Intersection short-time flow prediction and signal timing optimization method and system

Abstract

The application provides a kind of intersection short-time flow prediction and signal timing optimization method and system, the method collects road network information, vehicle data, queue length, channelization information, timing scheme and road condition data;Determine preliminary prediction period based on current timing scheme;Utilize real-time queue length to dynamically calculate upstream travel time, reversely determine prediction window and multi-level trace vehicle flow;Construct normalized turning weight matrix through historical same period data, combine upstream inflow and queue length compensation, calculate each turning predicted flow and phase length;Compare difference value of before and after timing scheme, if greater than threshold, then iteratively update until convergence.The application realizes prediction period self-adaptation, turning flow fine decomposition and prediction-timing closed-loop verification, improves short-time flow prediction accuracy and adaptability and robustness of signal timing scheme.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management and control technology, specifically to a method and system for short-term traffic flow prediction and signal timing optimization at intersections. Background Technology

[0002] Urban road intersections are critical nodes in the road network, and the rationality of their signal control schemes directly affects the traffic efficiency and service level of the area. Due to the significant time-varying and random nature of traffic flow demand at different times and in different directions, traditional fixed-time or time-segmented fixed-cycle signal timing schemes often struggle to match dynamic traffic demand in real time, easily leading to situations where green lights are ineffective in some directions while queues are excessively long in others. Therefore, how to optimize signal timing schemes in real time based on short-term traffic flow trends has become an important research direction for improving intersection capacity.

[0003] Currently, mainstream signal timing optimization methods can be divided into two categories: (1) End-side adaptive control method Radar, video, and geomagnetic detectors are deployed at intersections to collect real-time data on traffic flow, queue length, and time occupancy for each lane. Adaptive algorithms (such as self-organizing control and reinforcement learning) built into the intersection's traffic signal controllers then adjust timing parameters in real time. This method offers fast response times but requires a large number of high-precision roadside detection devices, resulting in high construction and maintenance costs. Furthermore, equipment failures can easily lead to system malfunctions, making it difficult to widely implement in large-scale road networks.

[0004] (2) Central offline optimization method By utilizing historical vehicle data collected from widely deployed electronic police and checkpoint devices in cities, fixed or time-segmented traffic timing schemes are generated through data statistics and modeling, and then distributed to traffic signal controllers at intersections for execution. This method is low-cost, but it relies solely on historical data and cannot respond in real time to short-term fluctuations in traffic flow. Especially during rush hours, emergencies, and inclement weather, the timing schemes lag significantly behind real-time demands, resulting in reduced control effectiveness.

[0005] In recent years, some studies have attempted to predict short-term arrival traffic flow at downstream intersections by combining traffic flow data from upstream intersections or road segments, and dynamically adjust signal timing schemes accordingly. For example, Chinese invention patent CN121861909A discloses a real-time adaptive signal timing optimization method at the center level of urban intersections. This method uses historical traffic flow data for each turn at the intersection and historical traffic flow data from upstream road segments, combined with an attention-based spatiotemporal graph convolutional network (ASTGCN) to predict short-term traffic flow, thereby generating a real-time signal timing scheme and inserting it into the daily schedule for execution. While this method achieves a certain degree of real-time response at the center level, its prediction period is based on a fixed cycle (the previous complete signal cycle), lacking adaptability to real-time physical conditions such as current queue length and road segment travel time. Furthermore, it does not perform refined spatiotemporal source tracing for traffic merging from upstream multi-level intersections, leaving room for improvement in prediction accuracy and the robustness of the timing scheme.

[0006] Furthermore, Chinese invention patent CN118486177B proposes a variable phase sequence overflow intersection signal coordination control method based on model predictive control (MPC). This method uses upstream vehicle travel data to predict the number of downstream arriving vehicles and calculates the queue length. It then solves the timing scheme using a dynamic programming model and employs a rolling optimization strategy. This method is primarily geared towards overflow control scenarios. However, its rolling optimization only executes the first stage of the optimization period before re-optimizing, lacking closed-loop verification of the deviation between predicted and actual arrival traffic. Furthermore, it does not involve steering-level traffic decomposition and iterative verification based on the steering weight matrix.

[0007] In summary, how to achieve low-cost, real-time, and high-precision short-term traffic flow prediction at intersections on a central platform, generate signal timing schemes that match the predicted traffic flow, and establish an iterative verification mechanism between prediction and timing are the technical problems that urgently need to be solved in the field of intelligent traffic signal control. Summary of the Invention

[0008] This invention aims to provide a method for short-term traffic flow prediction and signal timing optimization at intersections. It can dynamically adjust the prediction period based on real-time queue length, integrate the source of turning traffic flow and turning weight matrix of multi-level upstream intersections, and achieve self-consistency verification of prediction and timing through comparison and iteration of previous and subsequent signal timing schemes. This improves the accuracy of short-term traffic flow prediction and the adaptability of signal timing schemes, and further enhances the traffic efficiency of intersections.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A method for short-term traffic flow prediction and signal timing optimization at intersections includes the following steps: S1. Collect basic data required for short-term traffic flow prediction and signal timing at intersections; the basic data includes, but is not limited to, road network information, vehicle data from intersection electronic police or checkpoints, real-time queue lengths of turning points at intersections, channelization information of intersections, timing scheme information for different time periods at intersections, execution information of intersection timing schemes, and real-time internet traffic data of the road network. S2. Obtain a preliminary plan for predicting the next cycle of the intersection and the preliminary prediction time period for each phase using the basic data; the preliminary plan includes predicting the intersection. No. Predicted start time of each phase Take the period before the end of the current period If the time is in seconds, then the termination time is predicted. ,in, For the first part of the preliminary plan Each phase duration, For the predicted intersection, the phase number is used. Each upstream branch Its predicted start time Predict the termination time Take the maximum value among all phases, that is ; S3. Based on steps S1 and S2, determine the traffic flow of each branch of the predicted intersection and each branch of the primary intersection merging into the predicted intersection at the predicted time; the primary intersection To the predicted intersection exist Travel time at any moment , For the length of the road segment, This represents the real-time average queue length. The speed of the road segment; and the primary intersection is determined in reverse based on the travel time. Predicted start time and predicted termination time The decision to continue tracing upstream to secondary and higher-level intersections is determined by whether the following set of preset inequality conditions are met: For primary intersections... If the following inequalities are satisfied simultaneously, then the upstream tracing will not continue; otherwise, the first-level intersection will be used. upstream intersection As a new primary intersection, the same logic continues to be applied in tracing its origins: ; ; ; in: To predict intersections Upstream branch; Class I intersection To the predicted intersection At the predicted start time Travel time; For the first part of the preliminary plan Each phase duration; For the number of phases; Redundancy design duration for predicting computation time before the end of the cycle; upstream intersection Merging into the primary road intersection Real-time average queue length in each direction; Standard vehicle length; This represents the average headway. Class I intersection Branches The moment the green light turns on; Class I intersection The predicted termination time.

[0010] S4. Based on step S3, determine the traffic flow and signal duration for the preliminary prediction period of each phase at the intersection; wherein, the predicted turning traffic flow for each branch of the intersection is based on the turning weight matrix obtained from historical data of the same period. The predicted traffic flow is obtained by calculating the traffic flow of each branch of the upstream intersection. Specifically, this includes: constructing a turning weight matrix of traffic flow of each branch of the upstream intersection at the predicted intersection; solving the normalized turning weight matrix using historical data from the same period; and combining the upstream merging traffic flow and queue length compensation to obtain the predicted traffic flow of each branch and each turn at the predicted intersection. Then, according to the release mode adopted in the initial plan (returning traffic in opposite directions or one-way traffic), the duration of each phase is calculated using the corresponding phase duration calculation formula. S5. Determine the difference in phase changes between the previous and subsequent signal timing schemes. If the difference in phase duration changes is no greater than 1 second, then determine the final predicted flow rate and signal timing scheme; otherwise, use the timing scheme calculated in step S4 as the new initial scheme and return to step S2 to re-execute.

[0011] A short-term traffic flow prediction and signal timing optimization system for intersections includes: The target selection module is used to select the predicted intersection and its upstream intersection, and to designate the predicted intersection and its upstream intersection as the intersection to be tested; The data statistics module is used to acquire signal control schemes, vehicle passage data, real-time queue lengths of each branch turning, channelization information, average speed of the road segment, number of different turning lanes, and Internet traffic data for the intersection under test and its upstream intersections according to a predetermined time period. The calculation module is used to predict the traffic flow of each branch of the intersection and each branch merging into the predicted intersection at the predicted time, based on the signal control scheme parameters of the predicted intersection and its upstream intersection, the turning traffic flow of each branch, the queue length, the average speed of the road segment and lane data; and to calculate the signal timing scheme based on the predicted turning traffic flow of the intersection branches during the predicted time period. The judgment module is used to determine whether traffic flow from multiple upstream intersections can merge into the predicted intersection at the predicted time. It makes the judgment and outputs the judgment based on the travel time of the road segment from the upstream intersection to the predicted intersection and the waiting time at the traffic light at the upstream intersection. It is also used to compare each phase of the preceding and following signal timing schemes, determine the final signal timing scheme of the predicted intersection, and output it.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention collects and predicts the queue length and road speed of each branch turning at the intersection in real time, dynamically calculates the travel time from the upstream intersection to the predicted intersection, and determines the prediction time window for each upstream intersection in reverse. The prediction period can adapt to changes in traffic conditions (the window is advanced when congested and delayed when smooth), making short-term traffic flow prediction more in line with actual arrival patterns and effectively reducing prediction errors.

[0013] 2. This invention not only considers the first-level upstream intersections, but also uses preset inequality conditions to determine whether it is necessary to trace back to the second-level, third-level, or even higher-level upstream intersections until the condition that the traffic flow can arrive within the predicted time period is met. On this basis, it uses historical data from the same period to construct and solve the turning weight matrix, and combines the traffic flow of each branch upstream and real-time queue length compensation to accurately calculate the predicted traffic flow of each branch turning at the predicted intersection. It does not rely on a large amount of training data and is convenient for engineering deployment.

[0014] 3. After calculating the predicted traffic flow and preliminary timing scheme, this invention compares the difference in the duration of each phase with the initial scheme. Only when the difference in the duration of all phase changes is no greater than 1 second will the final scheme be output; otherwise, the new scheme will be used as the initial scheme, and the upstream will be returned to re-execute the prediction and timing calculation until convergence. This dual verification closed loop can automatically correct the scheme mismatch problem caused by prediction deviation, and significantly improve the robustness of the signal timing scheme and the actual control effect.

[0015] 4. This invention fully utilizes vehicle data collected by electronic police, checkpoints and other equipment already widely deployed in cities, as well as Internet traffic data. It eliminates the need to install a large number of additional roadside detectors. Multi-level source tracing, turning flow prediction and adaptive generation and iterative optimization of timing schemes can be completed on the central platform. Compared with edge adaptive control, hardware costs are significantly reduced. Compared with offline optimization driven by pure historical data, real-time response capability is significantly improved. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for short-term traffic flow prediction and signal timing optimization at intersections provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the budget prediction intersection and its upstream intersection provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the predicted travel time of traffic merging from upstream intersections provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the relationship between the predicted time of the predicted intersection and the upstream intersection and the merging traffic flow, provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The illustrative embodiments and descriptions of this invention are for illustrative purposes only and are not intended to limit the invention.

[0018] Example 1: Short-term traffic flow prediction and signal timing optimization method at intersections like Figure 1 As shown, this embodiment provides a method for short-term traffic flow prediction and signal timing optimization at intersections, using a crossroads in an urban road as the prediction intersection. For example, the specific steps include:

[0019] S1, Basic Data Collection Specifically, this embodiment collects the following basic data: Road network information: including the number of intersections, road segment lengths, lane function divisions, etc.; Traffic data from intersection cameras or checkpoints: records the time of passage, lane, license plate, and other information for each vehicle; Real-time queue length of each branch turn at the intersection: The real-time number of vehicles queuing for each turn (left turn (including U-turn), straight, and right turn) at each approach lane is obtained through electronic police or radar detectors and converted into queue length (meters). Intersection channelization information: number of lanes at each approach, number of dedicated turning lanes, configuration of mixed traffic lanes, etc.; Timing scheme information for intersections at different times: fixed timing schemes for off-peak, peak, and nighttime periods; Execution information of the intersection timing scheme: current cycle duration, phase scheme, and current phase; Real-time internet traffic data for the road network: Real-time average speed (m / s) for each road segment, which can be obtained through third-party map services.

[0020] S2. Determine the preliminary plan and preliminary forecast period. By determining the preliminary plan for the next cycle of the intersection and the preliminary prediction time period for each phase, when acquiring data for the signal timing plan in the next cycle, it is convenient to statistically analyze which time period of the intersection the traffic flow will be in based on the preliminary time domain and phase conditions. The preliminary prediction time period for each phase allows for a more refined statistical analysis of the traffic flow merging at the end of each phase, enabling the targeted design of the required phase duration for the traffic flow merging at that phase. This leads to faster calculations and improved data reliability.

[0021] First, obtain the current time. The current operating cycle will be The moment ended, among which To avoid redundant design time, this embodiment takes... It lasts for 3 seconds.

[0022] If the current period ends at time If the current periodic operation is still within the current time period, the preliminary plan will adopt the current operation plan; otherwise, the periodic operation plan for the next time period will be adopted.

[0023] After determining the preliminary plan, the predicted intersections are set. No. The prediction start time for each phase is before the end of the current cycle. Seconds, that is: (1); The predicted termination time is: (2); In formula (1)-(2): To predict intersections No. Predicted start time (seconds) for each phase; To predict intersections No. The predicted termination time (in seconds) for each phase; For redundancy, the duration is set to 3 seconds in this embodiment; To predict intersections Initial Scheme The duration of each phase (including green and yellow lights, in seconds); To predict intersections The number of phases in the initial scheme.

[0024] For each upstream branch Its prediction start time and prediction end time are the maximum values ​​among all phases: (3); (4); In formula (3)-(4): To predict intersections No. Predicted start time (seconds) for each phase; To predict intersections No. The predicted termination time (in seconds) for each phase; It is a Class I intersection; , Predicted intersections of Predicted start and end times of the directional approach lane; This indicates the operation of retrieving the maximum value.

[0025] S3, multi-level upstream intersection traffic flow tracing like Figure 2 As shown, a primary intersection is the adjacent intersection upstream of each branch of the predicted intersection in the road network (in this embodiment, the predicted intersection is a crossroads, and there are a total of 4 primary intersections). Except for the branch connecting to the predicted intersection, the right-turn, straight-ahead, and left-turn (including U-turn) traffic flows of the other branches of the primary intersection merge into the branches of the predicted intersection; a secondary intersection is the upstream intersection of the primary intersection. Except for the branch connecting to the primary intersection, the right-turn, straight-ahead, and left-turn (including U-turn) traffic flows of the other branches of the secondary intersection merge into the branches of the primary intersection; and so on.

[0026] S31, First-level intersection prediction time period Calculate the first-level intersection To the predicted intersection exist Travel time at each point in time: (5); In formula (5): Class I intersection To the predicted intersection At any moment The journey time (in seconds); Class I intersection To the predicted intersection The length of the road segment (meters); For the road section at time Real-time average speed (m / s); To predict intersections At any moment of The average queue length (in meters) at the directional approach lanes is obtained by averaging the queue lengths of the three turns: (6); In formula (6): , , The real-time average queue length (in meters) for left turn (including U-turns), straight, and right turn lanes are respectively.

[0027] S32. Determine the primary intersection based on the travel time. Predicted start time and predicted termination time : (7); (8); In formulas (7)-(8): , First-level intersections The prediction start time and prediction end time; , Predicted intersections Branches The prediction start time and prediction end time; Class I intersection To the predicted intersection At any moment The journey time (in seconds).

[0028] S33. Determine whether to continue tracing upstream. This embodiment sets the following three inequalities (9)-(11) to determine the first-level intersection. Whether the incoming traffic volume during the predicted time period can fully cover the predicted demand at the intersection will determine whether it is necessary to add traffic to secondary intersections. Tracing the source.

[0029] The travel time is sufficient to cover the entire forecast period: (9); In formula (9): Class I intersection To the predicted intersection At any moment The journey time (in seconds); For redundancy, the duration is set to 3 seconds in this embodiment; To predict intersections Initial Scheme The duration of each phase (including green and yellow lights, in seconds); To predict intersections The number of phases in the initial scheme.

[0030] Secondary intersection The time it takes for queued vehicles to be released can fill the remaining time slots: (10); In formula (10): Secondary intersection Merging into the primary road intersection Real-time average queue length (meters) in each direction; For the standard vehicle length, this embodiment uses 7 meters; Class I intersection of The average headway of vehicles entering the directional lane is generally taken as 2.0 to 3.5 seconds; the left side represents the total time (in seconds) required for all vehicles in the queue to be released.

[0031] The upstream intersection has a green light phase that allows vehicles to proceed: (11); In formula (11): Class I intersection No. Branches (i.e., secondary intersections) The green light start time (in seconds) for the direction is indicated; the right side is the predicted end time of the first-level intersection minus 2 seconds, ensuring that vehicles have at least 2 seconds of green light to exit.

[0032] If inequalities (9)-(11) all hold, then a first-level intersection is considered to be... itself and secondary intersections The queue of vehicles is sufficient to meet the predicted demand, so there is no need to trace back upstream. At this point, the primary intersection... During the forecast period Standard number of vehicles passing through the turn signal electronic police system And calculate the additional number of vehicles that can be released based on the green light time during that period: (12); In formula (12): , Class I intersection Branches exist The start and end times of the green light for the predicted direction of the intersection within the specified time period; Class I intersection Branches The direction of the merging is calculated based on the number of dedicated turning lanes at the predicted intersection. Class I intersection Branches The traffic volume (in vehicles) merging into the predicted intersection during the predicted time period.

[0033] If the inequality does not hold, then take the first-level intersection as an example. upstream intersection As a new first-level intersection, the same logic continues to trace back to the second-level, third-level, and so on, until the inequality conditions are met or the preset maximum tracing level is reached.

[0034] S34, Predict the total merging traffic flow of each branch at the intersection Ultimately, predict the intersection of The total number of standard vehicles predicted for the directional entrance lanes during the prediction period is: (13); In formula (13): Indicates the predicted intersections within the predicted time period Branches The predicted standard number of vehicles (units). , , The real-time average queue length (in meters) for left turn (including U-turn), straight, and right turn lanes, respectively. , , These are the dedicated lanes for left turns (including U-turns), straight ahead, and right turns, respectively. upstream intersection Branches Merging into the predicted intersection The traffic flow is obtained by summing the contributions from upstream intersections at each level; For the upstream intersection The total number of branches.

[0035] S4. Turning flow decomposition and timing scheme calculation S41. Construction of the Turning Weight Matrix For the upstream intersection Construct the traffic flow of each branch at the predicted intersection The turning weight matrix : (14); In formula (14): upstream intersection Branches Traffic flow at predicted intersections Weighting of left turns (including U-turns); The weight value for the straight line; This is the weight value for turning right.

[0036] S42, Historical Concurrent Data Calibration First, obtain and predict the time period. Historical turning traffic data from the same period, and constructing predictive intersections. Branches Turning traffic matrix : (15).

[0037] Then obtain the predicted time period [ Historical predictions of intersections Branches upstream intersection where it merges Traffic flow matrix of each branch : (16).

[0038] S43. Calculate the turning weight matrix of the predicted intersection during the prediction period: (17); In formula (17): This represents the matrix transpose operation.

[0039] Further solution The values ​​of each parameter are then determined. The steering weights of each branch are then normalized to obtain the normalized steering weight matrix. : (18).

[0040] S44, Predicting Turning Traffic Flow Based on the upstream intersection obtained in step S3 Each branch merges into the traffic flow matrix Calculate and predict intersections Turning traffic flow forecast : (19).

[0041] Then, by combining the current queue length compensation, the final predicted turning flow is obtained: (20); In formula (19)-(20): , , To predict the intersections respectively of Predicted traffic volume (in vehicles) for left turns (including U-turns), straight ahead, and right turns at the directional entrance lane. , , The real-time average queue length (in meters) for left turn (including U-turn), straight, and right turn lanes, respectively. , , These are the dedicated lanes for left turns (including U-turns), straight ahead, and right turns, respectively. , , From the upstream intersection Predicted traffic volume for merging and turning left (including U-turns), going straight, and turning right; The length of a single standard vehicle is typically 7 meters.

[0042] Phase duration calculation If the initial plan adopts a phased release approach (e.g., simultaneous release of east-west through traffic), the duration of each phase is calculated using the following formula: Left turn (including U-turn) phase duration: (twenty one); Linear phase duration: (twenty two); Right turn phase duration: (twenty three); If the initial scheme adopts a one-way release phase (single-port release), then the first... The duration of the phase is the maximum value of the times required for the three turns: (twenty four); In formulas (21)-(24): Forecast intersections within the forecast period Branches Allow the branch of the opposing phase to proceed; , , Forecast intersections within the forecast period Branches Forecast traffic volume for left turns (including U-turns), straight ahead, and right turns; , , Predicted intersections Branches Number of lanes for left turns (including U-turns), straight ahead, and right turns; , , Predicted intersections Branches The average headway for vehicles in left-turn (including U-turn), straight-ahead, and right-turn lanes is generally taken as 2.0-3.5 seconds. Indicates predicted intersection Branches Single-port release Phase duration; , , Predicted intersections Branches The duration of the left-turn (including U-turn), straight-ahead, and right-turn traffic flow phases; constant 3 indicates the yellow light duration (seconds). This indicates the operation of retrieving the maximum value.

[0043] S5, Comparison Iteration The duration of each phase in the preliminary scheme in step S2 The new phase duration calculated in step S4 Comparison: If the absolute value of the duration change of all phases is no greater than 1 second, the timing scheme is determined to have converged, and the predicted turning flow and timing scheme calculated in step S4 are output as the final result.

[0044] If any phase duration variation difference is greater than 1 second, the timing scheme calculated in step S4 is used as the new initial scheme, i.e., the timing is updated. Then return to step S2 and repeat steps S2 to S5 until the convergence condition is met.

[0045] Through the above iterations, the timing mismatch problem caused by prediction bias can be effectively eliminated, and the robustness of the final solution can be improved.

[0046] Example 2: Short-term traffic flow prediction and signal timing optimization system at intersections This embodiment provides a system corresponding to the above method, including the following modules: Target selection module: Receives user-specified predicted intersections Together with its upstream intersections (Level 1, Level 2, etc.), they form a set of intersections to be tested.

[0047] Data statistics module: Acquires signal control schemes, electronic police vehicle passage data, real-time queue lengths for each turn, channelization information, average speed of road segments, number of lanes, and Internet traffic data for each intersection to be tested according to a predetermined time period (e.g., every 2 minutes).

[0048] Calculation module: Executes steps S3 and S4 above to realize multi-level source-tracing traffic flow calculation, steering weight matrix construction and steering flow prediction, and phase duration calculation. The calculation module is further configured as follows: The travel time from the upstream intersection to the predicted intersection is dynamically calculated based on the real-time queue length. The predicted time period of each upstream intersection is determined in reverse. The decision on whether to continue tracing to a higher-level intersection is made by judging the preset inequality conditions. Based on historical data from the same period, a steering weight matrix is ​​constructed and solved. After normalization, combined with the traffic flow of each upstream branch and the current queue length compensation, the predicted traffic flow of each branch at the intersection is calculated.

[0049] Judgment module: Determines whether traffic flow from multiple upstream intersections can merge into the predicted intersection within the predicted time (based on travel time and traffic light waiting time); compares the phase duration difference between the initial scheme in step S2 and the new scheme in step S4. If any phase difference is greater than 1 second, the calculation module is triggered to recalculate using the current new scheme as the initial scheme until convergence, and outputs the final traffic flow prediction and signal timing scheme.

[0050] Example 3: Electronic Equipment The electronic device described in this embodiment includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are connected in sequence. The memory is used to store a computer program, which includes program instructions. The processor is configured to call the program instructions to execute the intersection short-term traffic flow prediction and signal timing optimization method described in Embodiment 1.

[0051] Example 4: Readable storage medium The readable storage medium described in this embodiment stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor performs the intersection short-term traffic flow prediction and signal timing optimization method described in Embodiment 1.

[0052] Technical effect verification The method described in this embodiment was used in a simulation test at an actual intersection in a city. Compared with fixed-period prediction, the method of this invention achieves significant improvements in the following indicators: The mean absolute percentage error (MAPE) of short-term flow forecasts was reduced by approximately 12%. The average vehicle delay per cycle decreased by approximately 8%; The number of phase switching operations is reduced, and the stability of the scheme is improved.

[0053] In addition, since no additional roadside radar or other detection equipment is required, the system construction cost is reduced by about 60% compared to the end-side adaptive solution.

[0054] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for short-term traffic flow prediction and signal timing optimization at intersections, characterized in that, Includes the following steps: S1. Collect basic data required for short-term traffic flow prediction and signal timing at intersections; the basic data includes, but is not limited to, road network information, vehicle data from intersection electronic police or checkpoints, real-time queue lengths of turning points at intersections, channelization information of intersections, timing scheme information for different time periods at intersections, execution information of intersection timing schemes, and real-time internet traffic data of the road network. S2. Obtain a preliminary plan for predicting the next cycle of the intersection and the preliminary prediction time period for each phase using the basic data; S3. Based on steps S1 and S2, determine the traffic flow of each branch of the predicted intersection and each branch of the primary intersection merging into the predicted intersection at the predicted time. S4. Based on step S3, determine the traffic flow and signal duration for the preliminary prediction period of each phase at the intersection; S5. Determine the difference in phase changes between the previous and subsequent signal timing schemes. If the difference in phase duration changes is no greater than 1 second, then determine the final predicted flow rate and signal timing scheme; otherwise, use the timing scheme calculated in step S4 as the new initial scheme and return to step S2 to re-execute.

2. The method for short-term traffic flow prediction and signal timing optimization at intersections according to claim 1, characterized in that, In step S2, the preliminary plan predicts the intersection. No. Predicted start time of each phase Take the period before the end of the current period If the time is less than a second, then the intersection is predicted. No. The prediction termination time for each phase is calculated using the following formula: ； in: To predict intersections No. The predicted termination time of each phase; The first in the preliminary plan Each phase duration, The number of phases; For predicted intersections Each upstream branch Its predicted start time Predict the termination time Take the maximum value among all phases, that is .

3. The method for short-term traffic flow prediction and signal timing optimization at intersections according to claim 1, characterized in that, Step S3 specifically includes the following sub-steps: S31. Calculate the first-level intersection To the predicted intersection exist Travel time at each point in time: ； in: The length of the road segment; This represents the real-time average queue length. For road segment speed; S32. Determine the primary intersection based on the travel time. Predicted start time and predicted termination time : ； ； S33, For Class I intersections If the following inequalities are satisfied simultaneously, then the upstream tracing will not continue; otherwise, the first-level intersection will be used. upstream intersection As a new primary intersection, the same logic continues to be applied in tracing its origins: ； ； ； in: To predict intersections Upstream branch; Class I intersection To the predicted intersection At the predicted start time Travel time; The first in the preliminary plan Each phase duration; The number of phases; Redundancy design duration for predicting computation time before the end of the cycle; upstream intersection Merging into the primary road intersection Real-time average queue length in each direction; Standard vehicle length; This represents the average headway. Class I intersection Branches The moment the green light turns on; Class I intersection The predicted termination time; S34. Predict the total merging traffic flow at each branch of the intersection: ； in: Indicates the predicted intersections within the predicted time period Branches The predicted standard number of vehicles; , , The real-time average queue length (in meters) for left turn (including U-turn), straight, and right turn lanes, respectively. , , These are the dedicated lanes for left turns (including U-turns), straight ahead, and right turns, respectively. upstream intersection Branches Merging into the predicted intersection The traffic flow is obtained by summing the contributions from upstream intersections at each level; For the upstream intersection The total number of branches.

4. The method according to claim 1, characterized in that, The specific steps in step S4 for constructing the steering weight matrix and calculating the steering traffic flow include: S41. Constructing Predictive Intersections upstream intersection Turning weight matrix of traffic flow at each branch intersection ; S42. Obtain historical predicted intersections for the same period in the prediction time. Branches Turning traffic flow data matrix and upstream intersection Traffic flow matrix of each branch ; S43, according to Solve The parameters in the matrix are determined, and the steering weights of each branch are normalized to obtain the normalized steering weight matrix. ; S44, according to Calculate the upstream intersection during the prediction period Merging into the predicted intersection The predicted turning traffic volume is combined with queue length compensation to obtain the predicted intersection. Branches The final predicted traffic flow for left turns (including U-turns), straight ahead, and right turns; the queue length compensation is calculated using the following formula: ； in: , , Forecast intersections within the forecast period Branches Predicted traffic flow for left turns (including U-turns), straight traffic, and right turns; , , These are the real-time average queue lengths for left-turn (including U-turns), straight-ahead, and right-turn lanes, respectively. , , This represents the number of lanes corresponding to the turn. , , From the upstream intersection Predicted traffic volume for merging and turning left (including U-turns), going straight, and turning right; This refers to the standard vehicle length.

5. The method for short-term traffic flow prediction and signal timing optimization at intersections according to claim 4, characterized in that, The method for calculating the predicted duration of each phase at the intersection in step S4 is as follows: If the initial plan adopts a reversing release phase, the phase duration is calculated using the following formula: ； ； ； If the initial plan adopts a one-way release phase, the phase duration is calculated using the following formula: ； in: Forecast intersections within the forecast period Branches Allow the branch of the opposing phase to proceed; , , Forecast intersections within the forecast period Branches Predicted traffic flow for left turns (including U-turns), straight traffic, and right turns; , , Forecast intersections within the forecast period Branches Forecast traffic volume for left turns (including U-turns), straight ahead, and right turns; , , Predicted intersections Branches Number of lanes for left turns (including U-turns), straight ahead, and right turns; , , Predicted intersections Branches Average headway for vehicles in left-turn (including U-turn), straight-ahead, and right-turn lanes; Indicates predicted intersection Branches Single-port release Phase duration; , , Predicted intersections Branches The duration of the left-turn (including U-turn), straight-ahead, and right-turn traffic flow phases; This indicates the operation of retrieving the maximum value.

6. The method for short-term traffic flow prediction and signal timing optimization at intersections according to claim 1, characterized in that, In step S5, the timing scheme calculated in step S4 is used as the new initial scheme, and the process returns to step S2 for re-execution. Specifically, this includes using the duration of each phase calculated in step S4 as the new... The preliminary plan for the next cycle of the intersection and the preliminary prediction time period for each phase are redefined, and steps S3 to S5 are repeated sequentially until the difference in duration of each phase is no greater than 1 second.

7. A short-term traffic flow prediction and signal timing optimization system for intersections, used to implement the short-term traffic flow prediction and signal timing optimization method for intersections as described in any one of claims 1-6, characterized in that, include: The target selection module is used to select the predicted intersection and its upstream intersection, and to designate the predicted intersection and its upstream intersection as the intersection to be tested; The data statistics module is used to acquire signal control schemes, vehicle passage data, real-time queue lengths of each branch turning, channelization information, average speed of the road segment, number of different turning lanes, and Internet traffic data for the intersection under test and its upstream intersections according to a predetermined time period. The calculation module is used to predict the traffic flow of each branch of the intersection and each branch merging into the predicted intersection at the predicted time step S3, based on the signal control scheme parameters of the predicted intersection and its upstream intersection, the turning traffic flow of each branch, the queue length, the average speed of the road segment and the lane data. And in step S4, the signal timing scheme is calculated based on the predicted branch turning traffic flow at the intersection during the predicted time period; The judgment module is used to determine whether traffic flow from multiple upstream intersections can merge into the predicted intersection at the predicted time. It makes the judgment and outputs the judgment based on the travel time of the road segment from the upstream intersection to the predicted intersection and the waiting time at the traffic light at the upstream intersection. It is also used to compare each phase of the preceding and following signal timing schemes, determine the final signal timing scheme of the predicted intersection, and output it. The preceding signal timing scheme is the initial scheme obtained in step S2, and the following signal timing scheme is the scheme calculated in step S4.

8. The intersection short-term traffic flow prediction and signal timing optimization system according to claim 7, characterized in that, The calculation module is further configured to: dynamically calculate the travel time from the upstream intersection to the predicted intersection based on the real-time queue length, reversely determine the predicted time period of each upstream intersection, and decide whether to continue tracing to a higher-level intersection by judging the preset inequality conditions.

9. The intersection short-term traffic flow prediction and signal timing optimization system according to claim 7, characterized in that, The calculation module is further configured to: construct and solve the turning weight matrix based on historical data from the same period, and after normalization, combine the traffic flow of each branch upstream and the current queue length compensation to calculate the predicted traffic flow of each branch at the intersection.

10. The intersection short-term traffic flow prediction and signal timing optimization system according to claim 7, characterized in that, The judgment module is further configured to: compare the difference between the duration of each phase in the initial scheme obtained in step S2 and the duration of each phase in the scheme calculated in step S4; if the difference of any phase duration is greater than 1 second, the calculation module is triggered to re-predict and recalculate the timing scheme using the currently calculated timing scheme as the new initial scheme, until the difference of all phase durations is no greater than 1 second, and output the final traffic prediction and signal timing scheme.

Images