Urban traffic optimization and planning method and system based on artificial intelligence data processing

Abstract

The invention relates to the technical field of traffic planning, and discloses an urban traffic optimization and planning method and system based on artificial intelligence data processing, and the method comprises the steps: collecting a time sequence perception data flow of a target intersection, and carrying out the extraction of an occupancy state tensor based on a time window, and obtaining the occupancy state tensor of the target intersection in the time window; using the occupancy mode analysis model to output occupancy mode probability distribution of the target intersection in the time window; and outputting the lane traffic capacity of the target intersection by using a lane traffic capacity calculation function which is corrected by introducing dynamic factors of a temporary parking type occupation reduction factor and a queuing type occupation delay factor, and dynamically adjusting a short-time traffic control strategy of the target intersection. According to the invention, through time sequence perception and occupation mode analysis, the temporary parking and queuing occupation modes are distinguished, and the traffic capacity of the lane is dynamically quantified and corrected, so that the short-time traffic control strategy is adjusted in real time, and refined and adaptive traffic control and optimization under complex disturbance conditions in a city are realized.

Description

Technical Field

[0001] This invention relates to the field of traffic planning through big data processing, and more particularly to a method and system for urban traffic optimization and planning based on artificial intelligence data processing. Background Technology

[0002] With the accelerating pace of urbanization and the continuous increase in the number of motor vehicles, the operational pressure on urban road traffic systems has significantly increased, with various urban intersections gradually evolving into key bottlenecks in traffic operation. Especially during morning and evening rush hours, traffic flow is concentrated and overlapping, easily leading to problems such as queue overflow and decreased traffic efficiency at intersections. Simultaneously, in daily operation, intersections are also commonly affected by a variety of unplanned factors, including sudden traffic incidents, temporary vehicle stops for loading and unloading, right-turning or U-turning vehicles yielding to pedestrians, and the mixing of non-motorized vehicles and pedestrians. This results in intersection traffic conditions exhibiting significant time-varying, random, and uncertain characteristics. These complex disturbances are often characterized by their suddenness, short duration, but significant impact, posing considerable challenges to traffic management and control. Furthermore, traditional traffic models and signal timing methods are mostly based on the assumption of stable flow, neglecting the impact of these unplanned occupancy events on intersection capacity. During peak hours, delivery riders and ride-hailing vehicles often temporarily park in curb spaces and intersection entrances. This short-term, frequent occupancy greatly affects the function of non-motorized vehicle lanes and right-turn lanes, rendering existing optimization methods ineffective.

[0003] In existing research, CN107563543B proposed a swarm intelligence-based urban traffic optimization service method and system. This method simulates local traffic conditions using swarm intelligence algorithms and constructs local scheduling strategies based on this simulation. Furthermore, it combines the simulation results of the global traffic situation to form a scheduling strategy that balances local and global factors, ultimately providing optimal route recommendations for vehicles and achieving macro-level traffic optimization. While this approach has certain advantages in route planning and collaboration among traffic participants, its core focus remains on vehicle-level travel routes and scheduling decisions. It primarily targets network-level or regional-level traffic situation optimization, lacking sufficient characterization of the micro-operational mechanisms at intersections. It struggles to distinguish between queuing congestion and unplanned temporary stops such as parking, loading / unloading, and right turns to avoid obstacles. The optimization results are mostly reflected at the route recommendation level and cannot be directly applied to intersection-level real-time control strategies such as signal control, resulting in limited responsiveness to sudden and localized traffic disturbances.

[0004] To address this problem, this invention proposes a method and system for urban traffic optimization and planning based on artificial intelligence data processing. This method dynamically optimizes and adjusts short-term traffic control strategies at intersections, improves the smoothness of urban traffic, and provides a basis for the optimization of future intelligent transportation systems. Summary of the Invention

[0005] This invention provides a method and system for urban traffic optimization and planning based on artificial intelligence data processing. Traditional traffic systems often rely on single-frame detection or statistics for control, which fails to reflect the temporal evolution of traffic conditions, resulting in a delayed response to short-term traffic fluctuations. Step S1 converts the temporal sensing data stream into a time-window-level occupancy state tensor, achieving a unified structured representation of the target intersection's traffic conditions across multiple time scales, thus solving the problem that raw sensing data cannot be directly used in intelligent decision-making. Step S2 introduces an occupancy pattern analysis model, outputting a time-window-level probability distribution of occupancy patterns, enabling a probabilistic distinction between temporary parking occupancy patterns and queuing occupancy patterns. To avoid misjudgment of control strategies due to local abnormal behavior, traditional lane capacity models are mostly based on ideal conditions or long-term statistical parameters, which cannot reflect the immediate reduction of capacity or queue expansion effect of temporary stopping events. Step S3 introduces temporary stopping occupancy reduction factors and queue occupancy delay factors to dynamically correct lane capacity, solving the problems of distorted capacity estimation and unreliable control basis in unplanned temporary stopping scenarios. Step S4 generates short-term traffic control strategies in real time based on the dynamically corrected lane capacity. The generated short-term traffic control strategies can proactively adapt to temporary stopping disturbances and queue evolution, avoiding ineffective green lights and the spread of local congestion.

[0006] To achieve the above objectives, this invention provides a method for urban traffic optimization and planning based on artificial intelligence data processing, comprising the following steps: S1: Collect the time-series sensing data stream of the target intersection, and extract the occupancy state tensor based on the time window from the time-series sensing data stream to obtain the occupancy state tensor of the target intersection in the time window. S2: Based on the occupancy state tensor, use the occupancy pattern analysis model to output the probability distribution of the occupancy pattern of the target intersection within the time window; S3: Based on the probability distribution of the occupancy pattern, the lane capacity calculation function is modified by introducing a temporary parking occupancy reduction factor and a queuing occupancy delay factor to output the lane capacity of the target intersection under the influence of unplanned temporary parking. S4: Based on the lane capacity of the target intersection, dynamically adjust the short-term traffic control strategy of the target intersection, and optimize the traffic at the target intersection according to the short-term traffic control strategy.

[0007] As a further improvement of the present invention: Further, the acquisition of the time-series sensing data stream of the target intersection in step S1 includes: S11: Deploy multiple traffic sensing devices at the lane positions of the target intersection. Each traffic sensing device is equipped with a unified clock to collect lane sensing data synchronized with the time. The lane sensing data includes the data collection time, the number of vehicles in the lane, and the average vehicle speed. S12: Construct time windows with consistent and non-overlapping time ranges. Based on the data acquisition time, insert the number of vehicles in the lane and the average vehicle speed in the lane perception data into the corresponding time window to obtain a time-series perception data stream.

[0008] Furthermore, step S1, which involves extracting the occupancy state tensor of the time-series-aware data stream based on a time window, also includes: S13: Based on the time-series sensing data stream, extract the maximum number of vehicles in the lane from all lane sensing data within the time window as the vehicle occupancy intensity of the target intersection within the time window. S14: Extract the mean of the average vehicle speed in all lane perception data within the time window, and perform reverse normalization on the mean of the average vehicle speed to serve as the lane speed occupancy intensity of the target intersection within the time window. Specifically, the formula for inversely normalizing the mean of the vehicle's average speed is as follows: ; in, This represents the average vehicle speed across all lane perception data within the time window. express The result of reverse normalization, Indicates the vehicle's reference speed; S15: The vehicle quantity occupancy intensity and lane speed occupancy intensity of the target intersection in the time window are concatenated to form the occupancy feature vector of the target intersection in the time window; S16: Based on the occupancy feature vector, a nonlinear cooperative occupancy state function is used to perform a nonlinear mapping on the vehicle quantity occupancy intensity and lane speed occupancy intensity to obtain the occupancy state of the target intersection in the time window; Specifically, the formula for nonlinear mapping using the nonlinear cooperative occupancy state function is as follows: ; This indicates the intensity of vehicle occupancy. Indicates lane occupancy intensity by speed. Indicates the intensity of vehicle occupancy. and lane speed occupancy intensity The occupancy status obtained by nonlinear mapping This represents the speed suppression amplification index, used to enhance the impact of low-speed conditions on occupancy. S17: Based on the occupancy status of the target intersection in the time window, determine the consistency of the occupancy status of adjacent time windows, and merge adjacent time windows that are determined to have the same occupancy status. S18: Calculate the occupancy statistics of the merged time window as the occupancy status tensor of the target intersection in the time window. The occupancy statistics include the time range length of the time window, the mean of the occupancy status, the standard deviation of the occupancy status, and the mean of the occupancy feature vector.

[0009] Furthermore, step S2, which uses the occupancy pattern analysis model to output the probability distribution of the occupancy pattern of the target intersection within a time window, also includes: S21: The occupancy pattern analysis model includes a feature embedding layer, a time-series awareness layer, and a pattern discrimination layer; S22: The feature embedding layer takes the occupancy state tensor of the target intersection in the time window as input, and uses a nonlinear mapping method based on the time range length embedding to map the occupancy state tensor to obtain a low-dimensional occupancy embedding vector. Specifically, the mapping formula for the occupancy state tensor is: ; in, This represents the occupancy state tensor of the target intersection in the r-th time window. This indicates the number of time windows after merging. Represents the occupied state tensor The low-dimensional occupancy embedding vector obtained by mapping Represents the occupied state tensor The length of the time range in the text. Indicates the length of the time range The embedded value, This is the Hadamard product operator, which means multiplying elements at corresponding positions. Represents the logarithmic function. This represents the parameters of the convolution matrix in the feature embedding layer. This represents the bias parameter in the feature embedding layer. This represents a non-linear activation function, set to the ReLU function. Represents the parameters of the temporal embedding matrix, where and The dimensions are consistent; S23: The time-series awareness layer converts the standard deviation of the occupancy state into a fluctuation gating coefficient, and uses a recursive update method based on fluctuation gating to update the state of the low-dimensional occupancy embedding vector under time-series awareness, thereby obtaining the time-series awareness vector. S24: The pattern discrimination layer is a fully connected layer structure. The pattern discrimination layer converts the time-series perception vector into the probability of temporary parking and queuing occupancy modes of the target intersection in the time window, which are used as the probability distribution of the occupancy modes of the target intersection in the time window.

[0010] Specifically, the conversion formula for the pattern discrimination layer to transform the time-series perception vector into the probability of temporary parking occupancy mode and the probability of queuing occupancy mode at the target intersection within the time window is as follows: ; ; ; in, Let represent the probability distribution of occupancy patterns at the target intersection in the r-th time window, where These represent the probabilities of the target intersection in the temporary parking mode and the queuing mode in the r-th time window, respectively. The probabilities of the temporary parking mode and the queuing mode represent the probabilities that the target intersection is in the temporary parking mode or the queuing mode in the time window, respectively. This represents an exponential function with the natural constant as its base. The parameters of the convolution matrix represent the temporary occupancy mode. The bias parameter represents the temporary occupancy mode. The parameters of the convolution matrix represent the queuing occupancy pattern. The bias parameter represents the queuing occupancy mode.

[0011] Furthermore, in step S23, the occupancy state standard deviation is converted into fluctuation gating coefficients. The update formula for the time-aware state update of the low-dimensional occupancy embedding vector using a recursive update method based on fluctuation gating is as follows: ; ; in, This represents the low-dimensional occupancy embedding vector of the target intersection in the r-th time window. This indicates the number of time windows after merging. Let represent the time-series sensing vector of the target intersection in the r-th time window. This represents the time-series perception vector of the target intersection in the (r-1)th time window. All of these represent the parameters of the weight sensing matrix in the time-aware layer. Represents a low-dimensional occupancy embedding vector Fluctuation gating, Indicates the volatility sensitivity coefficient. This represents an exponential function with the natural constant as its base. Represents the occupied state tensor Standard deviation of occupancy status in This represents the occupancy state tensor of the target intersection in the r-th time window.

[0012] Furthermore, in step S3, the lane capacity calculation function, modified by dynamic factors, outputs the lane capacity of the target intersection under the influence of unplanned temporary stops, including: S31: Based on the probability distribution of the occupancy mode of the target intersection in all time windows, the probability distribution of the occupancy mode is weighted and normalized according to the time range length of the comprehensive time window to obtain the comprehensive probability of the temporary parking occupancy mode and the comprehensive probability of the queuing occupancy mode of the target intersection. S32: Calculate the temporary parking occupancy reduction factor based on the comprehensive probability of the temporary parking occupancy mode at the target intersection; S33: Calculate the queuing occupancy delay factor based on the comprehensive probability of the queuing occupancy pattern at the target intersection; S34: Combining the temporary parking occupancy reduction factor and the queuing occupancy delay factor, construct a lane capacity calculation function with dynamic factor correction, and use the lane capacity calculation function to output the lane capacity of the target intersection under the influence of unplanned temporary parking.

[0013] Furthermore, the expression for the lane capacity calculation function is as follows: ; in, Indicates basic traffic capacity. This indicates the temporary occupancy reduction factor. This represents the queuing delay factor. This indicates the lane capacity of the target intersection under the influence of unplanned temporary stops.

[0014] Furthermore, in step S4, the short-term traffic control strategy for the target intersection is dynamically adjusted based on the lane capacity of the target intersection, including: S41: Calculate the lane release effectiveness index of the target intersection based on the lane capacity of the target intersection and the combined probability of temporary parking occupancy mode; S42: Convert the lane clearance effectiveness index into the green light duration ratio within a signal cycle; S43: Calculate the invalid green light suppression coefficient of the target intersection, dynamically suppress the green light duration ratio, and obtain the green light duration ratio after dynamic suppression; S44: Use the green light duration ratio after dynamic suppression as the short-term traffic control strategy for the target intersection to adjust and control the green light duration of the target intersection.

[0015] This invention also proposes an urban traffic optimization and planning system based on artificial intelligence data processing. The urban traffic optimization and planning system includes a data acquisition module, an occupancy state tensor construction module, an occupancy pattern analysis module, and a short-term traffic control optimization module, so as to realize the urban traffic optimization and planning method based on artificial intelligence data processing as described above.

[0016] Compared with existing technologies, this invention proposes a method and system for urban traffic optimization and planning based on artificial intelligence data processing. This technology has the following beneficial effects: First, in constructing the occupancy state tensor, this invention uses the maximum number of vehicles in a lane within a time window as the vehicle quantity occupancy intensity. This effectively captures extreme congestion or abnormal accumulation in local lanes, avoiding congestion weakening caused by averaging. Furthermore, by inversely normalizing the average vehicle speed, lane speed occupancy intensity is constructed, giving higher weight to low-speed operation states in occupancy assessment, thereby enhancing sensitivity to queuing congestion and temporary stop interference. Further, this invention employs a nonlinear collaborative occupancy state function to jointly map vehicle quantity occupancy intensity and lane speed occupancy intensity. This distinguishes between different traffic states such as high-density low-speed congestion, temporary stop interference, and normal traffic flow, reducing the risk of misjudgment. The consistency determination of occupancy state merges adjacent time windows, making the occupancy state tensor smoother and more continuous in the time dimension, weakening the impact of instantaneous noise on the judgment results. The final occupancy state tensor simultaneously retains information on occupancy intensity level and state stability.

[0017] Meanwhile, this invention achieves adaptive temporal perception modeling of traffic occupancy status at target intersections by introducing a fluctuation-gated recursive update mechanism based on the standard deviation of occupancy status. Specifically, this invention utilizes the standard deviation of occupancy status to construct a fluctuation gating coefficient, enabling the temporal update process to perceive the differences in stability and volatility of traffic status within different time windows: when occupancy status fluctuates significantly, the gating coefficient increases, enhancing the influence of the low-dimensional occupancy embedding vector of the current time window on the temporal status, thereby quickly responding to abnormal changes such as sudden congestion and temporary parking interference; when occupancy status is relatively stable, the gating coefficient decreases, strengthening the continuity of historical temporal perception vectors and suppressing the interference of short-term noise on status updates. Through the above recursive update method, while maintaining temporal continuity, it achieves differentiated processing of sudden changes and stable evolution of occupancy status, making the obtained temporal perception vector both sensitive and robust, improving the accuracy and robustness of subsequent traffic optimization processes. Attached Figure Description

[0018] Figure 1 A flowchart illustrating an urban traffic optimization and planning method based on artificial intelligence data processing, provided as an embodiment of the present invention; Figure 2This is a design diagram of a traffic sensing device deployment method provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the system structure of an urban traffic optimization and planning system provided in an embodiment of the present invention.

[0019] In the figure, the meanings of the attached labels are as follows: 100, Urban Traffic Optimization and Planning System; 101, Data Acquisition Module; 102, Occupancy State Tensor Construction Module; 103, Occupancy Pattern Analysis Module; 104, Short-Term Traffic Control Optimization Module.

[0020] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0022] This invention provides a method for urban traffic optimization and planning based on artificial intelligence data processing. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device, and the software may be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0023] Reference Figure 1 as well as Figure 2 As shown, Embodiment 1 of the present invention is as follows: A method for urban traffic optimization and planning based on artificial intelligence data processing includes the following steps: S1: Collect the time-series sensing data stream of the target intersection, and extract the occupancy state tensor based on the time window from the time-series sensing data stream to obtain the occupancy state tensor of the target intersection in the time window.

[0024] Specifically, step S1, which involves collecting the time-series sensing data stream of the target intersection, includes: S11: Deploy multiple traffic sensing devices at the lane positions of the target intersection. Each traffic sensing device is equipped with a unified clock to collect lane sensing data synchronized with the time. The lane sensing data includes the data collection time, the number of vehicles in the lane, and the average vehicle speed. As one embodiment of the present invention, the traffic sensing device includes a camera and a geomagnetic sensor, as shown in the figure below. Figure 2 The diagram shows a design for the deployment of traffic sensing equipment. By deploying cameras above or to the side of the target intersection, continuous imaging of the target lane is performed to obtain lane images of the target lane within the data acquisition time range. The target lane is a dedicated right-turn lane for motor vehicles. Using vehicle target detection and tracking algorithms, vehicle targets in lane images within the data acquisition time range are identified and counted to obtain the number of vehicles in the target lane at the time of data acquisition. Two sets of geomagnetic sensors with known spacing are deployed on the target lane. The vehicle speed is calculated by the time difference between the triggering of the geomagnetic sensors before and after the vehicle. The average speed of multiple vehicles within the data collection time range is used as the average speed of vehicles in the target lane at the data collection time. The data acquisition time range is a time range centered on the data acquisition time. Specifically, for data acquisition time t, the time range corresponding to data acquisition time t is: ,in Indicates the preset data collection time interval; Optionally, the vehicle target monitoring can use a target detection model based on a convolutional neural network to detect vehicle targets in the lane image, and output the bounding box position and category information of the vehicle target. The target detection model can be the YOLO model or its lightweight variant, which is used to quickly locate vehicle targets in a single frame image. It can also use state prediction based on Kalman filtering and matching rules based on distance or overlap to track the same vehicle target in continuous lane images. By assigning a unique identifier to each tracked vehicle and updating the vehicle state when the vehicle enters or leaves the preset lane area, the deduplication count of valid vehicle targets in the lane can be achieved, thereby obtaining the number of vehicles in the target lane at the time of data acquisition. S12: Construct time windows with consistent and non-overlapping time ranges. Based on the data acquisition time, insert the number of vehicles in the lane and the average vehicle speed in the lane perception data into the corresponding time window to obtain a time-series perception data stream.

[0025] Specifically, the traffic sensing device in the urban traffic optimization and planning system collects lane sensing data at preset data collection time intervals and periodically forms a time-series sensing data stream for the planning time, wherein the data collection time interval is set to 20 seconds, and the planning time... An example of a time-series-aware data stream is as follows: ; ; ; ; in, Indicates the time to be planned The time-series sensing data stream, These represent the lane perception data sequences within the first to M time windows, respectively. , This represents the lane perception data sequence within the m-th time window, where M represents the number of time windows. The lane perception data sequences are represented in sequence. In the middle of The lane perception data at each data acquisition time point, where N represents the number of lane perception data points contained in the formed time-series perception data stream. Represents the lane perception data sequence The starting time number of the middle lane perception data collection. Represents the lane perception data sequence The data collection time number at which the lane perception data terminates; mod represents the modulo operator. Indicates rounding down; Furthermore, select the distance from the time to be planned. The lane perception data from the N closest data acquisition times are used as the temporal perception data stream. Lane perception data in the data, and for time-series perception data stream. The middle lane perception data generates data collection time numbers from 1 to N, with N set to 90 and M set to 15.

[0026] Step S1, which involves extracting the occupancy state tensor of the time-series-aware data stream based on a time window, further includes: S13: Based on the time-series sensing data stream, extract the maximum number of vehicles in the lane from all lane sensing data within the time window as the vehicle occupancy intensity of the target intersection within the time window. S14: Extract the mean of the average vehicle speed in all lane perception data within the time window, and perform reverse normalization on the mean of the average vehicle speed to serve as the lane speed occupancy intensity of the target intersection within the time window. Specifically, the formula for inversely normalizing the mean of the vehicle's average speed is as follows: ; in, This represents the average vehicle speed across all lane perception data within the time window. express The result of reverse normalization, This indicates the vehicle's reference speed; the vehicle's reference speed is set to 50 kilometers per hour. S15: The vehicle quantity occupancy intensity and lane speed occupancy intensity of the target intersection in the time window are concatenated to form the occupancy feature vector of the target intersection in the time window; S16: Based on the occupancy feature vector, a nonlinear cooperative occupancy state function is used to perform a nonlinear mapping on the vehicle quantity occupancy intensity and lane speed occupancy intensity to obtain the occupancy state of the target intersection in the time window; Specifically, the formula for nonlinear mapping using the nonlinear cooperative occupancy state function is as follows: ; in, Indicates the intensity of vehicle occupancy. Indicates lane occupancy intensity by speed. Indicates the intensity of vehicle occupancy. and lane speed occupancy intensity The occupancy status obtained by nonlinear mapping This represents the speed suppression amplification index, used to enhance the impact of low-speed conditions on occupancy. (Setting...) It is 1.2; When both vehicle quantity intensity and lane speed intensity are high, the occupancy status increases rapidly, indicating high-density low-speed congestion; when vehicle quantity intensity is low but lane speed intensity is high, the occupancy status is moderate, indicating temporary parking interference; when lane speed intensity is low, the occupancy status is naturally suppressed to avoid misjudging the occupancy status. S17: Based on the occupancy status of the target intersection in the time window, determine the consistency of the occupancy status of adjacent time windows, and merge adjacent time windows that are determined to have the same occupancy status. As an embodiment of the present invention, the process for determining the consistency of occupancy status is as follows: Calculate the difference in occupancy status between adjacent time windows. If the absolute value of the difference is less than the preset occupancy status stability threshold, it is determined that the adjacent time windows are continuous and consistent in occupancy status, and a merging operation is performed. Optionally, the occupancy status stability threshold is set to 5. S18: Calculate the occupancy statistics of the merged time window as the occupancy status tensor of the target intersection in the time window. The occupancy statistics include the time range length of the time window, the mean of the occupancy status, the standard deviation of the occupancy status, and the mean of the occupancy feature vector.

[0027] Specifically, the occupancy state tensor of the target intersection within the time window is represented as follows: ; in, This represents the occupancy state tensor of the target intersection in the r-th time window, where the time windows described in steps S18, S2, S3, and S4 are all merged time windows. These represent the occupied state tensors respectively. The time range length, mean of occupied states, standard deviation of occupied states, and mean of occupied feature vectors are all included. This indicates the number of time windows after merging.

[0028] Specifically, the occupancy state tensor proposed in this invention, which combines time windows, provides a structured characterization of the lane occupancy state at the target intersection from the perspectives of temporal continuity and traffic state stability. This achieves a unified expression of lane occupancy intensity, duration, and evolutionary stability, providing a state representation with clear physical meaning for capacity correction and short-term traffic control.

[0029] Furthermore, the occupancy state tensor reflects the stability and sustained impact of occupancy state over time by recording the duration of the merged time window; the mean and standard deviation of occupancy state comprehensively characterize the overall congestion level and fluctuation characteristics of the lane during that time period, distinguishing between stable queuing occupancy and short-term disturbance occupancy; the mean of quantitative occupancy intensity characterizes the degree of vehicle spatial aggregation, reflecting the stability of lane load-bearing pressure; and the mean of speed inhibition occupancy factor reflects the degree of vehicle obstruction.

[0030] S2: Based on the occupancy state tensor, the occupancy pattern analysis model is used to output the probability distribution of the occupancy pattern of the target intersection within the time window.

[0031] The S2 step, which uses the occupancy pattern analysis model to output the probability distribution of the occupancy pattern of the target intersection within a time window, also includes: S21: The occupancy pattern analysis model includes a feature embedding layer, a time-series awareness layer, and a pattern discrimination layer; S22: The feature embedding layer takes the occupancy state tensor of the target intersection in the time window as input, and uses a nonlinear mapping method based on the time range length embedding to map the occupancy state tensor to obtain a low-dimensional occupancy embedding vector. Specifically, the mapping formula for the occupancy state tensor is: ; in, This represents the occupancy state tensor of the target intersection in the r-th time window. This indicates the number of time windows after merging. Represents the occupied state tensor The low-dimensional occupancy embedding vector obtained by mapping Represents the occupied state tensor The length of the time range in the text. Indicates the length of the time range The embedded value, This is the Hadamard product operator, which means multiplying elements at corresponding positions. Represents the logarithmic function. This represents the parameters of the convolution matrix in the feature embedding layer. This represents the bias parameter in the feature embedding layer. This represents a non-linear activation function, set to the ReLU function. Represents the parameters of the temporal embedding matrix, where and The dimensions are consistent; S23: The time-series awareness layer converts the standard deviation of the occupancy state into a fluctuation gating coefficient, and uses a recursive update method based on fluctuation gating to update the state of the low-dimensional occupancy embedding vector under time-series awareness, thereby obtaining the time-series awareness vector. S24: The pattern discrimination layer is a fully connected layer structure. The pattern discrimination layer converts the time-series perception vector into the probability of temporary parking and queuing occupancy modes of the target intersection in the time window, which are used as the probability distribution of the occupancy modes of the target intersection in the time window.

[0032] Specifically, the conversion formula for the pattern discrimination layer to transform the time-series perception vector into the probability of temporary parking occupancy mode and the probability of queuing occupancy mode at the target intersection within the time window is as follows: ; ; ; in, Let represent the probability distribution of occupancy patterns at the target intersection in the r-th time window, where These represent the probabilities of the target intersection in the temporary parking mode and the queuing mode in the r-th time window, respectively. The probabilities of the temporary parking mode and the queuing mode represent the probabilities that the target intersection is in the temporary parking mode or the queuing mode in the time window, respectively. This represents an exponential function with the natural constant as its base. The parameters of the convolution matrix represent the temporary occupancy mode. The bias parameter represents the temporary occupancy mode. The parameters of the convolution matrix represent the queuing occupancy pattern. The bias parameter represents the queuing occupancy mode.

[0033] In step S23, the occupancy state standard deviation is converted into fluctuation gating coefficients. The update formula for the time-aware state update of the low-dimensional occupancy embedding vector using a recursive update method based on fluctuation gating is as follows: ; ; in, This represents the low-dimensional occupancy embedding vector of the target intersection in the r-th time window. This indicates the number of time windows after merging. Let represent the time-series sensing vector of the target intersection in the r-th time window. This represents the time-series perception vector of the target intersection in the (r-1)th time window. All of these represent the parameters of the weight sensing matrix in the time-aware layer. Represents a low-dimensional occupancy embedding vector Fluctuation gating, Indicates the fluctuation sensitivity coefficient, set It is 0.3. This represents an exponential function with the natural constant as its base. Represents the occupied state tensor Standard deviation of occupancy status in This represents the occupancy state tensor of the target intersection in the r-th time window.

[0034] As an embodiment of the present invention, the occupancy state tensor of the intersection in the temporary parking occupancy mode or the queuing occupancy mode is obtained, and the occupancy mode of the intersection associated with the obtained occupancy state vector is taken as the actual occupancy mode of the occupancy state vector. The obtained occupancy state vector and the actual occupancy mode are used to construct a training dataset. The occupancy mode analysis model receives the obtained occupancy state vector and outputs the probability of the occupancy state vector in the actual occupancy mode. The training loss function is constructed to maximize the probability of the occupancy state vector in the actual occupancy mode. The training loss function is solved by using the gradient descent algorithm or the Adam optimizer to achieve the training optimization of the trainable parameters in the occupancy mode analysis model.

[0035] S3: Based on the probability distribution of the occupancy pattern, the lane capacity calculation function is modified by introducing a temporary parking occupancy reduction factor and a queuing occupancy delay factor to output the lane capacity of the target intersection under the influence of unplanned temporary parking.

[0036] The lane capacity calculation function modified by dynamic factors in step S3 outputs the lane capacity of the target intersection under the influence of unplanned temporary stops, including: S31: Based on the probability distribution of the occupancy mode of the target intersection in all time windows, the probability distribution of the occupancy mode is weighted and normalized according to the time range length of the comprehensive time window to obtain the comprehensive probability of the temporary parking occupancy mode and the comprehensive probability of the queuing occupancy mode of the target intersection. Specifically, the formulas for calculating the combined probability of the temporary parking occupancy mode and the combined probability of the queuing occupancy mode at the target intersection are as follows: ; ; ; in, This represents the overall probability of temporary parking occupancy patterns at the target intersection. This represents the overall probability of queuing patterns at the target intersection. This represents the weighted result of processing the probability of temporary parking occupancy patterns at the target intersection across all time windows. This represents the weighted result of the probability calculation of queuing occupancy patterns at the target intersection across all time windows. They represent respectively to as well as Normalization is performed. Represents the occupied state tensor The length of the time range of the medium time window. Indicates the time window decay coefficient, set It is 0.2; S32: Calculate the temporary parking occupancy reduction factor based on the comprehensive probability of the temporary parking occupancy mode at the target intersection; Specifically, the formula for calculating the temporary occupancy reduction factor is as follows: ; in, This indicates the temporary occupancy reduction factor. Indicates the temporary stop sensitivity coefficient, set It is 0.3. This represents the overall probability of temporary parking occupancy patterns at the target intersection. This represents the mean of occupied states in the occupied state tensor. This represents the occupancy state tensor of the target intersection in the r-th time window; S33: Calculate the queuing occupancy delay factor based on the comprehensive probability of the queuing occupancy pattern at the target intersection; Specifically, the formula for calculating the queuing occupancy delay factor is as follows: ; in, This represents the queuing delay factor. Indicates the queuing delay sensitivity coefficient, set It is 0.4; S34: Combining the temporary parking occupancy reduction factor and the queuing occupancy delay factor, construct a lane capacity calculation function with dynamic factor correction, and use the lane capacity calculation function to output the lane capacity of the target intersection under the influence of unplanned temporary parking.

[0037] It should be noted that this invention incorporates weighted normalization processing of the occupancy pattern probability distribution within each time window, using the duration of the time window and the exponential decay of the time window. This ensures that the occupancy pattern probability distribution of time windows with longer durations and closer to the planned time has a higher weight in the comprehensive probability calculation, thereby avoiding the interference of instantaneous fluctuations on the global judgment and enhancing the timeliness and stability of the occupancy pattern recognition results. Furthermore, based on the comprehensive probability of the temporary parking occupancy pattern, a temporary parking occupancy reduction factor is constructed, causing the lane capacity to decay exponentially when temporary parking interference increases, truly reflecting the direct weakening effect of temporary parking behavior on physical traffic conditions. Based on the comprehensive probability of the queuing occupancy pattern, a queuing occupancy delay factor is constructed to characterize the lag in traffic capacity caused by the restriction of queue release, avoiding misjudging queue congestion as a completely impassable state. By comprehensively introducing temporary parking occupancy reduction factors and queuing occupancy delay factors, the lane capacity is dynamically corrected, so that the output capacity can not only distinguish different occupancy causes, but also adaptively adjust with changes in traffic conditions. The lane capacity assessment improves the accuracy of lane capacity assessment in complex and unstable traffic scenarios, and provides a reliable decision-making basis for subsequent short-term traffic control strategies.

[0038] The expression for the lane capacity calculation function is: ; in, Indicates basic traffic capacity, settings =1, This indicates the temporary occupancy reduction factor. This represents the queuing delay factor. This indicates the lane capacity of the target intersection under the influence of unplanned temporary stops.

[0039] Furthermore, when the temporary parking occupancy mode is dominant, the lane capacity is mainly affected by space encroachment and shows a rapid decline; when the queuing occupancy mode is dominant, the lane capacity is mainly affected by release delay and shows a slow but continuous decline; when the two occupancy modes coexist, the reduction effect and the delay effect work together to achieve a precise characterization of the traffic status of complex intersections.

[0040] S4: Based on the lane capacity of the target intersection, dynamically adjust the short-term traffic control strategy of the target intersection, and optimize the traffic at the target intersection according to the short-term traffic control strategy.

[0041] In step S4, the short-term traffic control strategy for the target intersection is dynamically adjusted based on the lane capacity of the target intersection, including: S41: Calculate the lane release effectiveness index of the target intersection based on the lane capacity of the target intersection and the combined probability of temporary parking occupancy mode; Specifically, the formula for calculating the lane release effectiveness index of the target intersection is as follows: ; in, This represents the lane release effectiveness index at the target intersection. Indicates the temporary shutdown inhibition coefficient; S42: Convert the lane clearance effectiveness index into the green light duration ratio within a signal cycle; Specifically, the conversion formula for the green light duration ratio is as follows: ; in, This represents the proportion of green light duration within a signal cycle, and indicates the preset maximum lane clearance effectiveness index. It is 2. This indicates the minimum green light duration percentage. Indicates the maximum green light duration percentage; optionally, set Set to 0.15 It is 0.4; S43: Calculate the invalid green light suppression coefficient of the target intersection, dynamically suppress the green light duration ratio, and obtain the green light duration ratio after dynamic suppression; Specifically, the green light duration ratio The dynamic suppression formula is: ; in, Indicates the proportion of green light duration The dynamic inhibition results Represents the dynamic suppression coefficient, set It is 0.2; S44: Use the green light duration ratio after dynamic suppression as the short-term traffic control strategy for the target intersection to adjust and control the green light duration of the target intersection.

[0042] Specifically, the short-term traffic control strategy for the target intersection is set to last for 10 minutes. After the short-term traffic control strategy ends, the green light duration of the target intersection is restored to the initial setting.

[0043] It should be noted that this invention constructs a lane release effectiveness index, coupling the actual capacity of a lane with the probability of temporary stop occupancy patterns in a model. This index accurately reflects whether a lane has the actual release capacity under a given green light condition, fundamentally distinguishing between queuing congestion and temporary stop blockage. Furthermore, this invention maps the lane release effectiveness index to a green light duration ratio, transforming green light allocation from being driven by the number of vehicles or saturation to being driven by release effectiveness, avoiding blindly extending green lights under traffic-restricted conditions. Further, this invention introduces an ineffective green light suppression coefficient to suppress green lights at target intersections with a high probability of temporary stop occupancy patterns, reducing the waste of green lights caused by unplanned temporary stops. This prioritizes limited signal resources for intersections with actual traffic conditions. The resulting dynamically suppressed green light duration ratio serves as a short-term traffic control strategy, enabling adaptive adjustment of green light duration. This method effectively improves green light utilization efficiency, suppresses the amplified impact of abnormal occupancy on intersection operation, and enhances the overall traffic efficiency and operational stability of target intersections in complex traffic scenarios.

[0044] Example 2: A city traffic optimization and planning system 100 based on artificial intelligence data processing, as shown in the figure below. Figure 3 The schematic diagram of the urban traffic optimization and planning system 100 shown includes a data acquisition module 101, an occupancy state tensor construction module 102, an occupancy pattern analysis module 103, and a short-term traffic control optimization module 104, to implement an urban traffic optimization and planning method based on artificial intelligence data processing as described in Example 1. The functions of each module are as follows: The data acquisition module 101 is used to acquire the time-series sensing data stream of the target intersection; The occupancy state tensor construction module 102 is used to extract the occupancy state tensor based on the time window from the time-series sensing data stream to obtain the occupancy state tensor of the target intersection in the time window. The occupancy pattern analysis module 103 is used to output the probability distribution of occupancy patterns of the target intersection within a time window using the occupancy pattern analysis model; The short-term traffic control optimization module 104 is used to calculate the lane capacity by using a lane capacity calculation function that incorporates dynamic factors such as temporary parking occupancy reduction factor and queuing occupancy delay factor, outputs the lane capacity of the target intersection under the influence of unplanned temporary parking, and dynamically adjusts the short-term traffic control strategy of the target intersection.

[0045] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in terms of the scope of the patent invention.

[0046] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0047] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0048] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for urban traffic optimization and planning based on artificial intelligence data processing, characterized in that, The method includes: S1: Collect the time-series sensing data stream of the target intersection, and extract the occupancy state tensor based on the time window from the time-series sensing data stream to obtain the occupancy state tensor of the target intersection in the time window. S2: Based on the occupancy state tensor, use the occupancy pattern analysis model to output the probability distribution of the occupancy pattern of the target intersection within the time window; S3: Based on the probability distribution of the occupancy pattern, the lane capacity calculation function is modified by introducing a dynamic factor to correct the temporary parking occupancy reduction factor and the queuing occupancy delay factor, and outputs the lane capacity of the target intersection under the influence of unplanned temporary parking. S4: Based on the lane capacity of the target intersection, dynamically adjust the short-term traffic control strategy of the target intersection, and optimize the traffic at the target intersection according to the short-term traffic control strategy.

2. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 1, characterized in that, The step S1 involves collecting the time-series sensing data stream of the target intersection, including: S11: Deploy multiple traffic sensing devices at the lane positions of the target intersection. Each traffic sensing device is equipped with a unified clock to collect lane sensing data synchronized with the time. The lane sensing data includes the data collection time, the number of vehicles in the lane, and the average vehicle speed. S12: Construct time windows with consistent and non-overlapping time ranges. Based on the data acquisition time, insert the number of vehicles in the lane and the average vehicle speed in the lane perception data into the corresponding time window to obtain a time-series perception data stream.

3. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 2, characterized in that, Step S1, which involves extracting the occupancy state tensor of the time-series-aware data stream based on a time window, further includes: S13: Based on the time-series sensing data stream, extract the maximum number of vehicles in the lane from all lane sensing data within the time window as the vehicle occupancy intensity of the target intersection within the time window. S14: Extract the mean of the average vehicle speed in all lane perception data within the time window, and perform reverse normalization on the mean of the average vehicle speed to serve as the lane speed occupancy intensity of the target intersection within the time window. S15: The vehicle quantity occupancy intensity and lane speed occupancy intensity of the target intersection in the time window are concatenated to form the occupancy feature vector of the target intersection in the time window; S16: Based on the occupancy feature vector, a nonlinear cooperative occupancy state function is used to perform a nonlinear mapping on the vehicle quantity occupancy intensity and lane speed occupancy intensity to obtain the occupancy state of the target intersection in the time window; S17: Based on the occupancy status of the target intersection in the time window, determine the consistency of the occupancy status of adjacent time windows, and merge adjacent time windows that are determined to have the same occupancy status. S18: Calculate the occupancy statistics of the merged time window as the occupancy status tensor of the target intersection in the time window. The occupancy statistics include the time range length of the time window, the mean of the occupancy status, the standard deviation of the occupancy status, and the mean of the occupancy feature vector.

4. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 1, characterized in that, The S2 step, which uses the occupancy pattern analysis model to output the probability distribution of the occupancy pattern of the target intersection within a time window, also includes: S21: The occupancy pattern analysis model includes a feature embedding layer, a time-series awareness layer, and a pattern discrimination layer; S22: The feature embedding layer takes the occupancy state tensor of the target intersection in the time window as input, and uses a nonlinear mapping method based on the time range length embedding to map the occupancy state tensor to obtain a low-dimensional occupancy embedding vector. S23: The time-series awareness layer converts the standard deviation of the occupancy state into a fluctuation gating coefficient, and uses a recursive update method based on fluctuation gating to update the state of the low-dimensional occupancy embedding vector under time-series awareness, thereby obtaining the time-series awareness vector. S24: The pattern discrimination layer is a fully connected layer structure. The pattern discrimination layer converts the time-series perception vector into the probability of temporary parking and queuing occupancy modes of the target intersection in the time window, which are used as the probability distribution of the occupancy modes of the target intersection in the time window.

5. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 4, characterized in that, In step S23, the occupancy state standard deviation is converted into fluctuation gating coefficients. The update formula for the time-aware state update of the low-dimensional occupancy embedding vector using a recursive update method based on fluctuation gating is as follows: ； ； in, This represents the low-dimensional occupancy embedding vector of the target intersection in the r-th time window. This indicates the number of time windows after merging. Let represent the time-series sensing vector of the target intersection in the r-th time window. This represents the time-series perception vector of the target intersection in the (r-1)th time window. All of these represent the parameters of the weight sensing matrix in the time-aware layer. Represents a low-dimensional occupancy embedding vector Fluctuation gating, Indicates the volatility sensitivity coefficient. This represents an exponential function with the natural constant as its base. Represents the occupied state tensor Standard deviation of occupancy status in This represents the occupancy state tensor of the target intersection in the r-th time window.

6. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 1, characterized in that, The lane capacity calculation function modified by dynamic factors in step S3 outputs the lane capacity of the target intersection under the influence of unplanned temporary stops, including: S31: Based on the probability distribution of the occupancy mode of the target intersection in all time windows, the probability distribution of the occupancy mode is weighted and normalized according to the time range length of the comprehensive time window to obtain the comprehensive probability of the temporary parking occupancy mode and the comprehensive probability of the queuing occupancy mode of the target intersection. S32: Calculate the temporary parking occupancy reduction factor based on the comprehensive probability of the temporary parking occupancy mode at the target intersection; S33: Calculate the queuing occupancy delay factor based on the comprehensive probability of the queuing occupancy pattern at the target intersection; S34: Combining the temporary parking occupancy reduction factor and the queuing occupancy delay factor, construct a lane capacity calculation function with dynamic factor correction, and use the lane capacity calculation function to output the lane capacity of the target intersection under the influence of unplanned temporary parking.

7. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 6, characterized in that, The expression for the lane capacity calculation function is: ； Here, represents basic traffic capacity. This indicates the temporary occupancy reduction factor. This represents the queuing occupancy delay factor, indicating the lane capacity of the target intersection under the influence of unplanned temporary stops.

8. The urban traffic optimization and planning method based on artificial intelligence data processing as described in claim 7, characterized in that, In step S4, the short-term traffic control strategy for the target intersection is dynamically adjusted based on the lane capacity of the target intersection, including: S41: Calculate the lane release effectiveness index of the target intersection based on the lane capacity of the target intersection and the combined probability of temporary parking occupancy mode; S42: Convert the lane clearance effectiveness index into the green light duration ratio within a signal cycle; S43: Calculate the invalid green light suppression coefficient of the target intersection, dynamically suppress the green light duration ratio, and obtain the green light duration ratio after dynamic suppression; S44: Use the green light duration ratio after dynamic suppression as the short-term traffic control strategy for the target intersection to adjust and control the green light duration of the target intersection.

9. A city traffic optimization and planning system based on artificial intelligence data processing, characterized in that, The urban traffic optimization and planning system includes a data acquisition module, an occupancy state tensor construction module, an occupancy pattern analysis module, and a short-term traffic control optimization module, to realize an urban traffic optimization and planning method based on artificial intelligence data processing as described in any one of claims 1-8.

Images