Methods and Equipment for Collaborative Traffic Congestion Mitigation Based on Traffic Flow Prediction

Abstract

This invention discloses a method and device for coordinated regional traffic congestion mitigation based on traffic flow prediction, belonging to the field of traffic management technology. The method includes: acquiring multi-source traffic data of the target area's road network and performing congestion analysis to extract congestion information; collecting historical driver trajectories to construct a behavioral model, quantifying the induced response, and generating compliance probabilities; analyzing diversion paths based on congestion information to generate detour paths and obtaining remaining capacity; calculating an information diffusion coefficient by combining compliance probabilities and remaining path capacity; and generating a differentiated information push strategy based on the information diffusion coefficient. This invention solves the technical problem that existing traffic mitigation methods do not consider the probability of drivers complying with induced information, leading to blind and untargeted induced strategies that easily cause secondary congestion on detour paths. It achieves the technical effect of improving the accuracy and effectiveness of traffic induced information by calculating the information diffusion coefficient and generating differentiated push strategies, thereby avoiding excessive induced congestion.

Description

Technical Field

[0001] This invention relates to the field of traffic management technology, specifically to a method and equipment for coordinated regional traffic congestion mitigation based on traffic flow prediction. Background Technology

[0002] With the accelerating pace of urbanization and the continuous growth of motor vehicle ownership, the traffic load on urban road networks is increasingly heavy, and traffic congestion has become a significant issue affecting urban operational efficiency and residents' travel experience. Especially during rush hours, emergencies, road construction, and inclement weather, regional traffic congestion can easily form in localized sections and spread to surrounding road networks, leading to decreased road network efficiency, increased queue lengths, longer travel times, and even triggering chain reactions of traffic jams. Current traffic management methods primarily rely on fixed signal control, manual direction-of-care, broadcast reminders, or navigation platforms issuing detour suggestions to alleviate congestion. While some technical solutions can identify or predict congestion based on real-time traffic data and issue guidance information to drivers, they still have the following shortcomings in practical applications: First, existing solutions mostly focus on the road network's operational status itself, lacking in-depth analysis of drivers' historical travel behavior and information response differences, making it difficult to accurately assess drivers' actual compliance with guidance information; second, existing guidance information dissemination usually adopts a general broadcast or large-scale push approach, failing to fully consider the remaining capacity of detour routes, easily causing some detour roads to receive excessive diverted traffic in a short period, thus creating new congestion points; third, for the coordinated management of regional traffic congestion, there is a lack of a dynamic control mechanism that comprehensively combines traffic flow prediction, driver response probability, and detour capacity constraints, making it difficult to precisely control the intensity and coverage of information dissemination, resulting in limited management effectiveness. Therefore, there is an urgent need for a collaborative management method that integrates traffic flow prediction, driver behavior characteristic quantification, and road network capacity perception to solve congestion problems caused by excessive guidance or inaccurate compliance rate predictions. Summary of the Invention

[0003] This application provides a method and device for coordinated regional traffic congestion mitigation based on traffic flow prediction. The key feature is that it solves the technical problem that existing traffic mitigation methods do not consider the probability of drivers complying with guidance information, resulting in blind and untargeted guidance strategies that easily lead to secondary congestion on detour routes. It achieves the technical effect of improving the accuracy and effectiveness of traffic guidance information by calculating the information diffusion coefficient and generating differentiated push strategies, thereby avoiding excessive guidance that causes new congestion.

[0004] A first aspect of this application provides a method for coordinated regional traffic congestion mitigation based on traffic flow prediction, the method comprising:

[0005] The system acquires multi-source traffic data of the target area's road network, performs traffic congestion analysis based on the multi-source traffic data, and extracts traffic congestion information; collects historical driver behavior trajectory data, constructs a driver behavior model, quantifies driver guidance response characteristics, and generates predicted information compliance probabilities; based on the traffic congestion information, it performs traffic diversion path analysis on the target area's road network, generates detour paths, and collects the remaining capacity of the detour paths; it calculates an information diffusion coefficient by combining the predicted information compliance probability and the remaining capacity of the detour paths, the information diffusion coefficient being used to characterize the intensity of allowing guidance information to be released into the road network at the current moment; and generates differentiated information push strategies based on the information diffusion coefficient.

[0006] A second aspect of this application provides a regional traffic congestion coordination and mitigation device based on traffic flow prediction, the device comprising:

[0007] The congestion analysis module acquires multi-source traffic data of the target area's road network, performs traffic congestion analysis based on the multi-source traffic data, and extracts traffic congestion information. The feature quantification module collects historical driver behavior trajectory data, constructs a driver behavior model, quantifies driver guidance response features, and generates predicted information compliance probabilities. The traffic diversion path analysis module analyzes traffic diversion paths in the target area's road network based on the traffic congestion information, generates detour paths, and collects the remaining capacity of the detour paths. The diffusion coefficient calculation module combines the predicted information compliance probabilities and the remaining capacity of the detour paths to calculate the information diffusion coefficient, which characterizes the intensity of the permitted dissemination of guidance information into the road network at the current moment. The push strategy generation module generates differentiated information push strategies based on the information diffusion coefficient.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, multi-source traffic data from the target area's road network is collected to analyze the current traffic flow and identify congestion. Then, a behavioral model is constructed using historical driver behavior trajectories to assess drivers' response characteristics to guidance information, yielding the probability of compliance. Next, congestion information is combined with road network detour path analysis to generate feasible detour routes and assess their remaining capacity. Then, by integrating driver compliance probability and detour route capacity, an information diffusion coefficient is calculated to determine the intensity of guidance information dissemination. Finally, based on this coefficient, differentiated information delivery strategies are developed to achieve orderly traffic flow guidance and congestion relief. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the regional traffic congestion collaborative mitigation method based on traffic flow prediction provided in an embodiment of this application.

[0012] Figure 2 A schematic diagram of a regional traffic congestion coordination and mitigation device based on traffic flow prediction, provided in an embodiment of this application.

[0013] Figure labeling: 11. Congestion analysis module, 12. Feature quantification module, 13. Diversion path analysis module, 14. Diffuse coefficient calculation module, 15. Push strategy generation module. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown, this application provides a method for coordinated regional traffic congestion mitigation based on traffic flow prediction, wherein the method includes:

[0016] Obtain multi-source traffic data of the road network in the target area, perform traffic congestion analysis based on the multi-source traffic data, and extract traffic congestion information.

[0017] In this embodiment, traffic data is acquired within the target area road network using various data collection methods. This traffic data includes: traffic flow, speed, and occupancy data collected by roadside geomagnetic detectors, loop detectors, or radar equipment; vehicle queue length and traffic status information obtained through image recognition from video surveillance equipment; GPS-based floating car trajectory data; and real-time travel data transmitted from navigation platforms. Subsequently, the acquired multi-source traffic data undergoes preprocessing, including removing outliers and missing values, performing time alignment on data from different sources, mapping trajectory data to specific road segments using map matching algorithms, and aggregating data according to a unified time granularity to construct a standardized multi-source traffic dataset. Following this, traffic operation status analysis is performed on each road segment. Parameters such as average vehicle speed, traffic flow per unit time, and traffic density are calculated, and a short-term traffic flow prediction model is used to predict the traffic status of each road segment over a future period. Finally, based on the current and predicted traffic status results, key road segments or nodes exhibiting congestion or a congestion trend are identified, thus determining congestion risk points. Finally, information such as the specific location of the congestion, the degree of congestion, and the expected duration are extracted to form structured traffic congestion information, providing a basis for the formulation of subsequent traffic management and guidance strategies.

[0018] Furthermore, multi-source traffic data of the target area's road network is acquired, and traffic congestion analysis is performed based on the multi-source traffic data to extract traffic congestion information, including:

[0019] Based on the multi-source traffic data, the traffic state prediction results of each road segment in the target area road network are analyzed; based on the traffic state prediction results, congestion risk points are identified, and the congestion location, congestion degree and expected duration corresponding to the congestion risk points are extracted as the traffic congestion information.

[0020] Preferably, the target area's road network is first digitally modeled, dividing the network into several directed road segment units and establishing road segment topological relationships, including upstream, downstream, and adjacent road segment connections. For each road segment, traffic state time series data with fixed time intervals is constructed. This traffic state time series data comes from multi-source traffic data, including indicators such as vehicle speed, flow rate, occupancy rate, and vehicle travel time. Subsequently, the traffic state time series data of each road segment is processed to calculate key derived feature parameters, including the ratio of the road segment's average vehicle speed to free-flow vehicle speed, the rate of change in the number of vehicles passing through per unit time, the speed fluctuation coefficient, and the state difference values ​​of adjacent road segments. Then, spatial correlation features, such as the congestion index of upstream or downstream road segments, are introduced, and a multi-dimensional feature vector is formed through normalization and concatenation. Next, this multidimensional feature vector is input into a short-term traffic flow prediction model for prediction. This model is constructed based on time-series models (such as ARIMA), deep learning models (such as LSTM, graph neural networks), or a combination thereof. The model outputs the predicted vehicle speed and predicted congestion index sequence for each road segment within several future time steps, thus obtaining the traffic state evolution result of each road segment within the prediction time window. Then, the prediction results are judged according to the preset congestion judgment rules. When the predicted vehicle speed is lower than a set proportion of the free-flow vehicle speed of the road segment or the congestion index is higher than a threshold, the road segment is marked as congested and designated as a candidate congested road segment. Based on this, spatial connectivity analysis is performed on these candidate congested road segments. Based on the road network topology, adjacent and continuously congested road segments are clustered and merged to form congestion risk points. At the same time, combined with the prediction sequence, the start and end times of congestion are determined, thereby analyzing the formation and dissipation process of congestion. Finally, structured traffic congestion information is extracted and generated for each congestion risk point, including congestion location, congestion level, and expected duration. The congestion location is the corresponding road segment number or geographical coordinate range; the congestion level includes the average speed reduction percentage and congestion index; and the expected duration is the duration calculated from the predicted start and end times. This traffic congestion information can provide a basis for subsequent traffic management decisions, enabling accurate identification of regional traffic congestion and early perception of its development trends, thereby improving the scientific rigor and timeliness of traffic management decisions.

[0021] For short-term traffic flow prediction models, Long Short-Term Memory (LSTM) networks can be used. Specifically, firstly, road segments are used as the basic modeling unit to construct input data samples. Time series data is generated for each road segment at fixed time intervals, and T consecutive time steps are selected as the input window. For example, T=12 corresponds to the past 60 minutes of data. Input features include vehicle speed, flow rate, occupancy rate, and time characteristics. The predicted vehicle speed, congestion index, and expected end time for the next k time steps are used as the model output labels. All features are then normalized to improve model training stability. Next, an LSTM network structure is constructed, where the input layer dimension is set to T×F, where F is the feature dimension. The first layer is an LSTM layer with 64 hidden units to extract long-term dependency features from the time series. A second LSTM layer with 32 hidden units can be stacked after this to enhance feature representation. A fully connected layer is connected after the LSTM layers to map the hidden states to the predicted output. ReLU and linear functions can be used as activation functions. To prevent overfitting, Dropout layers can be added between the LSTM layers. Next, the model training parameters are set. The loss function can be the mean squared error, the optimizer uses the Adam optimization algorithm, the learning rate can be set to 0.001, the batch size can be set to 32 or 64, and the number of training epochs can be set to 50-100 depending on the convergence. During training, a validation set is used for early stopping to prevent overfitting. Finally, using historical traffic data, the model is trained offline through forward propagation, loss calculation, backpropagation, and parameter optimization. After training, the model is deployed for online prediction. In actual operation, the latest T time-step data are input in real time, and the traffic state prediction results for the next k time steps are output, thus providing a foundation for subsequent congestion identification and traffic management.

[0022] Historical driver behavior trajectory data is collected to construct a driver behavior model, quantify the driver's induced response characteristics, and generate predictive information compliance probability.

[0023] In one embodiment, historical driver behavior trajectory data is first collected. This data originates from in-vehicle navigation devices, mobile terminal positioning data, or vehicle-to-everything (V2X) platforms, and includes records such as vehicle travel routes, timestamps, road segment selections, and whether navigation or guidance information was received and executed. Subsequently, this historical driver behavior trajectory data is input into a pre-built driver behavior model. The guidance response characteristics of the historical driver behavior trajectory data are quantified according to the processing flow reserved in the driver behavior model, extracting guidance response characteristic parameters for each driver. Then, these guidance response characteristic parameters are statistically analyzed to obtain the driver's response tendency to guidance information, which is then converted into a predicted information compliance probability in the current traffic situation. This predicted information compliance probability serves as an important input parameter for subsequent traffic guidance strategy formulation, reflecting the overall response level of drivers to guidance information, thereby improving the rationality and effectiveness of traffic management decisions.

[0024] Furthermore, historical driver behavior trajectory data is collected to construct driver behavior models, quantify driver guidance response characteristics, and generate predicted compliance probabilities, including:

[0025] The characteristic parameters of each driver are extracted from the historical driver behavior trajectory data. The characteristic parameters include the proportion of drivers who actually follow the detour suggestions after receiving guidance information. Based on the characteristic parameters, the average information compliance probability of the group is calculated as the predicted information compliance probability.

[0026] Preferably, in the driver behavior model, the first step is to identify trajectory segments containing guidance information interactions from historical driver behavior trajectory data. This involves filtering driving records where navigation or traffic guidance information was received during the journey, segmenting these segments into continuous trajectory segments before and after receiving the guidance information, and then mapping the trajectory points to specific road segments using map matching methods to reconstruct the driver's actual driving path. Subsequently, for each driver, their actual behavioral responses after receiving detour guidance information are statistically analyzed across multiple trip records. Specifically, the recommended detour route is compared with the driver's actual driving path. If the overlap between the actual and recommended paths exceeds a preset ratio, a detour is considered executed; otherwise, it is considered not executed. Based on this determination, the driver's execution ratio is calculated—the ratio of the number of detours executed to the total number of times guidance information was received—and used as the driver's guidance response characteristic parameter. Finally, the characteristic parameters of all drivers are summarized and statistically analyzed, using either an average or weighted average method (such as weighted by the number of trips) to calculate the average information compliance rate at the group level. Finally, the average information compliance rate of the group is used as the output of the predicted information compliance probability to characterize the overall response level of drivers to guidance information under the current traffic conditions, thereby providing a basis for subsequent information release intensity control.

[0027] Based on the traffic congestion information, traffic diversion path analysis is performed on the road network in the target area to generate detour paths and collect the remaining capacity of the detour paths.

[0028] In one embodiment, based on the acquired traffic congestion information, the locations of key road segments or nodes experiencing congestion are determined and used as the starting point for traffic diversion. Subsequently, a road network graph model is established based on the topology of the target area's road network, where road segments are treated as edges and intersections as nodes, and each road segment is assigned corresponding weight parameters. Then, while avoiding congested road segments, path search analysis is performed based on the road network graph model. This path search can employ shortest path algorithms, such as Dijkstra's algorithm or A* algorithm, with the optimization objective of minimizing travel time or overall cost. One or more feasible detour paths are calculated from the starting point to the destination. The generated detour paths are then screened and evaluated, eliminating paths with potential congestion risks or poor traffic conditions, and retaining several preferred detour paths as alternative routes. Then, for each detour route, real-time traffic status data of each road segment it includes is collected, the current capacity and actual traffic load of the corresponding road segment are calculated, and the remaining capacity of each road segment is estimated based on the difference between the capacity and the traffic volume already carried. The remaining capacity of each road segment in the route is then summarized, and the minimum value is taken to obtain the remaining capacity of the detour route. This serves as an important basis for the subsequent generation and release of traffic guidance information, thereby achieving reasonable diversion of traffic flow and congestion relief.

[0029] Furthermore, based on the traffic congestion information, traffic diversion path analysis is performed on the road network in the target area to generate detour routes and collect the remaining capacity of the detour routes, including:

[0030] Based on the congestion location in the traffic congestion information and the road network topology, the detour route is generated using the shortest path algorithm; the real-time traffic status of the detour route is collected, the remaining capacity is estimated, and the remaining capacity is used as the remaining capacity of the detour route.

[0031] Preferably, the congested road segments corresponding to the congested locations are first determined based on traffic congestion information, and these segments are marked as high-cost edges or prohibited edges in the road network graph model. Simultaneously, the road network of the target area is represented as a graph structure, abstracting each intersection as a node and the road segments connecting adjacent intersections as directed edges, assigning a travel cost to each edge. This cost can use the real-time travel time of the road segment as a base weight, adjusted in conjunction with the congestion index. Next, the starting and ending nodes for the detour path search are determined. The starting node can be an upstream branching node before entering the congested area, and the ending node can be a downstream node that rejoins the original path after bypassing the congested area, or the node corresponding to the vehicle's original travel destination can be directly selected. Then, a shortest path algorithm is used for path search. Taking Dijkstra's algorithm as an example, the minimum cumulative cost of the starting node is first recorded as 0, and the remaining nodes are recorded as infinity. A set of nodes to be searched is then established to store nodes whose shortest cost has not yet been determined. Simultaneously, a predecessor node record table is established to store the previous node of each node under the current optimal path. Next, the path search process begins. In each round, the node with the smallest cumulative cost is selected from the set of nodes to be searched as the current expanded node, denoted as u. Then, all adjacent nodes v connected to node u are traversed, and the candidate cost of reaching v via u is calculated. If the candidate cost is less than the minimum cost currently recorded for node v, the minimum cost of node v is updated, and the predecessor node of node v is recorded as u. After all adjacent edges of the current node u are relaxed, node u is marked as the node with the determined optimal path and is not processed again. This process continues iterating until the destination node is selected or all reachable nodes have been searched. After the search is completed, based on the predecessor node record table, the process backtracks from the destination node to the starting node to obtain a candidate detour path with the smallest total cost. To avoid insufficient induced selection due to generating only a single path, after obtaining the first optimal path, Dijkstra's search can be performed again by increasing the edge weights of some edges in the selected path or temporarily removing key edges, thereby generating 2-3 alternative detour paths for subsequent induced strategy selection. This approach can balance path length, real-time traffic efficiency, and diversion requirements. After obtaining the detour route, real-time traffic conditions along the route are collected to estimate its remaining capacity. Specifically, for each road segment included in the route, parameters such as real-time traffic flow, average speed, number of lanes, and road design capacity are obtained, and the current remaining capacity of that road segment is calculated. The remaining capacity of a single road segment can be expressed as the difference between the maximum capacity of that road segment in the current time period and the current actual traffic flow of that road segment. Since the overall traffic flow capacity of the detour route is usually limited by the weakest road segment, the minimum remaining capacity of all road segments in the route can be taken as the remaining capacity of the detour route.Finally, the detour routes obtained from the search and their corresponding remaining capacity are associated and stored to form a mapping relationship between detour routes and the transfer traffic that can be received. This provides a basic support for subsequent calculation of information diffusion coefficient and push of differentiated guidance information, thereby ensuring that the number of transferred vehicles after guidance does not exceed the carrying capacity of the detour route, and improving the rationality of the diversion route generation and the actual application effect.

[0032] The information diffusion coefficient is calculated by combining the predicted information compliance probability and the remaining capacity of the detour path. The information diffusion coefficient is used to characterize the intensity of the guidance information allowed to be released into the road network at the current moment.

[0033] In one embodiment, the predicted information compliance probability at the current moment is first obtained to characterize the likelihood that a driver will actually take a detour after receiving the guidance information in the current traffic situation. Simultaneously, the remaining capacity of each generated detour path is obtained to reflect the additional traffic flow that each detour path can handle. Subsequently, based on the proportion of vehicles that have received guidance information in the current road network, the predicted transfer flow is matched with the remaining capacity of the detour paths. The intensity of the guidance information dissemination is quantified using a preset mapping rule, and an information diffusion coefficient is calculated. This information diffusion coefficient is used as a control variable to guide the scale and coverage of subsequent guidance information pushes, thereby achieving orderly traffic diversion while ensuring that detour paths are not overloaded, and improving the coordination and effectiveness of traffic management.

[0034] Furthermore, the information diffusion coefficient is calculated by combining the predicted information compliance probability and the remaining capacity of the detour path, including:

[0035] The proportion of vehicles that have received guidance information in the road network at the current moment is obtained as the current information coverage rate; the information diffusion coefficient is calculated according to the remaining capacity of the detour path, the current information coverage rate, and the predicted information compliance probability through a preset mapping rule.

[0036] Preferably, the process begins by counting the number of vehicles that have received guidance information in the target area's road network at the current moment, and obtaining the total number of vehicles in the road network at the current moment. The number of vehicles that have received guidance information can be counted through feedback from in-vehicle terminals, records issued by the navigation platform, or vehicle-to-everything (V2X) communication records. The total number of vehicles can be estimated based on real-time detection data, floating car data, or road segment traffic flow statistics within the area. Then, the number of vehicles that have received guidance information is divided by the total number of vehicles in the road network at the current moment to obtain the current information coverage rate, which characterizes the propagation range and coverage level of the guidance information in the regional road network at the current moment. Next, the remaining capacity of the generated detour routes is obtained, and combined with the predicted information compliance probability, the potential traffic volume to be diverted after further guidance information is issued is estimated. Specifically, the estimated scale of vehicles that may actually enter the detour routes can be estimated by multiplying the number of newly guided vehicles by the predicted information compliance probability. Then, a mapping relationship is established between the information diffusion coefficient and the remaining capacity of the detour path, the current information coverage rate, and the predicted information compliance probability. Based on this mapping relationship, the maximum number of additional vehicles allowed to be accommodated is determined according to the remaining capacity of the detour path. Then, combining the current information coverage rate and the predicted information compliance probability, the proportion of information that can be further increased at the current moment is deduced, and this proportion is mapped to the information diffusion coefficient in the range of 0 to 1. Finally, the calculated information diffusion coefficient is used as a control parameter for induced information dissemination and input into the subsequent push strategy generation process. This ensures that the intensity of information dissemination matches the carrying capacity of the detour path, avoids excessive induction causing secondary congestion, and improves the coordination and effectiveness of traffic management.

[0037] Furthermore, the mapping rule aims to ensure that the induced transfer flow does not exceed the remaining capacity of the detour path, and dynamically adjusts the value of the information diffusion coefficient with the optimization objective of minimizing the total travel time of the road network.

[0038] Optionally, for this mapping rule, let the total number of vehicles in the current road network be N, the current information coverage rate be η, the predicted information compliance probability be p, and the information diffusion coefficient be β. Then, the proportion of newly added vehicles receiving guidance information at the current moment is β·(1−η). Correspondingly, the number of vehicles that actually undergo path transfer can be expressed as: transfer flow Q=N×β×(1−η)×p. Subsequently, the remaining capacity C of the detour path is obtained, and constraints are established to ensure that the transfer flow generated by the inducement does not exceed the carrying capacity of the detour path, that is, Q≤C. Thus, the upper bound constraint of the information diffusion coefficient can be obtained, β≤C / [N×(1−η)×p]. In the calculation process, β can be truncated to limit its value range to [0,1], thereby ensuring its physical meaning is reasonable. Based on this, with the minimization of the total travel time of the road network as the optimization objective, an objective function is constructed, that is, the sum of the travel times of each road segment in the road network is used as the evaluation index. After introducing guidance information, some traffic flow will shift from congested road sections to detour routes, thereby changing the flow distribution and corresponding travel time of each road section. Then, under the constraint Q≤C, the information diffusion coefficient β is searched. That is, using a step-by-step search or simple gradient adjustment method, the state of traffic flow redistribution after the road network is simulated or estimated under different β values, the corresponding total travel time T is calculated, and the β that minimizes T is selected as the optimal information diffusion coefficient for the current moment. Finally, this optimal information diffusion coefficient is used as a control parameter for the intensity of guidance information dissemination, achieving dynamic adjustment that neither exceeds the carrying capacity of detour routes nor reduces the overall total travel time of the road network, thereby improving the overall efficiency and coordination of regional traffic management.

[0039] A differentiated information push strategy is generated based on the information diffusion coefficient.

[0040] In one embodiment, the intensity of the guidance information release and the target information coverage are first determined based on the calculated information diffusion coefficient. Then, drivers are screened and grouped within the target area road network. Specifically, based on the vehicle's current location, driving direction, destination distribution, and spatial relationship to congested road sections, vehicles about to enter congested areas or with detour options are prioritized as target recipients. A differentiated push strategy is then formulated based on the grouping results. For example, for drivers with a high probability of compliance and close to congested areas, explicit detour route suggestions can be prioritized; for drivers with a low probability of compliance or far away, suggestive or weak guidance information can be used. Furthermore, different recommendation intensities or proportions can be assigned to different routes based on the remaining capacity of detour routes to achieve multi-path diversion. Subsequently, during the information push process, the push rhythm and frequency can be dynamically controlled. For example, the push scope can be gradually expanded in batches, and the changes in road network status and vehicle response can be monitored in real time. The push strategy can be adjusted according to the feedback results, thereby forming a differentiated information push strategy that matches the current road network status and carrying capacity. This enables refined guidance for different areas and different driver groups, improves the effectiveness of guidance information, avoids excessive concentration in certain areas, and enhances the overall traffic management effect.

[0041] Furthermore, generating differentiated information push strategies based on the aforementioned information diffusion coefficient includes:

[0042] The target information coverage rate for this guidance is determined based on the information diffusion coefficient; based on the target information coverage rate, detour guidance information is partially pushed to drivers on the road network of the target area.

[0043] Preferably, the target information coverage rate for this guidance information release is first determined based on the information diffusion coefficient calculated at the current moment. Specifically, the information diffusion coefficient is used as a control parameter for the information diffusion intensity of the current round, and the expected target information coverage rate after this guidance is calculated in conjunction with the current information coverage rate. For example, if the current information coverage rate is η and the information diffusion coefficient is β, then the target information coverage rate for this guidance can be expressed as: Target information coverage rate = η + β × (1 − η); where, the larger β is, the higher the proportion of drivers allowed to be newly covered in this round; the smaller β is, the more the guidance push is only performed on drivers within a smaller range in this round. In this way, the information release range can be adapted to the carrying capacity of the current detour route. After determining the target information coverage rate, the total number of drivers currently capable of receiving guidance information within the target area's road network is counted. Combined with the number of drivers who have already received guidance information, the number of target drivers to be added in this round is calculated. Drivers capable of receiving guidance information are limited to vehicles located within the target area's road network, traveling in a direction related to the congestion-affected area, and not yet having received guidance information in this round. The number of new drivers to be added can be determined based on the difference between the target information coverage rate and the current coverage level, thus clarifying the scale of this round's partial push. Next, specific push targets are selected from the candidate drivers, prioritizing drivers more likely to be affected by congestion and more suitable for detours. In cases where multiple detour routes exist, different drivers can be assigned to different push sets corresponding to different routes based on the remaining capacity of each detour route, achieving decentralized traffic management. Based on this, detour guidance information is sent to a selected group of drivers. This information includes congestion alerts, recommended detour routes, estimated time savings, and corresponding route guidance. For other drivers not included in this round of target coverage, no push notifications are sent, or only general traffic condition information is sent, thus achieving partial push rather than full broadcast. Finally, after completing this round of push notifications, driver responses to the guidance information and changes in road network traffic conditions can be monitored. The information diffusion coefficient and target information coverage are recalculated at the next time step to dynamically adjust the scope and target of subsequent push notifications, thereby improving the accuracy of guidance information dissemination and traffic management effectiveness while controlling detour traffic flow.

[0044] In summary, the embodiments of this application have at least the following technical effects:

[0045] First, multi-source traffic data of the target area's road network is acquired, and traffic congestion analysis is performed based on this data to extract congestion information. Then, historical driver behavior trajectory data is collected to construct a driver behavior model, quantifying driver guidance response characteristics and generating predicted information compliance probabilities. Next, based on the traffic congestion information, traffic diversion path analysis is performed on the target area's road network to generate detour paths, and the remaining capacity of these detour paths is collected. Then, an information diffusion coefficient is calculated by combining the predicted information compliance probabilities and the remaining capacity of the detour paths. This information diffusion coefficient characterizes the intensity at which guidance information is allowed to be disseminated into the road network at the current moment. Finally, a differentiated information push strategy is generated based on the information diffusion coefficient. This approach solves the technical problem that existing traffic diversion methods do not consider the driver's compliance probability with guidance information, leading to blind and untargeted guidance strategies that easily cause secondary congestion on detour paths. It achieves the technical effect of improving the accuracy and effectiveness of traffic guidance information by calculating the information diffusion coefficient and generating differentiated push strategies, thereby avoiding excessive guidance that causes new congestion.

[0046] Example 2, based on the same inventive concept as the regional traffic congestion collaborative mitigation method based on traffic flow prediction in the previous examples, such as... Figure 2 As shown, this application provides a regional traffic congestion collaborative mitigation device based on traffic flow prediction, wherein the device includes:

[0047] Congestion Analysis Module 11: Acquires multi-source traffic data of the target area road network, performs traffic congestion analysis based on the multi-source traffic data, and extracts traffic congestion information; Feature Quantization Module 12: Collects historical driver behavior trajectory data, constructs a driver behavior model, quantifies driver guidance response features, and generates predicted information compliance probabilities; Traffic Diversion Path Analysis Module 13: Based on the traffic congestion information, performs traffic diversion path analysis on the target area road network, generates detour paths, and collects the remaining capacity of the detour paths; Diffuse Coefficient Calculation Module 14: Combines the predicted information compliance probability and the remaining capacity of the detour paths to calculate the information diffuse coefficient, which is used to characterize the intensity of the guidance information allowed to be released into the road network at the current moment; Push Strategy Generation Module 15: Generates differentiated information push strategies based on the information diffuse coefficient.

[0048] Furthermore, the congestion analysis module 11 is used to perform the following methods:

[0049] Based on the multi-source traffic data, the traffic state prediction results of each road segment in the target area road network are analyzed; based on the traffic state prediction results, congestion risk points are identified, and the congestion location, congestion degree and expected duration corresponding to the congestion risk points are extracted as the traffic congestion information.

[0050] Furthermore, the feature quantization module 12 is used to perform the following method:

[0051] The characteristic parameters of each driver are extracted from the historical driver behavior trajectory data. The characteristic parameters include the proportion of drivers who actually follow the detour suggestions after receiving guidance information. Based on the characteristic parameters, the average information compliance probability of the group is calculated as the predicted information compliance probability.

[0052] Furthermore, the evacuation path analysis module 13 is used to perform the following methods:

[0053] Based on the congestion location in the traffic congestion information and the road network topology, the detour route is generated using the shortest path algorithm; the real-time traffic status of the detour route is collected, the remaining capacity is estimated, and the remaining capacity is used as the remaining capacity of the detour route.

[0054] Furthermore, the diffusion coefficient calculation module 14 is used to perform the following method:

[0055] The proportion of vehicles that have received guidance information in the road network at the current moment is obtained as the current information coverage rate; the information diffusion coefficient is calculated according to the remaining capacity of the detour path, the current information coverage rate, and the predicted information compliance probability through a preset mapping rule.

[0056] Furthermore, the diffusion coefficient calculation module 14 is used to perform the following method:

[0057] The mapping rule is to ensure that the induced transfer flow does not exceed the remaining capacity of the detour path, and to dynamically adjust the value of the information diffusion coefficient with the optimization objective of minimizing the total travel time of the road network.

[0058] Furthermore, the push strategy generation module 15 is used to perform the following method:

[0059] The target information coverage rate for this guidance is determined based on the information diffusion coefficient; based on the target information coverage rate, detour guidance information is partially pushed to drivers on the road network of the target area.

[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for regional traffic congestion collaborative relief based on traffic flow prediction, characterized in that, include: Acquire multi-source traffic data of the road network in the target area, perform traffic congestion analysis based on the multi-source traffic data, and extract traffic congestion information; Collect historical driver behavior trajectory data, construct driver behavior models, quantify driver guidance response characteristics, and generate predicted information compliance probabilities; Based on the traffic congestion information, traffic diversion path analysis is performed on the road network in the target area to generate detour paths and collect the remaining capacity of the detour paths. The information diffusion coefficient is calculated by combining the predicted information compliance probability and the remaining capacity of the detour path. The information diffusion coefficient is used to characterize the strength of the guidance information allowed to be released into the road network at the current moment. A differentiated information push strategy is generated based on the information diffusion coefficient. 2.The method of claim 1, wherein, Acquire multi-source traffic data of the road network in the target area, perform traffic congestion analysis based on the multi-source traffic data, and extract traffic congestion information, including: Based on the analysis of the multi-source traffic data, the traffic status prediction results of each road segment in the target area road network are obtained; Based on the traffic condition prediction results, congestion risk points are identified, and the congestion location, congestion level, and expected duration corresponding to the congestion risk points are extracted as the traffic congestion information. 3.The method of claim 1, wherein, Historical driver behavior trajectory data is collected to construct a driver behavior model, quantify the driver's induced response characteristics, and generate predicted compliance probabilities, including: The characteristic parameters of each driver are extracted from the historical driver behavior trajectory data, including the proportion of drivers who actually followed the detour suggestions after receiving guidance information; Based on the aforementioned feature parameters, the average information compliance probability of the group is statistically calculated and used as the predicted information compliance probability. 4.The method of claim 1, wherein, Based on the traffic congestion information, traffic diversion path analysis is performed on the road network in the target area to generate detour routes and collect the remaining capacity of the detour routes, including: Based on the congestion location in the traffic congestion information and combined with the road network topology, the detour route is generated using the shortest path algorithm. The real-time traffic status of the detour route is collected, the remaining capacity is estimated, and the remaining capacity is used as the remaining capacity of the detour route. 5.The method of claim 1, wherein, The information diffusion coefficient is calculated by combining the predicted information compliance probability and the remaining capacity of the detour path, including: The current information coverage rate is calculated as the proportion of vehicles in the road network that have received guidance information at the current moment. The information diffusion coefficient is calculated based on the remaining capacity of the detour path, the current information coverage, and the predicted information compliance probability, using a preset mapping rule. 6.The method of claim 5, wherein, The mapping rule is to ensure that the induced transfer flow does not exceed the remaining capacity of the detour path, and to dynamically adjust the value of the information diffusion coefficient with the optimization objective of minimizing the total travel time of the road network. 7.The method of claim 1, wherein, Based on the information diffusion coefficient, a differentiated information push strategy is generated, including: The target information coverage rate for this induction is determined based on the information diffusion coefficient. Based on the target information coverage, partial detour guidance information is pushed to drivers on the road network in the target area.

8. A regional traffic congestion coordination and mitigation device based on traffic flow prediction, characterized in that, The method for coordinated regional traffic congestion mitigation based on traffic flow prediction as described in any one of claims 1-7 includes: Congestion Analysis Module: Acquires multi-source traffic data of the road network in the target area, performs traffic congestion analysis based on the multi-source traffic data, and extracts traffic congestion information; Feature quantization module: Collects historical driver behavior trajectory data, constructs driver behavior models, quantifies driver guidance response features, and generates predicted information compliance probabilities; Traffic diversion path analysis module: Based on the traffic congestion information, it performs traffic diversion path analysis on the road network in the target area, generates detour paths, and collects the remaining capacity of the detour paths; The diffusion coefficient calculation module calculates the information diffusion coefficient by combining the predicted information compliance probability and the remaining capacity of the detour path. The information diffusion coefficient is used to characterize the intensity of the guidance information allowed to be released into the road network at the current moment. Push strategy generation module: Generates differentiated information push strategies based on the information diffusion coefficient.

Images