If the traffic status value is lower than the preset threshold, the signal light parameters are updated through the parameter adjustment amount to obtain a new state improvement degree. The fifth major step (step S1075) adjusts the data acquisition frequency according to the state improvement degree and the neighborhood influence radius to obtain the optimized sensor data stream. Among them:

[0200] Specifically, when determining the smoothing rate allocation range through the time window boundary, it is first necessary to clearly define the time window boundary, which is usually based on the traffic characteristics of the peak period or the trough period.

[0201] For example, the peak time window of a certain intersection is set to 7:00 to 8:00. Through historical data analysis, the smoothing rate allocation range can be preliminarily set to 0.6 to 0.8.

[0202] For example, the initial traffic light parameter adjustment can be set to 20 seconds based on a smoothing rate lower limit of 0.6, and 30 seconds based on an upper limit of 0.8. This range provides a foundation for subsequent optimization. Acquiring real-time traffic flow data and vehicle speed changes from sensor data streams is a key step.

[0203] Specifically, the sensor collected data every 5 seconds, recording that the eastbound traffic volume increased from 60 to 80 vehicles and the speed decreased from 45 km / h to 38 km / h.

[0204] Understandably, these data reflect real-time changes in intersection pressure.

[0205] In one possible implementation, if eastbound traffic surges while southbound traffic remains at 40 vehicles with a stable speed, it indicates that the eastbound traffic puts more pressure on the neighborhood.

[0206] It's important to note that intersection relevance is determined by the degree of traffic overflow. If stranded eastbound vehicles affect southbound traffic, the impact radius could extend to 50 meters. When using a support vector machine algorithm to classify real-time traffic flow data and vehicle speed changes, the principle is to map the data into a high-dimensional space and classify traffic states.

[0207] For example, eastbound traffic of 80 vehicles at a speed of 38 km / h is classified as "congested", while southbound traffic of 40 vehicles at a speed of 45 km / h is "smooth".

[0208] In one embodiment, the traffic status value is quantified from 0 to 1, with congestion being 0.4 and unobstructed being 0.7, and the preset threshold is set to 0.5.

[0209] Preferably, when the eastbound state value 0.4 is lower than 0.5, a parameter adjustment is triggered to increase the green light duration from 20 seconds to 25 seconds. If the traffic state value is lower than the threshold, the signal light parameters are updated by the parameter adjustment amount, and the degree of state improvement can be observed.

[0210] For example, after the eastbound green light was increased to 25 seconds, the traffic volume dropped to 70 vehicles, the status value rose to 0.55, and the improvement was 0.15.

[0211] In one embodiment, the data collection frequency is adjusted based on the improvement degree and the neighborhood influence radius. If the improvement degree is less than 0.2 and the influence radius exceeds 50 meters, the frequency is increased from every 5 seconds to every 3 seconds. This adjustment improves data accuracy and ensures that the optimization solution meets actual needs.

[0212] In one possible implementation, the effects of the adjustments were verified across multiple directions. After the eastbound green light was extended to 25 seconds, the number of stranded vehicles dropped to 65, the southbound speed remained at 45 km / h, and the westbound wait time increased by only 2 seconds, demonstrating that neighborhood impacts were effectively controlled.

[0213] Preferably, if traffic decreases during holidays, the eastbound duration can be flexibly adjusted back to 20 seconds. This dynamic feature improves the adaptability of the solution and ensures traffic efficiency in different scenarios.

[0214] It is understandable that the entire process is closely linked, from the time window boundary to the sensor data flow, to the algorithm classification and parameter adjustment.

[0215] For example, increasing the state improvement not only alleviates congestion in a single direction but also avoids chain reactions at neighboring intersections. This approach ensures flexible and precise traffic management by driving decisions with real-time data.

[0216] This step outputs the optimized baseline solution, which includes seven major steps, namely, the first major step (step S1081) obtains traffic density data through the prediction model, extracts the change trend and deviation data, and obtains the initial analysis results. The second major step (step S1082) calculates the traffic propagation coefficient based on the initial analysis results and the dynamic update frequency, and determines the updated propagation parameters. The third major step (step S1083) adjusts the traffic propagation model through the deviation data if the updated propagation parameter exceeds the preset threshold value to obtain the corrected coefficient. The fourth major step (step S1084) uses the corrected coefficient to update the green light duration calculation logic to obtain the adjusted duration value. The fifth major step (step S1085) generates a baseline solution through the adjusted duration value to determine whether it meets the dynamic update requirements. The sixth major step (step S1086) If the baseline solution meets the requirements, the solution is optimized in combination with the traffic density change trend to obtain the final adjustment result. The seventh major step (step S1087) updates the green light duration allocation logic according to the final adjustment result and outputs the optimized baseline solution. Among them:

[0217] Specifically, obtaining traffic density data through prediction models is a basic step in optimizing traffic light control.

[0218] For example, historical traffic data can be combined with variables such as weather and time period to predict the traffic density at a certain intersection during peak hours.

[0219] For example, the predicted traffic density for a main road during the morning rush hour is 2,000 vehicles per hour, but drops to 800 vehicles per hour during off-peak hours. Extracting trend and deviation data reveals a gradual decrease in traffic density after the peak period. Deviation data also reflects the difference between the predicted and actual values, such as a 10% overestimation. This deviation analysis helps refine the model's accuracy. One possible implementation approach is to calculate the traffic propagation coefficient based on the initial analysis results combined with a dynamic update frequency. The traffic propagation coefficient reflects the extent to which traffic flow from one intersection spreads to adjacent intersections.

[0220] Specifically, if the traffic density at a certain intersection increases, the propagation coefficient may be adjusted from 0.6 to 0.8, indicating that downstream intersections are more affected.

[0221] It should be noted that the dynamic update frequency can be adjusted based on the speed of traffic flow, for example, every 5 minutes during peak hours and every 15 minutes during off-peak hours. If the updated propagation parameter exceeds a preset threshold, such as 0.9, the model is adjusted based on the deviation data.

[0222] For example, if the deviation indicates that the predicted traffic flow is spreading too fast, the coefficient can be reduced to 0.7, which is closer to reality. The revised coefficient is used to update the green light duration calculation logic.

[0223] It is understandable that the duration of the green light is directly related to the traffic density.

[0224] In one embodiment, if the traffic density at an intersection is 1,500 vehicles per hour and the correction factor is 0.75, the green light duration can be adjusted from 30 seconds to 40 seconds, improving traffic efficiency. After generating a baseline solution based on the adjusted duration value, it is determined whether it meets the dynamic update requirements.

[0225] For example, if the baseline solution shows a 15% reduction in waiting time, it meets the requirements, otherwise further adjustments are required.

[0226] Preferably, the optimization plan is combined with the changing trend of traffic density.

[0227] For example, if the forecast shows that traffic density will continue to decline over the next 30 minutes, the green light duration can be appropriately shortened to 35 seconds to save resources while maintaining smooth operation. The final adjustment results update the green light duration allocation logic and output the optimized baseline solution.

[0228] In one possible implementation, the eastbound green light duration at an intersection is adjusted from 40 seconds to 45 seconds, while the westbound green light duration is simultaneously shortened to 35 seconds, creating a coordinated control system that effectively balances traffic flow in all directions.

[0229] Specifically, the prediction model relies on real-time sensor data collection, with deviations adjusted by comparing historical data. The traffic flow propagation coefficient is calculated by referencing speed changes at adjacent intersections. For example, if the speed at an upstream intersection drops to 20 km / h, the downstream intersection will also experience a 15% decrease. Green light duration adjustments are primarily updated on a minute-by-minute basis to ensure real-time performance.

[0230] For example, this method can reduce the duration of intersection congestion by about 20% during peak hours, avoid unnecessary waiting during off-peak hours, and improve overall traffic flow.

[0231] It can be understood that each link supports each other, forming a complete logical chain from prediction to adjustment, ensuring that the plan is both practical and flexible.

[0232] like Figure 2As shown, at the same time, the present invention also provides an intelligent traffic signal control system based on the Internet of Things, including module 101, module 102, module 103, module 104, module 105, module 106, module 107, and module 108. The present invention discloses an intelligent traffic signal control method based on the Internet of Things and deep learning. The method collects traffic flow, vehicle speed and queue length data in real time through road sensors, and uses time series analysis and deep learning prediction models to calculate the trend of traffic density changes. According to the flow propagation coefficient, edge delay time and flow balancing factor, the present invention adopts a collaborative optimization algorithm to dynamically adjust the green light duration, and optimizes the phase switching interval through real-time feedback gain data. When the degree of improvement in traffic conditions is lower than expected, the present invention can extract deviation data from the prediction model, recalculate the flow propagation coefficient, and realize adaptive adjustment of the green light duration. This method can effectively alleviate urban traffic congestion, improve road traffic efficiency, and provide an innovative solution for smart city traffic management.

[0233] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. An intelligent traffic signal control method based on the Internet of Things, characterized in that: include: S102: Based on node traffic and historical data, the traffic density change trend is calculated using a deep learning prediction model. Key features are extracted from the traffic density change prediction value, and the current traffic density is adjusted through data comparison. The traffic propagation coefficient is calculated based on the corrected traffic density sequence to obtain the traffic propagation coefficient. S103: Based on the traffic propagation coefficient, the traffic state variation range is determined through data fusion analysis. The intersection state is analyzed and judged to obtain a preliminary saturation determination result. The initial quantized value of the edge delay time is obtained by extracting the fluctuation characteristics. The quantized value is adjusted using a linear regression algorithm to obtain a corrected edge delay time. The presence of a local congestion point is determined based on the intersection state. The direction of flow adjustment is determined using a support vector machine algorithm to achieve dynamic flow distribution. S104: Based on the quantized edge delay values, an adjacency weight matrix is calculated using a graph network algorithm. Traffic concentration areas within the region boundaries are divided using a clustering algorithm. The average delay value within each region boundary is calculated. Traffic fluctuation characteristics between intersections are obtained from the factor distribution. The traffic balancing factor is adjusted using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction. S105: Based on the distribution of flow balancing factors and the trend of traffic density changes, a traffic density adjustment coefficient is calculated using a collaborative optimization algorithm. A baseline adjustment offset is calculated based on the trend data. The adjustment coefficient is integrated using a linear regression algorithm. Fluctuation characteristics are extracted from the traffic density. The green light duration distribution is compared with the duration plan based on a preset threshold to determine whether to adjust the green light duration distribution. The traffic density adjustment result after collaborative optimization is calculated by adjusting the green light duration distribution to obtain the final green light duration for the current direction. S106: Based on the gain data obtained from real-time feedback, the verification process result is obtained by determining whether the green light duration is lower than the minimum duration baseline. The time window boundary is determined by adjusting the interval range. The smoothing rate distribution is calculated based on the time window boundary. The adjusted interval offset is obtained by adjusting the baseline through gain data fusion. The light position switching characteristics are updated by adjusting the interval offset. The smoothing rate change trend is compared with the minimum baseline to obtain the final green light switching interval for the current direction. S107: Based on the smoothing rate fluctuation range, the smoothing rate allocation range is determined by the time window boundary. By determining the scope of the intersection relevance on the neighborhood influence radius, the real-time traffic data and vehicle speed change values are classified using the support vector machine algorithm. The signal light parameters are updated based on the parameter adjustment amount. The optimized sensor data stream is obtained by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval. S108: Based on the full-time predicted intersection traffic density, the initial analysis results are obtained by extracting the change trend and deviation data, and the traffic propagation coefficient is calculated in combination with the dynamic update frequency. The traffic propagation model is adjusted by the deviation data, and the green light duration calculation logic is updated with the corrected coefficient. The baseline plan is generated with the adjusted duration value, and the plan is optimized in combination with the traffic density change trend. The green light duration allocation logic is updated according to the final adjustment result to output the optimized baseline plan.

2. The intelligent traffic signal control method according to claim 1, characterized in that: The intelligent traffic signal control method further includes: S101: Based on the collected real-time traffic data of multiple groups of different types at intersections, a multi-dimensional parameter set for the current direction is obtained through multi-dimensional parameter combination. A smoothed parameter sequence is obtained through time series analysis. The traffic distribution characteristics are determined by calculating the node traffic values in each direction to achieve data preprocessing. The step S101 further includes: S1011: Based on the collected real-time current direction traffic, current direction vehicle speed, and current direction length, a multi-dimensional parameter set of the current direction including a timestamp is obtained through multi-dimensional parameter aggregation; S1012: Based on the current direction flow of the multi-dimensional parameter set of the current direction, obtain the average flow of the current sliding window by performing a segmented averaging process based on the current direction flow within the sliding window range; and generate a current direction average flow sequence by combining the average flows of the current sliding window based on multiple segmented averaging processes; Based on the current direction vehicle speed of the multi-dimensional parameter set of the current direction, the average vehicle speed of the current sliding window is obtained by performing a segmented average processing of the current direction vehicle speed within the sliding window range; based on the average vehicle speed of the current sliding window processed multiple segments, a sequence of average vehicle speeds of the current direction is generated by combining the window average vehicle speeds; Based on the current direction length of the multi-dimensional parameter set of the current direction, the average length of the current sliding window is obtained by segmented averaging of the lengths of the current direction within the sliding window; based on the average lengths of the current sliding window processed multiple times by segmented averaging, the average lengths of the current sliding window are combined to generate a sequence of average lengths of the current direction; S1013: Based on the average flow sequence in the current direction, obtain the average flow value and the flow distribution in the current direction through multi-segment averaging processing; based on the average flow value in the current direction, obtain the congestion level of the main flow direction of the current intersection by comparing the average flow values in the current direction; Based on the average speed sequence of the current direction, the average speed value and the speed distribution of the current direction are obtained through multi-segment averaging. Based on the average speed value of the current direction, the congestion level of the main speed direction of the current intersection is obtained by comparing the average speed values of the current direction. Based on the average length sequence of the current direction, the average length value and the distribution of the length in the current direction are obtained through multi-segment averaging. Based on the average length value of the current direction, the congestion level of the main length direction of the current intersection is obtained by comparing the average length values of the current directions. Based on the current direction flow distribution, current direction speed distribution, and current direction queue distribution, the accuracy of the current intersection parameter distribution and the congestion degree of the main direction is determined through a multi-parameter verification method of the congestion degree of the current main flow direction of the intersection, the congestion degree of the current main speed direction of the intersection, and the congestion degree of the current main queue direction; Based on the parameter distribution of the current intersection and the parameter distribution of historical intersections in the same time period, the credibility of the parameter distribution and congestion level of the current intersection is determined through a multi-method verification method based on the congestion level of the main direction of the current intersection and the congestion level of the main direction of the historical intersection in the same time period; Based on the parameter distribution of the current intersection and the parameter distribution of the historical intersection in different time periods, the regularity of the parameter distribution and the congestion degree of the current intersection is determined by verifying the regularity of the congestion degree of the main direction of the current intersection and the congestion degree of the main direction of the historical intersection in different time periods.

3. The intelligent traffic signal control method according to claim 1, characterized in that: The step S102 further includes: S1021: Based on the parameter distribution of the current intersection and the parameter distribution of historical intersections in the current time period, and according to the regularity of historical data and the collected real-time intersection traffic data, the traffic density in the direction of the current intersection in the next window is predicted using a deep learning prediction model; S1022: Based on the traffic flow density in the current direction of the current segment window, the peaks and valleys of the traffic flow density in the current direction of each segment window are extracted to determine a predicted fluctuation range of the current window as a traffic flow density trough to a traffic flow density peak; S1023: Based on the current direction traffic density of the current segment window, a comparison is made between the predicted fluctuation range of the current segment window and a preset window traffic density threshold range to determine whether the range is exceeded. If the range is exceeded, a correction value of the current direction traffic density is obtained based on the current direction traffic density of the current segment window by multi-segment averaging of the current segment window and other segment windows. S1024: Based on the traffic propagation coefficient in the current direction, the traffic propagation coefficient in the other direction is obtained through the linear relationship between the traffic density and the traffic propagation coefficient.

4. The intelligent traffic signal control method according to claim 1, characterized in that: The step S103 further includes: S1031: If the traffic propagation coefficient in the current direction exceeds the traffic aggregation threshold, determine the traffic change state in the current direction based on the traffic propagation coefficient in the current direction, according to the change in vehicle speed in the current direction and the change in the length of the vehicle in the current direction; S1032: Based on the traffic change state in the current direction, determine the saturation factor of the current direction using the linear relationship between the flow propagation coefficient in the current direction and the saturation factor in the current direction; determine whether the saturation factor in the current direction has reached a saturation suppression threshold; if so, determine that the preliminary saturation state in the current direction is a critical state close to full load; S1033: If the initial saturation state in the current direction is a critical state close to full load, then based on the average vehicle speed sequence in the current direction, the fluctuation range of the average vehicle speed in the current direction is determined by extracting the peaks and valleys of the average vehicle speed sequence in the current direction; based on the fluctuation range of the average vehicle speed in the current direction, the initial side delay time is determined according to the increasing and decreasing relationship between the fluctuation range of the average vehicle speed in the current direction and the initial side delay time; S1034: Based on the initial side delay time, a linear regression adjustment is performed between the initial side delay time and the length of the current direction length to obtain a corrected initial side delay time; S1035: If the corrected initial side delay time exceeds the delay threshold, determine the local congestion state of the current direction based on the corrected initial side delay time and the current direction saturation factor; S1036: Based on the local congestion status of the current direction, the direction of intersection adjustment is determined by the vector machine algorithm; based on the direction of intersection adjustment, according to the flow rate in the adjustment direction, the traffic change status in the adjustment direction, and the initial edge delay time of the adjustment direction, a dynamic flow distribution plan for the adjustment direction is obtained.

5. The intelligent traffic signal control method according to claim 1, characterized in that: The step S104 further includes: S1041: Based on the corrected edge delay time in each direction of the intersection, weights are assigned according to the traffic interaction frequency and the distance between intersections, and a graph network algorithm is used to establish a weight matrix for each direction of the intersection. Based on the weight matrix for each direction of the intersection, a flow balancing factor distribution for each direction is generated according to the linear relationship between the weight of each direction of the intersection and the flow balancing factor for each direction of the intersection. S1042: Based on the distribution of flow balancing factors in each direction and the correlation of directions with similar weights, the flow correlation in each direction is determined; based on the intersection direction of the flow correlation in each direction, different flow concentration areas are generated through K-means clustering; S1043: Based on each traffic concentration area, average the corrected edge delay times of each traffic concentration area to obtain an average delay value of each traffic concentration area; S1044: If the average delay value of the current traffic concentration area exceeds the delay threshold, the fluctuation range of the current direction traffic balancing factor is determined based on the current direction traffic balancing factor sequence by extracting the peaks and valleys of the current direction traffic balancing factor sequence; if the fluctuation range of the current direction traffic balancing factor exceeds the factor fluctuation threshold, the load capacity of the current direction is determined to be high; otherwise, the load capacity of the current direction is determined to be low. S1045: If the current direction load capacity is high load, based on the current direction flow balancing factor sequence, the current direction flow balancing factor sequence and the weight matrix are linearly regressed to obtain an optimized current direction flow balancing factor.

6. The intelligent traffic signal control method according to claim 1, characterized in that: The step S105 further includes: S1051: Based on the historical intersection traffic data and the real-time intersection traffic data, according to the optimized linear relationship between the current direction traffic flow balancing factor and the current direction traffic flow density adjustment coefficient, a current direction traffic flow density adjustment coefficient is obtained; based on the current direction traffic flow density adjustment coefficient, according to the linear relationship between the current direction traffic flow density adjustment coefficient and the current direction green light duration, a preliminary current direction green light duration is obtained; S1052: Based on the traffic change state in the current direction, according to the offset relationship between the traffic density in the current direction and the preliminary green light duration in the current direction, determine the optimized green light duration in the current direction; S1053: Based on the optimized green light duration in the current direction, an updated current direction traffic flow balancing factor is obtained by linear regression adjustment of the optimized green light duration in the current direction and the current direction traffic flow balancing factor; based on the updated current direction traffic flow balancing factor, an optimized current direction traffic flow density is obtained by linear regression adjustment of the intersection distance and the current direction vehicle speed and the current direction traffic flow density adjustment coefficient; S1054: Based on the traffic density sequence in the current direction, the fluctuation range of the traffic density in the current direction is determined by extracting the peaks and valleys of the traffic density sequence in the current direction; if the fluctuation range of the traffic density in the current direction is greater than the traffic density fluctuation threshold, the traffic density area in the current direction is determined to be a high traffic density area; S1055: If the traffic density area in the current direction is a high traffic density area, then based on the current direction traffic flow balancing factor, according to the optimized linear relationship between the traffic density in the current direction and the traffic flow balancing factor, an updated current direction traffic flow balancing factor is obtained; based on the updated current direction traffic flow balancing factor, according to the updated linear relationship between the traffic flow balancing factor in the current direction and the green light duration in the current direction, an optimized green light duration in the current direction is obtained; determine whether the optimized green light duration in the current direction exceeds a green light duration threshold; if so, adjust the current direction green light duration based on the linear relationship between the distance between intersections and the current direction vehicle speed and the green light duration in the current direction to obtain an adjusted green light duration in the current direction; S1056: Based on the adjusted green light duration for the current direction, a final green light duration for the current direction is determined by a linear relationship between the adjusted green light duration for the current direction and the traffic density for the current direction.

7. The intelligent traffic signal control method according to claim 1, characterized in that: The step S106 further includes: S1061: Based on the final green light duration in the current direction and using real-time incremental traffic feedback in the current direction, determine whether the final green light duration in the current direction is lower than the minimum baseline for the green light duration. If so, obtain a green light duration adjustment plan. S1062: Based on the green light duration adjustment scheme, extract the light position switching characteristics of the green light duration adjustment scheme to obtain the green light switching interval for the current direction. Based on the green light switching interval for the current direction, determine whether there are still vehicles stranded when the green light for the current direction ends. If there are still vehicles stranded, preliminarily adjust the green light switching interval for the current direction. S1063: Based on the preliminary adjustment of the green light switching interval in the current direction, a time window boundary for the peak period is set, and the smoothing rate fluctuation range is obtained by extracting the peaks and valleys of the average traffic sequence in the current direction based on the time window boundary; S1064: Based on the smoothing rate fluctuation range, if the smoothing rate fluctuation range exceeds the smoothing rate threshold, a green light interval offset is obtained based on the vehicle speed increment in the current direction and a linear relationship between the vehicle speed increment in the current direction and the green light duration increment; S1065: Based on the green light switching interval of the current direction, the green light switching interval of the current direction is compensated by the green light interval offset to obtain a compensated green light switching interval of the current direction; based on the compensated green light switching interval of the current direction, the time window boundary of the peak period is set again, and the peak and valley of the average traffic sequence of the current direction based on the time window boundary are extracted again to obtain the verified smoothing rate fluctuation range; S1066: Based on the verified smoothing rate fluctuation range, determine whether the verified smoothing rate fluctuation range meets the smoothing rate standard. If it does not meet the smoothing rate standard, adjust the green light switching interval for the current direction to generate a final green light switching interval for the current direction.

8. The intelligent traffic signal control method according to claim 1, characterized in that: The step S107 further includes: S1071: Based on the smoothing rate fluctuation range, the initial signal light parameter adjustment amount is obtained by extracting the current direction green light duration and the current direction green light switching interval; S1072: Based on the current direction traffic flow, the current direction vehicle speed, and the correlation between the traffic flows in each direction, determine whether there is a vehicle backlog in the current direction. If there is a vehicle backlog, determine the scope of the neighborhood influence radius as the degree of vehicle overflow in the current direction. S1073: Based on the current direction traffic flow, the current direction vehicle speed, and the current direction traffic overflow degree, the current direction intersection state value is quantized in the range of 0-1 using a support vector machine algorithm; S1074: Based on the current direction intersection state value, if the current direction intersection state value is lower than the intersection state threshold, an intersection state error value between the current direction intersection state value and the intersection state threshold is obtained, and based on the intersection state error value, the current direction green light duration and the current direction green light switching interval are dynamically adjusted according to a linear relationship between the current direction green light duration and the current direction green light switching interval and the intersection state error value; S1075: Based on the dynamically adjusted current direction green light duration and current direction green light switching interval, and according to the overflow degree of traffic in the current direction, the data collection frequency is adjusted to obtain optimized current direction traffic flow and current direction vehicle speed.

9. The intelligent traffic signal control method according to claim 1, characterized in that: The step S108 further includes: S1081: Based on the predicted traffic flow density at the intersection for all time periods, a change trend of the traffic flow density in the current direction of the current time period is obtained by comparing the predicted traffic flow density in the current direction of the current time period with the predicted traffic flow density in the current direction during the peak period; and a predicted deviation of the traffic flow density in the current direction is obtained by comparing the predicted traffic flow density in the current direction with the actual traffic flow density in the current direction. S1082: Based on the predicted deviation of the traffic flow density in the current direction and the linear relationship between the traffic flow density change trend in the current direction and the traffic propagation coefficient, determine the traffic propagation coefficient in the current direction; S1083: Based on the prediction deviation of the traffic flow density in the current direction, the optimized traffic propagation coefficient in the current direction is determined by correcting the coefficient of the traffic propagation coefficient in the current direction and the coefficient of the prediction deviation; S1084: Based on the optimized traffic propagation coefficient of the current direction and a linear relationship between the optimized traffic propagation coefficient of the current direction and the green light duration of the current direction, a green light duration of the current direction is determined, and at the same time, a temporary baseline plan for the current direction is generated; S1085: Based on the temporary baseline plan for the current direction, determine whether it meets the dynamic update requirements by comparing it with the actual traffic density feedback for the current direction; S1086: If the dynamic update requirements are met, the temporary baseline solution for the current direction is converted into the official baseline solution for the current direction. Through repeated iterative prediction, the official baseline solution for the next direction is generated. S1087: Based on the predicted traffic density in the current direction and the traffic propagation coefficient in the current direction, the green light duration in the other direction is predicted through repeated iterations to form the final baseline solution for the current intersection.

10. An intelligent traffic signal control system based on the Internet of Things, characterized in that: include: Based on the intelligent traffic signal control method according to any one of claims 1 to 9, the intelligent traffic signal control system includes: Module 101: Based on the collected real-time traffic data of multiple groups of different types at intersections, a multi-dimensional parameter set for the current direction is obtained through multi-dimensional parameter combination, a smoothed parameter sequence is obtained through time series analysis, and traffic distribution characteristics are determined by calculating the node traffic values in each direction to achieve data preprocessing; Module 102: Calculates the traffic density change trend based on node traffic and historical data using a deep learning prediction model, extracts key features from the traffic density change prediction value, adjusts the current traffic density through data comparison, and calculates the traffic propagation coefficient based on the corrected traffic density sequence to obtain the traffic propagation coefficient; Module 103: Based on the traffic propagation coefficient, data fusion analysis is used to determine the range of traffic state changes. Analysis and judgment of the intersection state are used to obtain a preliminary saturation determination result. Initial quantitative values of the edge delay time are obtained by extracting fluctuation characteristics. The quantitative values are adjusted using a linear regression algorithm to obtain a corrected edge delay time. Based on the intersection state, the presence of local congestion points is determined. A support vector machine algorithm is used to determine the direction of flow adjustment to achieve dynamic flow distribution. Module 104: Calculates an adjacency weight matrix based on the quantized edge delay time using a graph network algorithm, divides the traffic concentration areas within the area boundary using a clustering algorithm, calculates the average delay value within each area boundary, obtains traffic fluctuation characteristics between intersections from the factor distribution, and adjusts the traffic balancing factor using a linear regression algorithm to obtain an optimized traffic balancing factor for the current direction; Module 105: Calculates the traffic density adjustment coefficient using a collaborative optimization algorithm based on the distribution of traffic balancing factors and the trend of traffic density changes. Calculates the baseline adjustment offset based on the trend data. Fusions the adjustment coefficient using a linear regression algorithm. Extracts fluctuation characteristics from traffic density. Determines whether to adjust the green light duration distribution based on a preset threshold and comparison with the duration scheme. Calculates the traffic density adjustment result after collaborative optimization by adjusting the green light duration distribution to obtain the final green light duration for the current direction. Module 106: Based on the gain data obtained from real-time feedback, it determines whether the green light duration is lower than the minimum duration baseline to obtain the verification process result, determines the time window boundary by adjusting the interval range, calculates the smoothing rate distribution based on the time window boundary, obtains the adjusted interval offset by adjusting the baseline through gain data fusion, updates the light position switching characteristics based on the adjusted interval offset, and compares the smoothing rate change trend with the minimum baseline to obtain the final green light switching interval for the current direction; Module 107: used to determine the smoothing rate allocation range based on the smoothing rate fluctuation range and the time window boundary, classify the real-time traffic data and vehicle speed change values by judging the scope of the intersection relevance on the neighborhood influence radius, update the traffic light parameters by parameter adjustment, and obtain the optimized sensor data stream by adjusting the data collection frequency to achieve coordinated optimization of the green light duration and the green light switching interval; Module 108: It is used to obtain the initial analysis results based on the traffic density of the intersection predicted for the entire period by extracting the change trend and deviation data, calculate the traffic propagation coefficient in combination with the dynamic update frequency, adjust the traffic propagation model through the deviation data, update the green light duration calculation logic through the corrected coefficient, generate the baseline plan through the adjusted duration value, optimize the plan in combination with the traffic density change trend, and update the green light duration allocation logic according to the final adjustment result to output the optimized baseline plan.

Images