Holographic perception-based single intersection signal lamp dynamic timing method and system

Abstract

The present application relates to the technical field of intelligent traffic control, and more particularly to a single intersection signal lamp dynamic timing method and system based on holographic perception, comprising: real-time collection of single intersection and surrounding traffic participant information through multi-source holographic perception equipment, preprocessing to obtain standardized traffic data set; based on the data set, combining vehicle, pedestrian and non-motor vehicle demand weight, constructing a traffic demand evaluation model, calculating the demand priority coefficient of each entrance; according to the priority coefficient, combining the intersection traffic capacity constraint condition, taking the minimization of the total vehicle waiting time at the intersection, the minimization of the pedestrian crossing waiting time and the maximization of the intersection unit time traffic volume as the target, constructing a signal lamp dynamic timing optimization objective function, solving to obtain the optimal signal lamp timing parameters and issuing to the control terminal; real-time collection of intersection traffic state feedback data, and if the change exceeds the preset threshold, re-timing optimization. It can reduce waiting time, improve intersection traffic efficiency and relieve traffic congestion.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent traffic control, and in particular to a method and system for dynamic timing of traffic lights at a single intersection based on holographic perception. Background Technology

[0002] With the acceleration of urbanization, urban traffic flow continues to rise, and traffic congestion has become a key issue restricting the efficient operation of cities and affecting residents' travel experience. Traditional traffic signal control systems mostly adopt a timed control mode. Their core flaw is that they operate solely based on preset fixed timing schemes and cannot dynamically respond to real-time changes in traffic flow. At the same time, they lack consideration for multiple traffic participants such as pedestrians and non-motorized vehicles, resulting in excessively long waiting times for vehicles at intersections, low traffic efficiency, and exacerbating urban traffic congestion.

[0003] While some existing adaptive signal control schemes attempt to adjust timing by combining traffic flow data, most schemes rely on a single sensing device to collect data, resulting in limited sensing range and single data dimension, leading to insufficient timing optimization accuracy. In addition, the control logic of existing technologies is mostly based on simple traffic flow threshold judgment, without constructing a dynamic mapping model between multi-dimensional traffic demand and signal timing, making it difficult to achieve global optimization of intersection traffic efficiency. Summary of the Invention

[0004] This invention provides a method and system for dynamic timing of traffic lights at a single intersection based on holographic perception, which can dynamically respond to real-time traffic changes, reduce vehicle waiting time, and improve intersection traffic efficiency, effectively solving the problems in the background art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for dynamic timing of traffic lights at a single intersection based on holographic perception, comprising:

[0006] The system uses multi-source holographic sensing equipment to collect real-time information on vehicles, pedestrians, and non-motorized vehicles within a single intersection and its surrounding preset range.

[0007] The collected vehicle, pedestrian, and non-motorized vehicle information is cleaned, fused, and standardized to obtain a standardized traffic dataset.

[0008] Based on a standardized traffic dataset, and taking into account the weights of vehicle demand, pedestrian demand, and non-motorized vehicle demand, a single-intersection traffic demand assessment model is constructed to calculate the traffic demand priority coefficients of each approach lane at the intersection.

[0009] Based on the traffic demand priority coefficient and the intersection capacity constraints, a dynamic traffic light timing optimization objective function is constructed with the optimization objectives of minimizing the total vehicle waiting time, minimizing the pedestrian crossing waiting time, and maximizing the intersection's throughput per unit time.

[0010] An improved particle swarm optimization algorithm is used to solve the objective function to obtain the optimal signal timing parameters for each approach lane under the current traffic conditions.

[0011] The optimal traffic light timing parameters obtained from the solution are sent to the traffic light control terminal at a single intersection to control the traffic lights to operate according to the timing parameters. The traffic status feedback data of the intersection is collected in real time. If the feedback data shows that the traffic status has changed above a preset threshold, the timing optimization method is re-executed.

[0012] In conjunction with the first aspect, in one possible design, vehicle information includes traffic flow, vehicle type, vehicle speed, and vehicle queue length; pedestrian information includes pedestrian flow and pedestrian waiting time to cross the street; and non-motorized vehicle information includes non-motorized vehicle flow and non-motorized vehicle queue length.

[0013] In conjunction with the first aspect, in one possible design, the multi-source holographic sensing device includes a high-definition camera, millimeter-wave radar, and lidar. The high-definition camera is used to collect data on vehicle type, pedestrian flow, and non-motorized vehicle flow. The millimeter-wave radar is used to collect data on vehicle speed, vehicle queue length, and non-motorized vehicle queue length. The lidar is used to assist in correcting vehicle position and pedestrian crossing trajectory data.

[0014] In conjunction with the first aspect, in one possible design, the vehicle demand weight is adjusted by the road resource occupancy rate corresponding to the vehicle type, the pedestrian demand weight is adjusted by the pedestrian crossing waiting time, and the non-motorized vehicle demand weight is adjusted by the non-motorized vehicle queue length.

[0015] In conjunction with the first aspect, in one possible design, the method for calculating the traffic demand priority coefficient is as follows:

[0016] Suppose there are n approach lanes at a single intersection. For the i-th approach lane (i = 1, 2, ..., n), its traffic demand priority coefficient P is... i The calculation formula is:

[0017] P i =α·W v,i +β·W p,i +γ·W b,i ;

[0018] Where α is the vehicle demand weighting coefficient, β is the pedestrian demand weighting coefficient, and γ is the non-motorized vehicle demand weighting coefficient, and α + β + γ = 1; W v,iW represents the vehicle demand weight for the i-th entrance lane. p,i W represents the pedestrian demand weight for the i-th approach lane. b,i The non-motorized vehicle demand weight for the i-th entrance lane;

[0019] Vehicle demand weight W v,i The calculation formula is:

[0020]

[0021] Among them, Q v,i Let ω be the real-time traffic flow of the i-th approach lane. v,i Let be the vehicle type correction factor for the i-th import lane. The vehicle type correction factor is determined based on the proportion of large vehicles, medium vehicles, and small vehicles. The correction factor for large vehicles is greater than that for medium vehicles, and the correction factor for medium vehicles is greater than that for small vehicles.

[0022] Pedestrian demand weight W p,i The calculation formula is:

[0023]

[0024] Among them, Q p,i Let τ be the pedestrian flow corresponding to the i-th approach lane. p,i Let be the average pedestrian waiting time for the i-th approach lane;

[0025] Non-motorized vehicle demand weight W b,i The calculation formula is:

[0026]

[0027] Among them, Q b,i Let L be the non-motorized vehicle traffic flow at the i-th approach lane. b,i Let be the queue length of non-motorized vehicles at the i-th entrance lane.

[0028] In conjunction with the first aspect, in one possible design, the traffic capacity constraints at the intersection include:

[0029] The number of vehicles passing through each approach lane per unit time shall not exceed the maximum design capacity of that approach lane;

[0030] The green light duration for each entrance lane shall not be less than the preset minimum green light duration and shall not be greater than the preset maximum green light duration.

[0031] The green light duration for pedestrians crossing the street shall not be less than the minimum duration required for pedestrians to safely cross the street.

[0032] In conjunction with the first aspect, in one possible design, the expression for the objective function is:

[0033] minF=λ1·Tv +λ2·T p -λ3·C;

[0034] Where F is the value of the objective function, and T v T represents the total vehicle waiting time at the intersection. p Let C be the average waiting time for pedestrians to cross the street, and let C be the traffic volume per unit time at the intersection. λ1, λ2, and λ3 are the weighting coefficients for vehicle waiting time, pedestrian waiting time, and traffic volume per unit time, respectively, and satisfy λ1+λ2+λ3=1.

[0035] Total vehicle waiting time T at the intersection v The calculation formula is:

[0036]

[0037] Among them, L v,i Let v be the queue length of vehicles at the i-th entrance lane. v,i Let g be the average speed of vehicles at the i-th entrance lane. i The green light duration for the i-th lane;

[0038] Average pedestrian waiting time T p The calculation formula is:

[0039]

[0040] The formula for calculating the traffic volume C per unit time at an intersection is:

[0041]

[0042] Among them, G i Let G be the signal light cycle duration for the i-th approach lane. i =g i +y i +r i

[0043] y i The duration of the yellow light, r i This refers to the duration of the red light.

[0044] In conjunction with the first aspect, in one possible design, the improvement of the modified particle swarm optimization algorithm lies in:

[0045] A dynamic inertia weight factor is introduced, which decreases linearly with the number of iterations to balance the algorithm's global search capability and local optimization capability. At the same time, a crossover and mutation operator is introduced to perform crossover and mutation operations on particles with low fitness values ​​in the particle swarm to avoid the algorithm getting trapped in local optima.

[0046] In conjunction with the first aspect, in one possible design, the formula for calculating the dynamic inertia weighting factor is:

[0047]

[0048] Where t is the current iteration number, t max ω represents the maximum number of iterations. max For the maximum inertia weight, ω min This represents the minimum inertia weight.

[0049] Secondly, the present invention also provides a method and system for dynamic timing of traffic lights at a single intersection based on holographic perception, comprising:

[0050] The sensing and acquisition module is used to deploy multi-source holographic sensing devices to collect vehicle, pedestrian and non-motorized vehicle information in real time at a single intersection and within a preset range.

[0051] The data preprocessing module receives the raw vehicle, pedestrian and non-motorized vehicle information output by the sensing and acquisition module, and generates a standardized traffic dataset through data cleaning, data fusion and data standardization operations.

[0052] The demand assessment module is used to construct a single-intersection traffic demand assessment model based on a standardized traffic dataset, by comprehensively setting and incorporating vehicle demand weights, pedestrian demand weights, and non-motorized vehicle demand weights. This model is used to calculate the traffic demand priority coefficients for each approach lane of the intersection.

[0053] The timing target construction module is used to construct a dynamic timing optimization objective function for traffic lights based on the traffic demand priority coefficients of each approach lane output by the demand assessment module, combined with the intersection capacity constraints, with the core optimization objectives being to minimize the total vehicle waiting time at the intersection, minimize the pedestrian crossing waiting time, and maximize the traffic volume per unit time at the intersection.

[0054] The objective solution module is used to solve the objective function of dynamic traffic light timing optimization using an improved particle swarm optimization algorithm. Through iterative search of the algorithm, it finds the optimal traffic light timing parameters for each approach lane that meet the constraints and achieve the optimization objective under the current traffic conditions.

[0055] The timing control module is used to send the optimal traffic light timing parameters obtained by the objective solution module to the traffic light control terminal at a single intersection, control the traffic lights to operate according to the timing parameters, collect traffic status feedback data at the intersection in real time, and monitor and analyze the feedback data; if the feedback data shows that the traffic status has changed above a preset threshold, the full-process timing optimization method will be re-executed.

[0056] The technical solution of this invention can achieve the following technical effects:

[0057] By collaborating with multiple devices including high-definition cameras, millimeter-wave radar, and lidar, comprehensive multi-dimensional data on three types of traffic participants—vehicles, pedestrians, and non-motorized vehicles—is achieved. This addresses the limitations of traditional solutions, such as narrow perception range and limited data dimensions, providing a holistic and precise data foundation for subsequent timing optimization. Dynamic adjustments to the demand weights of these three traffic participants are made by introducing vehicle type correction coefficients, pedestrian waiting times, and non-motorized vehicle queue lengths. Weighted calculations yield traffic demand priority coefficients for each approach lane, enabling refined quantification of the demands of different traffic entities. This is achieved by minimizing total vehicle waiting time and minimizing... The optimization objectives are to optimize pedestrian waiting time and maximize the traffic volume per unit time at the intersection. The objective function is constructed by combining rigid constraints such as the maximum capacity of the approach lanes, the upper and lower limits of the green light duration, and the safe pedestrian crossing time. At the same time, an improved particle swarm optimization algorithm with dynamic inertia weights and cross-mutation operators is used to solve the problem. This ensures that the timing scheme can meet the core needs of various traffic participants and adapt to the actual traffic capacity of the intersection. By collecting traffic status feedback data in real time and setting thresholds to trigger re-optimization, a closed-loop control is formed, which can solve the problem that traditional timed control or single adaptive schemes cannot respond to sudden changes in traffic flow. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of the single-intersection traffic light dynamic timing method based on holographic perception in this invention;

[0060] Figure 2 This is a structural diagram of the single-intersection traffic light dynamic timing system based on holographic perception in this invention. Detailed Implementation

[0061] This application will now be described with reference to the accompanying drawings.

[0062] like Figure 1 As shown, the single-intersection traffic light dynamic timing method based on holographic perception of the present invention specifically includes the following steps:

[0063] Step S1: Real-time collection of vehicle information, pedestrian information and non-motorized vehicle information within a single intersection and its surrounding preset range using multi-source holographic sensing equipment;

[0064] Step S2: Perform data cleaning, data fusion, and data standardization on the collected vehicle information, pedestrian information, and non-motorized vehicle information to obtain a standardized traffic dataset;

[0065] Step S3: Based on the standardized traffic dataset, and taking into account the weights of vehicle demand, pedestrian demand, and non-motorized vehicle demand, construct a single-intersection traffic demand assessment model and calculate the traffic demand priority coefficients of each approach lane at the intersection.

[0066] Step S4: Based on the traffic demand priority coefficient and the intersection capacity constraints, construct the traffic light dynamic timing optimization objective function with the optimization objectives of minimizing the total vehicle waiting time, minimizing the pedestrian crossing waiting time, and maximizing the intersection's per unit time throughput.

[0067] Step S5: The improved particle swarm optimization algorithm is used to solve the objective function to obtain the optimal signal timing parameters for each approach lane under the current traffic conditions.

[0068] Step S6: Send the obtained optimal traffic light timing parameters to the single-intersection traffic light control terminal, control the traffic lights to operate according to the timing parameters, and collect traffic status feedback data of the intersection in real time. If the feedback data shows that the traffic status has changed above the preset threshold, the timing optimization method is re-executed.

[0069] In this embodiment, by deploying multi-source holographic sensing equipment, real-time collection of traffic information from all dimensions of vehicles, pedestrians, and non-motorized vehicles at a single intersection can be achieved, providing a more comprehensive data foundation for timing optimization. Through data cleaning, fusion, and standardization, data quality can be improved, avoiding timing deviations caused by insufficient data accuracy. The constructed traffic demand assessment model comprehensively considers the needs of three types of traffic participants and calculates the priority coefficient of the approach lane, taking into account the coordination problem of multiple traffic participants and making the allocation of timing resources more equitable. By constructing an optimization function with multiple objectives and combining it with priority coefficients, an improved particle swarm optimization algorithm is used to efficiently solve for the optimal timing parameters, which can more accurately match real-time traffic demand, reduce vehicle and pedestrian waiting time, and increase intersection throughput. Through real-time feedback and cyclic optimization mechanisms, it can quickly respond to changes in traffic conditions such as tidal traffic flow and temporary congestion, improve adaptability to real-time changes in traffic flow, and further ensure the continuous smooth flow of traffic at the intersection.

[0070] In some embodiments of the present invention, the information generated by vehicles, pedestrians, and non-motorized vehicles within a single intersection and its surrounding preset range can reflect comprehensive data on their traffic behavior, status, and needs. By fully covering all traffic participants at the intersection, it can ensure that the timing scheme can balance the passage needs of various participants. Specifically, vehicle information focuses on various types of motor vehicles passing through the intersection, collecting key data that reflects vehicle passage status and needs, including:

[0071] A1) Traffic flow: The total number of vehicles passing through the stop line of the entrance lane or entering the waiting area of ​​the entrance lane per unit time, which directly reflects the traffic pressure in that direction.

[0072] A2) Vehicle type: Classified by vehicle size, weight and road resource occupancy rate. Different vehicle types have significantly different requirements for lane space and speed. The timing resource allocation logic needs to be corrected by vehicle type data to avoid low traffic efficiency of large vehicles due to the average allocation of timing based on the number of vehicles.

[0073] A3) Vehicle speed: The real-time average speed of motor vehicles in the entrance lane can reflect whether there is hidden congestion in that direction, and traffic flow needs to be eased through timing adjustments.

[0074] A4) Vehicle queue length: The total length of the queue of motor vehicles waiting for red light in the entrance lane directly reflects the urgency of the vehicle waiting demand. The longer the queue, the longer the waiting time for vehicles in that direction. Green light resources should be prioritized in the timing to reduce vehicle delays.

[0075] Pedestrian information is collected from pedestrians crossing the street at crosswalks, gathering data that reflects their crossing needs and waiting status, including:

[0076] B1) Pedestrian flow: The total number of pedestrians arriving at the zebra crossing of the entrance lane and preparing to cross the street per unit time. This can determine the strength of pedestrian demand for crossing the street in that direction. Focusing only on vehicle flow may lead to insufficient green light time for pedestrians crossing the street, causing safety hazards such as pedestrians running red lights. It is necessary to balance the rights of vehicles and pedestrians through pedestrian flow data.

[0077] B2) Pedestrian crossing waiting time: The average time it takes for a pedestrian to wait for the current traffic light cycle to end before crossing the street after arriving at the zebra crossing. If the waiting time is too long, it will reduce the pedestrian's travel experience and even cause traffic disorder. This data needs to be used to trigger the adjustment of pedestrian priority timing to ensure the safety and efficiency of pedestrian crossing.

[0078] Non-motorized vehicle information targets entities that travel using non-motorized modes of transportation such as bicycles, electric bicycles, and electric scooters, collecting data that reflects their traffic status and queuing needs, including:

[0079] C1) Non-motorized vehicle flow: The total number of vehicles entering the non-motorized vehicle waiting area of ​​the entrance lane or driving along the non-motorized vehicle lane towards the intersection within a unit of time. If its flow data is ignored, it is easy for non-motorized vehicles to overflow into the motorized vehicle lane, affecting the overall traffic efficiency.

[0080] C2) Non-motorized vehicle queue length: The total length of the non-motorized vehicle queue waiting for the red light in the approach lane. If the non-motorized vehicle queue is too long, it will occupy the space of the intersection approach lane and reduce the passage space of motor vehicles. This data is needed to adjust the green light duration for non-motorized vehicles to reduce conflicts with motor vehicles and pedestrians.

[0081] By collecting the above information, traffic light timing can be transformed from serving only vehicles to balancing the needs of various stakeholders, ultimately achieving overall efficiency and order in intersection traffic.

[0082] To achieve accurate collection of traffic participant information across all dimensions, multi-source holographic sensing equipment needs to be deployed according to the principle of full coverage of approach lanes and key area monitoring. This equipment may include high-definition cameras, millimeter-wave radar, and lidar. The deployment principle is to place one high-definition camera and one millimeter-wave radar 5-10 meters behind the stop line at each approach lane, with the high-definition camera lens facing the direction of oncoming traffic and the millimeter-wave radar's detection angle covering the entire approach lane. At the center of the intersection, one lidar is deployed at a height of no less than 8 meters, with a horizontal detection angle of 360° to cover all approach lanes, exit lanes, and pedestrian crossing areas.

[0083] In multi-source holographic sensing devices, high-definition cameras are responsible for collecting category recognition information, namely vehicle type, pedestrian flow, and non-motorized vehicle flow. Based on deep learning image recognition algorithms, they can classify and count targets by analyzing the intersection video stream captured by the cameras in real time. The specific content and operations collected are as follows:

[0084] Vehicle type category: The algorithm identifies the outline size and appearance features of vehicles in the video stream, classifies them according to preset rules, and counts the proportion of each vehicle type in the entrance lane;

[0085] Pedestrian traffic: The algorithm identifies pedestrian targets in the video stream, counts pedestrians entering the zebra crossing area, and calculates the pedestrian traffic corresponding to the entrance lane according to the predetermined frequency;

[0086] Non-motorized vehicle traffic flow: The algorithm identifies non-motorized vehicles by their two-wheel structure and open body, counts the number of vehicles entering the non-motorized vehicle waiting area of ​​the entrance lane, and counts the non-motorized vehicle traffic flow according to the pre-determined frequency;

[0087] Millimeter-wave radar is responsible for collecting dynamic parameter information, such as vehicle speed, vehicle queue length, and non-motorized vehicle queue length. Based on the Doppler effect and frequency-modulated continuous wave technology, it transmits high-frequency electromagnetic waves and receives the echo signals reflected from targets to calculate parameters such as target speed, distance, and position. Specific data collection and operation are as follows:

[0088] Vehicle speed: The radar monitors the radial speed of each vehicle in the approach lane in real time. After removing the data of stationary vehicles, the average speed of all vehicles in the approach lane is calculated and updated according to the predetermined frequency.

[0089] Vehicle queue length: The radar calculates the number of vehicles in the queue by measuring the distance between the frontmost vehicle in the queue and the stop line, combined with the average spacing between vehicles in a single lane, and then multiplies it by the average vehicle length to obtain the vehicle queue length, which is updated according to the scheduled frequency.

[0090] Non-motorized vehicle queue length: The radar distinguishes non-motorized vehicle queues by identifying the moving speed and body shape characteristics of non-motorized vehicles, measures the distance between the frontmost non-motorized vehicle and the stop line, and calculates the queue length by combining the average spacing and average length of non-motorized vehicles, and updates it according to the predetermined frequency.

[0091] The lidar is responsible for collecting position correction information, namely, assisting in correcting vehicle positions and pedestrian crossing trajectories. It scans the target area by emitting laser beams to generate high-density point cloud data. Based on point cloud matching and coordinate transformation, it corrects positional deviations of vehicles and pedestrians. The specific data collection and operations are as follows:

[0092] Vehicle position correction: High-definition cameras are prone to errors in judging vehicle position due to perspective. LiDAR obtains the three-dimensional coordinates of the vehicle through point cloud data and matches them with the vehicle position identified by the camera to correct the actual position of the vehicle in the lane, ensuring the accuracy of vehicle spacing in queue length calculation.

[0093] Pedestrian crossing trajectory correction: Millimeter-wave radar has low accuracy in detecting pedestrian positions. LiDAR tracks pedestrian movement trajectories through point clouds and corrects the real-time position of pedestrians during the crossing process, avoiding deviations in pedestrian flow statistics due to misjudgment of position.

[0094] The raw data collected by the above three types of devices can control the timestamp error of all devices to within 10ms through the Network Time Protocol, ensuring that the vehicle and pedestrian data at the same time correspond and match. Through the edge computing unit, the vehicle type / flow data of the high-definition camera, the speed / queue length data of the millimeter-wave radar, and the position correction data of the lidar are associated into a complete traffic participant information record based on the entrance lane number + timestamp.

[0095] In some embodiments of the present invention, the raw data collected by the multi-source holographic sensing device may contain outliers, missing values, and inconsistencies in units. Therefore, the data needs to be preprocessed to obtain a standardized traffic dataset, specifically including:

[0096] Step S21, Data Cleaning: Outliers in vehicle speed, flow rate, and other data can be removed using the 3σ criterion; missing data can be filled in using the mean of adjacent time points to ensure data continuity.

[0097] Step S22, Data Fusion: Based on the Kalman filter algorithm, the same type of data collected by high-definition cameras and millimeter-wave radar, such as traffic flow and queue length, can be fused to improve data accuracy by utilizing the complementarity of the two devices.

[0098] Step S23, Data Standardization: The min-max standardization method can be used to transform feature data of different dimensions such as traffic flow, speed, and queue length to the [0,1] interval. The formula is as follows:

[0099]

[0100] Where x is the original data, x min x is the minimum value of the data in this dimension. max x represents the maximum value of the data in this dimension. norm This is the standardized data.

[0101] In some embodiments of the present invention, a traffic demand priority assessment model is constructed based on preprocessed feature data, and the priority coefficient of each approach lane is calculated to reflect the urgency of traffic demand at different approach lanes, specifically including:

[0102] Step S31: Based on the traffic flow characteristics of the intersection, preset the vehicle demand weight coefficient α, pedestrian demand weight coefficient β, and non-motorized vehicle demand weight coefficient γ, and satisfy α+β+γ=1; for example, β can be increased at intersections in commercial areas and α can be increased at intersections in industrial areas.

[0103] Step S32: Suppose there are n approach lanes at a single intersection. For the i-th approach lane (i = 1, 2, ..., n), considering the difference in road resource usage by vehicle type, a vehicle type correction coefficient ω is introduced. v,i Through formula Calculate vehicle demand weights to ensure that import lanes with a high proportion of large vehicles receive higher vehicle demand weights, where Q v,i Let ω be the real-time traffic flow of the i-th approach lane. v,i Let be the vehicle type correction factor for the i-th import lane. The vehicle type correction factor is determined based on the proportion of large vehicles, medium vehicles, and small vehicles. The correction factor for large vehicles is greater than that for medium vehicles, and the correction factor for medium vehicles is greater than that for small vehicles.

[0104] Combining pedestrian flow and waiting time, using the formula The pedestrian demand weight is calculated, with higher weights for approach lanes where waiting times are longer, to protect pedestrians' right to cross the street. Q is one such weight. p,i Let τ be the pedestrian flow corresponding to the i-th approach lane. p,i Let be the average pedestrian waiting time for the i-th approach lane;

[0105] Combining non-motorized vehicle traffic flow and queue length, through The demand weight for non-motorized vehicles is calculated, and the longer the queue at the entrance, the higher the demand weight for non-motorized vehicles, thus alleviating congestion for non-motorized vehicles. Where Q... b,i Let L be the non-motorized vehicle traffic flow at the i-th approach lane. b,i Let be the queue length of non-motorized vehicles at the i-th entrance lane.

[0106] Step S33, using formula P i =α·W v,i +β·W p,i +γ·W b,i Calculate the traffic demand priority coefficient. The higher the coefficient, the more urgent the traffic demand at that entrance lane is, and the longer the green light duration should be allocated.

[0107] In some embodiments of the present invention, constructing an optimization objective function specifically includes:

[0108] Through formula Calculate the total vehicle waiting time at the intersection, where L v,i Let v be the queue length of vehicles at the i-th entrance lane. v,i Let g be the average speed of vehicles at the i-th entrance lane. i The green light duration for the i-th lane;

[0109] Through formula Calculate the average waiting time for pedestrians to cross the street to ensure a better pedestrian experience;

[0110] The combined traffic volume of vehicles, pedestrians, and non-motorized vehicles is calculated using the formula. Calculate the traffic volume per unit time at the intersection, where G i Let G be the signal light cycle duration for the i-th approach lane. i =g i +y i +r i y i The duration of the yellow light, r i This refers to the duration of the red light.

[0111] The formula minF=λ1·T is used. v +λ2·T p -λ3·C integrates the three sub-objectives into a unified optimization objective function, where λ1, λ2, and λ3 are the weight coefficients of vehicle waiting time, pedestrian waiting time, and traffic volume per unit time, respectively, and satisfy λ1+λ2+λ3=1. It can be dynamically adjusted according to the intersection type, such as increasing λ2 at intersections around schools. Through weight allocation, the optimization focus can be switched in different scenarios.

[0112] In constructing the objective function, it is necessary to combine the intersection capacity constraints, including vehicle traffic volume constraints, green light duration constraints, and pedestrian crossing constraints. The specific constraints are as follows:

[0113] Traffic volume constraints: Traffic volume at each approach lane per unit time The maximum design capacity of the approach lane shall not exceed the maximum design capacity of the approach lane; the maximum design capacity is determined by the road width and the number of lanes to avoid overloading road resources;

[0114] Green light duration constraint: Green light duration g for each entrance lane i The green light duration shall not be less than the preset minimum green light duration and shall not be greater than the preset maximum green light duration. The minimum green light duration shall be set based on the shortest time for vehicles to safely pass through the intersection, and the maximum green light duration shall be set based on the maximum tolerable waiting time of other approach lanes, in order to prevent long-term congestion in one direction.

[0115] Pedestrian crossing constraints: The green light duration for pedestrian crossings shall not be less than the minimum duration required for pedestrians to safely cross the street; the safe crossing time for pedestrians shall be calculated based on the intersection width and the average walking speed of pedestrians to ensure pedestrian safety.

[0116] In some embodiments of the present invention, an improved particle swarm optimization algorithm is used to solve the objective function to obtain the optimal green light duration g for each entrance lane. i Yellow light duration y i , is the red light duration r i The improvement of this algorithm lies in:

[0117] pass Introduce an inertia weight that decreases linearly with the number of iterations, where t is the current iteration number. max ω represents the maximum number of iterations. max For the maximum inertia weight, ω min The minimum inertia weight; in the early stage of iteration, ω(t) is close to ω max For example, 0.9 can enhance the global search capability to explore a wider solution space; in the later stages of iteration, ω(t) approaches ω min For example, 0.4 can enhance the local optimization capability to accurately locate the optimal solution;

[0118] For the bottom 30% of the particles in each generation of the particle swarm, i.e. the objective function value F, single-point crossover and Gaussian mutation operations can be used. That is, the crossover operation randomly selects two particles to exchange genes in some dimensions, and the mutation operation slightly perturbs the value of a certain dimension of the particle to avoid the algorithm getting trapped in local optima.

[0119] The specific solution process of this algorithm includes:

[0120] Step S51: Set the particle population size and the maximum number of iterations t. maxAnd the particle dimension, randomly generate an initial particle swarm; the particle swarm size can be set to 50, and the maximum number of iterations t max It can be set to 100; the particle dimension is equal to the number of inlet lanes, and each dimension corresponds to the green light duration of one inlet lane; the green light duration of each particle in the randomly generated initial particle swarm is between the minimum and maximum green light duration.

[0121] Step S52: Substitute the green light duration corresponding to each particle into the optimization objective function and calculate its fitness value F;

[0122] Step S53: Based on the solution corresponding to the particle's own historical best fitness and the solution corresponding to the population's historical best fitness, and combined with the dynamic inertia weight ω(t), update the particle's velocity and position:

[0123] v k+1 =ω(t)·v k +c1·r1·(p best -x k )+c2·r2·(g best -x k );

[0124] x k+1 =x k +v k+1 ;

[0125] Among them, v k x k Here, c1 and c2 are the particle velocity and position in the kth iteration, respectively; c1 and c2 are learning factors, defaulted to 2; r1 and r2 are random numbers between [0,1], and p... best For the individual optimal solution, g best This is the globally optimal solution;

[0126] Step S54: Perform crossover and mutation operations on particles with low fitness values ​​to update the particle population;

[0127] Step S55: If the number of iterations reaches t max If the global optimal solution shows no improvement after 10 consecutive iterations, then stop iterating and output the green light duration g for each entrance lane corresponding to the global optimal solution. i And according to the preset y i G is dynamically set according to the traffic flow density at the intersection. i Calculate the red light duration r i =

[0128] G i -g i -y i .

[0129] By sending the optimal timing parameters obtained from the solution to the traffic light control terminal at a single intersection, the control terminal can drive the traffic lights to switch according to the timing parameters. Due to the dynamic fluctuations in traffic flow at intersections, such as sudden vehicle cutting in, short-term pedestrian peaks, and temporary delivery vehicle stops, even the optimal timing parameters obtained through previous optimization may no longer be suitable for the current traffic conditions over time. Therefore, it is necessary to ensure that the traffic light timing always "resonates" with traffic demand through real-time monitoring and threshold judgment. By collecting traffic state data after the timing is implemented in real time, dynamic fluctuations can be quickly captured. If the queue length of vehicles at a certain approach lane increases by more than a preset threshold compared to before the timing was implemented, or the pedestrian waiting time exceeds a preset tolerance value, it is determined that the traffic state has changed significantly, triggering a new round of timing optimization process. If the traffic state is stable and the rate of change is less than the threshold, the current timing parameters are maintained until the next preset optimization cycle is automatically triggered for optimization. Through the dual mechanism of threshold judgment and periodic triggering, it is possible to avoid traffic light switching disorder caused by frequent optimization triggers, ensure the predictability of traffic lights for drivers and pedestrians, and achieve closed-loop control with real-time adaptation and continuous optimization.

[0130] like Figure 2 As shown, the present invention also provides a dynamic timing system for traffic lights at a single intersection based on holographic perception, specifically including the following modules:

[0131] The sensing and acquisition module is used to deploy multi-source holographic sensing devices to collect vehicle, pedestrian and non-motorized vehicle information in real time at a single intersection and within a preset range.

[0132] The data preprocessing module receives the raw vehicle, pedestrian and non-motorized vehicle information output by the sensing and acquisition module, and generates a standardized traffic dataset through data cleaning, data fusion and data standardization operations.

[0133] The demand assessment module is used to construct a single-intersection traffic demand assessment model based on a standardized traffic dataset, by comprehensively setting and incorporating vehicle demand weights, pedestrian demand weights, and non-motorized vehicle demand weights. This model is used to calculate the traffic demand priority coefficients for each approach lane of the intersection.

[0134] The timing target construction module is used to construct a dynamic timing optimization objective function for traffic lights based on the traffic demand priority coefficients of each approach lane output by the demand assessment module, combined with the intersection capacity constraints, with the core optimization objectives being to minimize the total vehicle waiting time at the intersection, minimize the pedestrian crossing waiting time, and maximize the traffic volume per unit time at the intersection.

[0135] The objective solution module is used to solve the objective function of dynamic traffic light timing optimization using an improved particle swarm optimization algorithm. Through iterative search of the algorithm, it finds the optimal traffic light timing parameters for each approach lane that meet the constraints and achieve the optimization objective under the current traffic conditions.

[0136] The timing control module is used to send the optimal traffic light timing parameters obtained by the objective solution module to the traffic light control terminal at a single intersection, control the traffic lights to operate according to the timing parameters, collect traffic status feedback data at the intersection in real time, and monitor and analyze the feedback data; if the feedback data shows that the traffic status has changed above a preset threshold, the full-process timing optimization method will be re-executed.

[0137] In this embodiment, by deploying multi-source holographic equipment through the perception and acquisition module, real-time acquisition of information from all dimensions of vehicles, pedestrians, and non-motorized vehicles can be achieved. Combined with the data preprocessing module's cleaning, fusion, and standardization operations, abnormal data is effectively eliminated, complementary information from multiple devices is integrated, and a unified dimension is established. The demand assessment module constructs an assessment model by integrating the demand weights of the three types of traffic participants. By calculating the traffic demand priority coefficient of the approach lanes, the urgency of passage at different approach lanes is accurately quantified, avoiding the drawbacks of indiscriminate resource allocation. The timing target construction module combines this priority coefficient with intersection capacity constraints to construct an optimization function with multiple objectives: reducing vehicle and pedestrian waiting time and increasing traffic volume. This ensures that the timing targets are closely aligned with the actual traffic needs at the intersection, rather than solely pursuing the efficiency of a particular type of participant. The target solution module employs an improved particle swarm optimization algorithm, which balances global and local optimization capabilities through dynamic inertial weights and avoids getting trapped in local optima through cross-mutation operators. This allows it to find the optimal timing parameters that fit the current traffic conditions more quickly, improving the efficiency of timing scheme generation. The timing control module not only accurately issues timing parameters to control the operation of traffic lights, but also triggers a full-process re-optimization when the state changes exceed a threshold by monitoring traffic status feedback data in real time. This allows it to flexibly respond to dynamic scenarios such as tidal traffic flow and temporary congestion, continuously ensuring the efficiency, order, and fairness of traffic at the intersection.

[0138] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic timing of traffic lights at a single intersection based on holographic perception, characterized in that, include: The system uses multi-source holographic sensing equipment to collect real-time information on vehicles, pedestrians, and non-motorized vehicles within a single intersection and its surrounding preset range. The collected vehicle, pedestrian, and non-motorized vehicle information is cleaned, fused, and standardized to obtain a standardized traffic dataset. Based on a standardized traffic dataset, and taking into account the weights of vehicle demand, pedestrian demand, and non-motorized vehicle demand, a single-intersection traffic demand assessment model is constructed to calculate the traffic demand priority coefficients of each approach lane at the intersection. Based on the traffic demand priority coefficient and the intersection capacity constraints, a dynamic traffic light timing optimization objective function is constructed with the optimization objectives of minimizing the total vehicle waiting time, minimizing the pedestrian crossing waiting time, and maximizing the intersection's throughput per unit time. An improved particle swarm optimization algorithm is used to solve the optimization objective function to obtain the optimal signal timing parameters for each approach lane under the current traffic conditions. The optimal traffic light timing parameters obtained from the solution are sent to the traffic light control terminal at a single intersection. The traffic lights are controlled to operate according to the timing parameters, and traffic status feedback data at the intersection is collected in real time. If the feedback data shows that the traffic status has changed above a preset threshold, the timing optimization method is re-executed. Vehicle demand weight is adjusted by the road resource occupancy rate corresponding to vehicle type; pedestrian demand weight is adjusted by pedestrian crossing waiting time; and non-motorized vehicle demand weight is adjusted by non-motorized vehicle queue length. The method for calculating the traffic demand priority coefficient is as follows: Suppose there are n approach lanes at a single intersection. For the i-th approach lane (i=1,2,...,n), its traffic demand priority coefficient is... The calculation formula is: ； in, This is the weighting coefficient for vehicle demand. This is the pedestrian demand weighting coefficient. This is the weighting coefficient for non-motorized vehicle demand, and it satisfies... + + =1; Let i be the vehicle demand weight for the i-th entrance lane. Let i be the pedestrian demand weight for the i-th approach lane. The non-motorized vehicle demand weight for the i-th entrance lane; Vehicle demand weight The calculation formula is: ； in, Let i be the real-time traffic flow of the i-th entrance lane. The vehicle type correction coefficient is defined for the i-th inlet lane. The vehicle type correction coefficient is determined based on the proportion of large vehicles, medium vehicles, and small vehicles. The correction coefficient for large vehicles is greater than that for medium vehicles, and the correction coefficient for medium vehicles is greater than that for small vehicles. pedestrian demand weight The calculation formula is: ； in, Let i be the pedestrian flow corresponding to the i-th entrance. Let be the average pedestrian waiting time for the i-th approach lane; The non-motorized vehicle demand weight The calculation formula is: ； in, Let i be the non-motorized vehicle traffic flow at the i-th approach lane. Let be the queue length of non-motorized vehicles at the i-th entrance lane; The expression for the optimization objective function is: ； in, To optimize the objective function value, This represents the total vehicle waiting time at the intersection. The average waiting time for pedestrians to cross the street. Traffic volume per unit time at the intersection; These are the weighting coefficients for vehicle waiting time, pedestrian waiting time, and traffic volume per unit time, respectively, and they satisfy the following conditions: ; Total vehicle waiting time at the intersection The calculation formula is: ； in, Let be the queue length of vehicles at the i-th entrance lane. Let be the average speed of vehicles at the i-th entrance lane. The green light duration for the i-th lane; The average waiting time for pedestrians to cross the street The calculation formula is: ； The formula for calculating the traffic volume C per unit time at the intersection is: ； in, Let be the signal light cycle duration for the i-th approach lane. , Yellow light duration This refers to the duration of the red light.

2. The method for dynamic timing of traffic lights at a single intersection based on holographic perception according to claim 1, characterized in that, The vehicle information includes traffic flow, vehicle type, vehicle speed, and vehicle queue length; the pedestrian information includes pedestrian flow and pedestrian waiting time to cross the street; and the non-motorized vehicle information includes non-motorized vehicle flow and non-motorized vehicle queue length.

3. The method for dynamic timing of traffic lights at a single intersection based on holographic perception according to claim 1, characterized in that, The multi-source holographic sensing device includes a high-definition camera, millimeter-wave radar, and lidar. The high-definition camera is used to collect vehicle type, pedestrian flow, and non-motorized vehicle flow data. The millimeter-wave radar is used to collect vehicle speed, vehicle queue length, and non-motorized vehicle queue length data. The lidar is used to assist in correcting vehicle position and pedestrian crossing trajectory data.

4. The method for dynamic timing of traffic lights at a single intersection based on holographic perception according to claim 1, characterized in that, The intersection capacity constraints include: The number of vehicles passing through each approach lane per unit time shall not exceed the maximum design capacity of that approach lane; The green light duration for each entrance lane shall not be less than the preset minimum green light duration and shall not be greater than the preset maximum green light duration. The green light duration for pedestrians crossing the street shall not be less than the minimum duration required for pedestrians to safely cross the street.

5. The method for dynamic timing of traffic lights at a single intersection based on holographic perception according to claim 1, characterized in that, The improvement of the improved particle swarm optimization algorithm lies in: A dynamic inertia weight factor is introduced, which decreases linearly with the number of iterations to balance the algorithm's global search capability and local optimization capability. At the same time, a crossover and mutation operator is introduced to perform crossover and mutation operations on particles with low fitness values ​​in the particle swarm to avoid the algorithm getting trapped in local optima.

6. The method for dynamic timing of traffic lights at a single intersection based on holographic perception according to claim 5, characterized in that, The formula for calculating the dynamic inertia weight factor is as follows: ； Where t is the current iteration number, The maximum number of iterations, For maximum inertia weight, This represents the minimum inertia weight.

7. A single-intersection traffic light dynamic timing system based on holographic perception, used to execute the method of claim 1, characterized in that, include: The sensing and acquisition module is used to deploy multi-source holographic sensing devices to collect vehicle, pedestrian and non-motorized vehicle information in real time at a single intersection and within a preset range. The data preprocessing module receives the raw vehicle, pedestrian and non-motorized vehicle information output by the sensing and acquisition module, and generates a standardized traffic dataset through data cleaning, data fusion and data standardization operations. The demand assessment module is used to construct a single-intersection traffic demand assessment model based on a standardized traffic dataset, by comprehensively setting and incorporating vehicle demand weights, pedestrian demand weights, and non-motorized vehicle demand weights. This model is used to calculate the traffic demand priority coefficients for each approach lane of the intersection. The timing target construction module is used to construct a dynamic timing optimization objective function for traffic lights based on the traffic demand priority coefficients of each approach lane output by the demand assessment module, combined with the intersection capacity constraints, with the core optimization objectives being to minimize the total vehicle waiting time at the intersection, minimize the pedestrian crossing waiting time, and maximize the traffic volume per unit time at the intersection. The objective solution module is used to solve the objective function of dynamic traffic light timing optimization using an improved particle swarm optimization algorithm. Through iterative search of the algorithm, it finds the optimal traffic light timing parameters for each approach lane that meet the constraints and achieve the optimization objective under the current traffic conditions. The timing control module is used to send the optimal traffic light timing parameters obtained by the objective solution module to the traffic light control terminal at a single intersection, control the traffic lights to operate according to the timing parameters, collect traffic status feedback data at the intersection in real time, and monitor and analyze the feedback data; if the feedback data shows that the traffic status has changed above a preset threshold, the full-process timing optimization method will be re-executed.

Images