An intelligent traffic management system based on driver yielding intention analysis

Abstract

The application discloses an intelligent traffic management system based on driver yielding willingness analysis, and belongs to the technical field of intelligent traffic management, and comprises a data acquisition module, a data processing module, a model construction module, an analysis verification module and a decision output module.Through the technical scheme of data fusion-model construction-dynamic response, the three problems of data one-sidedness, model limitation and intervention passivity in the prior art are directly solved, and accurate analysis and active control of the yielding willingness of drivers at a non-signalized intersection are realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent traffic management, and particularly relates to an intelligent traffic management system based on driver yielding willingness analysis. BACKGROUND

[0002] Existing researches mainly focus on the safety of pedestrian crossing and the macro analysis of vehicle-pedestrian conflict, and explore the influencing factors of driver yielding behavior through quantitative models. For example, research based on Logit model shows that vehicle speed, traffic density and the number of pedestrians significantly affect the driver's yielding decision; the Bayesian multilevel logistic regression model reveals the dynamic characteristics of the pedestrian crossing behavior over time. In addition, the improvement measures of road facility design (such as trapezoidal speed hump and self-luminous pedestrian crossing) are proved to reduce vehicle speed and reduce the phenomenon of rushing through by physical intervention. Some researches also start from the individual characteristics of drivers (such as gender and personality type), and use Logistic regression model to analyze the influence on waiting time threshold, which provides data support for the safety of pedestrian crossing. These achievements lay a theoretical foundation for understanding the interaction mechanism of vehicle-pedestrian and optimizing traffic management strategy.

[0003] Although the existing researches have made certain progress, there are still significant limitations. First, existing analysis is mainly concentrated on urban roads or signal-controlled intersections, and there is a serious lack of research on the special scene of non-signalized intersection on village road, and the driver behavior pattern and conflict mechanism of the latter have unique characteristics due to mixed traffic, complex road conditions and weak infrastructure. Second, although the existing model can identify the influence of external environmental variables (such as the number of lanes and pedestrian behavior), the exploration of subjective psychological factors of drivers (such as risk perception, sensation seeking and moral constraints) is very limited, especially the lack of systematic analysis combined with psychological theories (such as the theory of planned behavior and deterrence theory). In addition, the existing road facility optimization scheme (such as speed hump) mainly relies on passive intervention, and cannot form a closed loop with the dynamic decision-making process of drivers, resulting in limited actual effect of the measures. SUMMARY

[0004] To solve the above technical problems, the application provides an intelligent traffic management system based on driver yielding willingness analysis to solve the problems existing in the prior art.

[0005] To achieve the above purpose, the application provides an intelligent traffic management system based on driver yielding willingness analysis, which comprises:

[0006] A data acquisition module is configured to acquire original data affecting the yielding willingness of drivers through a vehicle terminal, a mobile terminal and a roadside sensing device, wherein the original data comprises driver behavior data, pedestrian dynamic information and environmental parameters.

[0007] a data processing module, configured to clean and normalize the original data to generate a structured data set;

[0008] a model construction module, configured to construct a structural equation model of driver yielding intention based on latent variables and the structured data set;

[0009] an analysis and verification module, configured to perform reliability test, validity test and fitting degree verification on the structural equation model, and output key influence factors of driver yielding intention;

[0010] a decision output module, configured to generate a dynamic warning instruction according to the key influence factors and send the dynamic warning instruction to a roadside interaction device of the unsignalized intersection.

[0011] Preferably, the data collection module comprises:

[0012] an on-board OBD interface, configured to collect driver behavior data in real time, the driver behavior data including vehicle speed, brake frequency and steering angle data;

[0013] a roadside infrared camera, configured to monitor pedestrian dynamic information, the pedestrian dynamic information including street crossing density, distraction behavior and pedestrian quantity;

[0014] an environmental sensor, configured to collect environmental parameters, the environmental parameters including road grade, weather condition and traffic density.

[0015] Preferably, the data processing module comprises:

[0016] a normalization processing unit, configured to perform normalization processing on the original data to eliminate dimensional differences;

[0017] an abnormal data elimination unit, configured to detect and eliminate abnormal driving behavior data by an isolation forest algorithm to obtain cleaned data;

[0018] a data mapping unit, configured to map the cleaned data to a preset latent variable dimension, the latent variable dimension including pedestrian street crossing intention, external environment perception, behavior intention, subjective norm, travel experience, punishment mechanism, perceived behavior control and behavior attitude.

[0019] Preferably, the model construction module comprises:

[0020] a measurement model unit, configured to define a corresponding relationship between observation variables and latent variables;

[0021] a structural model unit, configured to define path assumptions between latent variables and calculate path coefficients by a maximum likelihood estimation method.

[0022] Preferably, the analysis and verification module:

[0023] a reliability verification unit configured to calculate Cronbach's α coefficient and composite reliability to complete reliability test;

[0024] a confirmatory factor analysis unit configured to obtain discriminant validity and convergent validity through confirmatory factor analysis;

[0025] a goodness-of-fit analysis unit configured to compare GFI, RMSEA and CFI indexes with preset threshold values to verify model adaptability.

[0026] Preferably, the decision output module comprises:

[0027] a first decision unit configured to trigger a roadside LED warning screen to display dynamic text prompts when detecting pedestrian distraction behavior;

[0028] a second decision unit configured to control the flashing frequency of the self-luminous pedestrian crossing LED floor tiles according to pedestrian crossing density;

[0029] a third decision unit configured to push voice warning information to the driver's vehicle terminal through V2X communication.

[0030] Preferably, the system further comprises:

[0031] an edge computing node deployed near the village road intersection, configured to process sensor data in real time and run a lightweight structural equation model;

[0032] a cloud management platform configured to store historical data, optimize model parameters and generate a safety evaluation report.

[0033] Preferably, the edge computing node comprises:

[0034] a first optimization unit configured to adjust the weight of latent variables according to real-time pedestrian crossing behavior data;

[0035] a second optimization unit configured to update the path coefficient of the penalty mechanism through an incremental learning algorithm.

[0036] Preferably, the system further comprises:

[0037] a self-luminous pedestrian crossing device whose luminous intensity is dynamically adjusted according to the driver's willingness to yield prediction result;

[0038] a trapezoidal speed hump linked with the decision output module to raise the hump height to force speed reduction when detecting low willingness to yield.

[0039] Preferably, the closed-loop control of the system comprises:

[0040] a coefficient correction unit configured to correct the path coefficient of the structural equation model according to the actual execution rate of the driver's yielding behavior;

[0041] a dynamic optimization unit for optimizing the dynamic warning strategy by a reinforcement learning algorithm.

[0042] Compared with the prior art, the present application has the following advantages and technical effects:

[0043] The application provides an intelligent traffic management system based on driver yielding intention analysis, comprising: a data acquisition module for acquiring original data affecting driver yielding intention through a vehicle terminal, a mobile terminal and a roadside sensing device, wherein the original data comprises driver behavior data, pedestrian dynamic information and environmental parameters; a data processing module for cleaning and standardizing the original data to generate a structured data set; a model construction module for constructing a structural equation model of driver yielding intention based on latent variables and the structured data set; an analysis and verification module for performing reliability test, validity test and fitting degree verification on the structural equation model, and outputting key influence factors affecting driver yielding intention; and a decision output module for generating a dynamic warning instruction according to the key influence factors and sending the dynamic warning instruction to a roadside interaction device of a non-signalized intersection.

[0044] In view of the technical problem that the prior art lacks comprehensive collection of driver subjective psychological factors and external environmental variables, resulting in single analysis dimension, the application integrates a vehicle terminal, a roadside sensing device and a mobile terminal through a data acquisition module, and the system first realizes synchronous collection of driver psychological characteristics (such as attitude and norm) and environmental parameters (such as pedestrian distraction and road grade), thereby providing a data basis for multi-dimensional analysis.

[0045] In view of the technical problem that the prior model ignores driver psychological mechanisms (such as subjective norm and perceived behavioral control), resulting in insufficient explanatory power for yielding intention, the model construction module in the application defines latent variables (such as behavioral attitude and punishment mechanism) and their path relationships, and first introduces a psychological framework into the analysis of non-signalized intersections on village roads, thereby significantly improving the ability of the model to depict the decision logic of drivers.

[0046] In view of the technical problem that traditional facilities (such as speed hump) rely on passive intervention and cannot adapt to dynamic traffic scenarios, the decision output module in the application generates a warning instruction in real time according to the model analysis result, actively intervenes in high-risk scenarios directly, and reduces the probability of human-vehicle conflict.

[0047] The application directly solves the three problems of data one-sidedness, model limitation and passive intervention in the prior art through the technical scheme of data fusion-model construction-dynamic response, and realizes accurate analysis and active control of driver yielding intention at non-signalized intersections. BRIEF DESCRIPTION OF DRAWINGS

[0048] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrations, together with their description, serve to explain the application without limiting its aspects. In the drawings:

[0049] Figure 1 Amos-based structural equation model diagram of an embodiment of the present application;

[0050] Figure 2 System schematic diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0051] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0052] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a group of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0053] Embodiment one

[0054] As shown in Figures 1-2 , the present embodiment provides an intelligent traffic management system based on driver yielding intention analysis, which includes five modules of data acquisition module, data processing module, model construction module, analysis verification module and decision output module, and specifically includes:

[0055] The data acquisition module is used to obtain original data affecting the driver yielding intention through the vehicle terminal, mobile terminal and roadside sensing device, and the original data includes driver behavior data, pedestrian dynamic information and environmental parameters.

[0056] Further, the data acquisition module includes:

[0057] The vehicle OBD interface is used to collect driver behavior data in real time, and the driver behavior data includes vehicle speed, brake frequency and steering angle data.

[0058] The roadside infrared camera is used to monitor pedestrian dynamic information, and the pedestrian dynamic information includes street crossing density, distraction behavior and pedestrian quantity.

[0059] The environmental sensor is used to collect environmental parameters, and the environmental parameters include road grade, weather condition and traffic density.

[0060] The data processing module is used to clean and standardize the original data to generate a structured data set.

[0061] Further, the data processing module comprises:

[0062] a normalization processing unit configured to normalize the original data to eliminate dimension differences;

[0063] an abnormal data elimination unit configured to detect and eliminate abnormal driving behavior data by an isolation forest algorithm to obtain cleaned data;

[0064] a data mapping unit configured to map the cleaned data to a preset latent variable dimension, the latent variable dimension comprising a pedestrian crossing willingness, an external environment perception, a behavior intention, a subjective norm, a travel experience, a punishment mechanism, a perceived behavior control, and a behavior attitude.

[0065] Specifically, eight latent variables for the structural equation model are: a behavior intention (BI), a pedestrian crossing willingness (WTC), a subjective norm (SN), an external environment perception (SP), a travel experience or habit strength (TE), a punishment mechanism (PM), a perceived behavior control (PBC), and a behavior attitude (AB). These latent variables and corresponding observed variables are structured data sets, which are listed in detail in Table 1.

[0066] Table 1

[0067]

[0068] In Table 1, the subjective norm specifically refers to the expectations and pressures from the surrounding environment (such as family, friends) and traffic rules, social morality, etc. on the driver's yielding behavior when deciding whether to yield to pedestrians. The perceived behavior control reflects the driver's subjective assessment of the difficulty of implementing the yielding behavior, which takes into account personal driving skills, experience, and external traffic conditions (e.g., road conditions, traffic signals, pedestrian dynamics).

[0069] a model construction module configured to construct a structural equation model of the driver's yielding intention based on the latent variables and the structured data set;

[0070] Further, the model construction module comprises:

[0071] a measurement model unit configured to define the correspondence between observed variables and latent variables;

[0072] a structural model unit configured to define path assumptions between latent variables and calculate path coefficients by maximum likelihood estimation.

[0073] Specifically, the embodiment proposes path assumptions between latent variables:

[0074] H1: SN→AB The subjective norm positively influences the behavior attitude;

[0075] H2: TE→AB Travel experience or habit strength positively influences behavior attitude;

[0076] H3: AB→BI Behavior attitude positively influences behavior intention;

[0077] H4: PBC→AB Perceived behavior control positively influences behavior attitude;

[0078] H5: PM→TE Punishment mechanism positively influences travel experience and habit;

[0079] H6: PM→SN Punishment mechanism positively influences subjective norm;

[0080] H7: PM→BI Punishment mechanism positively influences behavior intention;

[0081] H8: SP→SN Surrounding environment perception positively influences subjective norm;

[0082] H9: SP→PBC Surrounding environment perception positively influences perceived behavior control;

[0083] H10: SP→BI Surrounding environment perception positively influences behavior intention;

[0084] H11: WTC→AB Pedestrian crossing willingness positively influences driver yielding attitude;

[0085] H12: WTC→SN Pedestrian crossing willingness positively influences subjective norm.

[0086] The structural equation model is as shown in FIG. 1. Figure 1 As shown in FIG. 1, the pedestrian crossing willingness (WTC), the punishment mechanism (PM) and the surrounding environment perception (SP) are defined as exogenous latent variables, which are starting points / independent variables of the model and have initial influences on the internal dynamics of the system. The rest of the latent variables, including the behavior intention (BI), the subjective norm (SN), the travel experience or habit strength (TE), the perceived behavior control (PBC) and the behavior attitude (AB), are regarded as endogenous latent variables, which are effects / dependent variables and their states are influenced and restricted by other variables in the model.

[0087] The analysis and verification module is configured to perform reliability test, validity test and fitting degree verification on the structural equation model, and output key influence factors of the driver yielding willingness;

[0088] Further, the analysis and verification module comprises:

[0089] The reliability verification unit is configured to calculate the Cronbach's a coefficient and the combined reliability to complete the reliability test.

[0090] Specifically, to ensure the reliability and stability of the structured data set, a rigorous reliability test was conducted. The data presented in Table 2 shows that the Cronbach's alpha coefficient and the composite reliability (CR) of each dimension exceed the threshold of 0.7, and the average variance extracted value (AVE) is also higher than 0.5. These indicators have reached the recognized standard in the academic community. This result has proven that the structured data set has good reliability in each dimension, laying a solid foundation for the subsequent factor analysis.

[0091] Table 2

[0092]

[0093] The confirmatory factor analysis unit is used to obtain discriminant validity and convergent validity through confirmatory factor analysis;

[0094] Specifically, before conducting factor analysis, the appropriateness of the data set was comprehensively tested. With the help of Bartlett's Sphericity Test and Kaiser-Meyer-Olkin (KMO) measure tools, the results presented in Table 3 clearly show that the analyzed data set exhibits high factor analysis suitability (specifically, the KMO value exceeds the benchmark of 0.6, reaching the "excellent" level; at the same time, the p-value obtained by Bartlett's test is less than 0.05, which has statistical significance). This finding not only reveals the possibility of shared factors among variables, but also strongly proves that the sample size is sufficient to support the effective conduct of subsequent factor analysis.

[0095] Table 3

[0096]

[0097] Factor analysis uses the principal component analysis (PCA) technique to analyze the underlying structure hidden behind the questionnaire items, thereby simplifying the data and improving the analysis efficiency. PCA reduces the data dimension to maximize the explanation of the data variance, and strictly follows the Kaiser criterion to retain only the factors with eigenvalues greater than 1 to ensure the effectiveness of the analysis. To further enhance the interpretability of the results, the Varimax rotation method is used to optimize and adjust the factor loadings. The factor loading matrix clearly shows the distribution of each variable on different factors, providing a strong basis for the naming of factors. Finally, the results of factor analysis not only clarify the dimensions of measurement, but also lay a solid foundation for subsequent hypothesis testing and in-depth explanation of phenomena.

[0098] Table 4

[0099]

[0100] The Variance Extraction Rate (VER) is a measure that quantifies the degree to which the extracted factors explain the variance of the original dataset variables, thereby assessing the representativeness of the factors. As the data presented in Table 4 shows, the Variance Extraction Rate obtained in this analysis exceeds the threshold of 50%, a result that strongly indicates the effectiveness of the factor extraction process and its good representativeness of the dataset.

[0101] Table 5

[0102]

[0103] The overview of the rotated component matrix is shown in Table 5. After the maximum variance rotation process, the relationship between the factors and each research content is clearly defined. It is worth noting that the communality of all research items exceeds the threshold of 0.4, which strongly proves the close relationship between the factors and the research content, and also verifies the effectiveness of the information extraction process.

[0104] Confirmatory Factor Analysis (CFA) plays a crucial role in the construction and validation of SEM. Its core purpose is to rigorously test whether the correspondence between observed indicators and latent variables fits the pre-established theoretical framework, thereby ensuring the accuracy and effectiveness of the measurement tool. The analysis scope of CFA covers two important dimensions of validity assessment: Discriminant Validity and Convergent Validity. Discriminant Validity aims to verify whether different latent variables can maintain their conceptual independence and distinguishability, while Convergent Validity focuses on ensuring that a group of observed variables consistently and reliably reflects the core characteristics of the corresponding latent variable. With the help of CFA, this research conducts a comprehensive and systematic fit assessment of the measurement model, not only investigating the adaptation degree between the model and the data, but also laying a solid and reliable foundation for subsequent structural model analysis. This process not only enhances the reliability and validity of the research results, but also further improves the explanatory power and predictive power of the theoretical model in practical applications.

[0105] Table 6

[0106]

[0107] The results of the discriminant validity analysis, as shown in Table 6, show that the square root of the AVE of each dimension in the questionnaire, i.e. the diagonal values, are consistently higher than the other non-diagonal values in the column. This result strongly proves that the questionnaire performs well in terms of discriminant validity, meaning that different dimensions can be clearly distinguished from each other, each independently measuring different constructs, thereby ensuring the effectiveness and accuracy of the measurement tool.

[0108] Table 7

[0109]

[0110] The convergent validity analysis is an important step in evaluating the internal consistency of each dimension of the questionnaire and the degree of correlation between them. When the correlation between items is high, it means that they have high consistency in measuring the same latent concept, thereby enhancing the reliability and robustness of the study. On the contrary, if the correlation is low, the design rationality of the items and the measurement method used need to be carefully examined to identify and correct potential biases or misunderstandings. In the specific evaluation, the composite reliability CR is a key indicator for measuring the consistency and stability of the measurement tool, and its value should not be less than 0.6 to ensure the reliability of the measurement. On the other hand, AVE is used to evaluate internal consistency and detect common method bias. When the AVE value reaches or exceeds 0.5, it indicates that the internal consistency of the measurement dimension is good, thereby verifying the high efficiency of the measurement tool. The results of the convergent validity analysis, as shown in Table 7, show that the CR and AVE values of all dimensions in this study have reached the established standard threshold. This result strongly proves that the questionnaire performs excellently in terms of convergent validity, and the data collected is not only accurate and reliable, but also provides solid support for the conclusions of the study.

[0111] The fit analysis unit is used to compare the GFI, RMSEA, and CFI indicators with the preset threshold to verify the model fit.

[0112] Specifically, the Amos 26.0 software was used to conduct in-depth analysis of the constructed structural equation model. The model fit analysis results, as shown in Table 8, show that all the fit indicators of the model fall within the acceptable range through comprehensive consideration of the results. This finding strongly proves the rationality and effectiveness of the model construction. Based on this, the research can smoothly proceed to the subsequent path analysis stage.

[0113] Table 8

[0114]

[0115] This example conducts a comprehensive test of the preset 12 hypotheses. The path coefficients are shown in Table 9, which details the relationship framework of these hypotheses, the corresponding standardized regression coefficients β, p values, and other key statistical indicators, as well as the explicit judgment of whether each hypothesis is supported by the data based on these indicators.

[0116] Table 9

[0117]

[0118] The non-standardized coefficient results show that all path coefficients are greater than 0, there is a positive influence between variables, and the P value is less than 0.05, indicating that the influence is significant.

[0119] All latent variables affect the willingness of drivers to yield to pedestrians at unsignalized intersections. Among these factors, the distraction behavior of pedestrians, the existence of a punishment mechanism, and the behavior attitude of drivers are identified as the most critical factors. There is a strong correlation between travel habits and behavior attitudes, which indicates that drivers' daily travel patterns are closely related to their inherent behavior attitudes. Similarly, there is a significant correlation between the perception of the external environment and perceived behavioral control, which reveals how environmental factors affect the decision-making process of drivers. In addition, the link between the willingness of pedestrians to cross the street and the behavior attitude of drivers cannot be ignored, which further emphasizes the complexity of the interaction between the two parties.

[0120] Subjective norms have a significant positive impact on behavior attitude, which strongly supports hypothesis H1. Specifically, it reveals the important role of external pressure factors such as moral norms in shaping the willingness of drivers to yield on rural roads. This indicates that when drivers perceive strong expectations and norms from society, they are more likely to exhibit positive yielding behavior. Therefore, to further improve the level of traffic safety on rural roads, it is particularly important to strengthen relevant publicity and education work. By enhancing the social responsibility of drivers, they can be more conscious of obeying traffic rules, thereby forming a more harmonious and safe traffic environment on rural roads.

[0121] Travel experience or habit strength has a significant positive impact on behavior attitude, which provides strong support for hypothesis H2. Specifically, it reveals that drivers with good driving habits are more likely to exhibit a positive and responsible yielding attitude when faced with pedestrians crossing the road. Given this, strengthening driver training and education to improve their driving skills and cultivate safer driving habits is an important strategy to enhance traffic safety on rural roads. This includes teaching drivers how to strictly control their speed when encountering pedestrians crossing, avoiding speeding, and enhancing their ability to observe dynamic conditions at intersections and zebra crossings, as well as actively taking measures to slow down and avoid.

[0122] Behavior attitude has a positive impact on behavior intention, which provides strong support for hypothesis H3. This conclusion is highly consistent with the core view of the theory of planned behavior, further verifying the applicability of the theory in practical situations. Specifically, it reveals that drivers' positive attitude towards yielding behavior can significantly improve their willingness to take yielding actions in actual driving.

[0123] Perceived behavioral control has a significant positive effect on behavioral attitude, which provides strong support for hypothesis H4. Specifically, when drivers have weak self-control, such as thinking that it is easy to cut in and difficult to control not to cut in, their willingness to yield will decrease significantly. This phenomenon is highly consistent with Bandura's self-efficacy theory, which emphasizes the key role of individuals' beliefs in their ability to make behavioral choices. The lack of perceived behavioral control reflects the low self-restraint ability of drivers, which further affects their formation of positive yielding attitude. Therefore, in order to improve the willingness of drivers to yield, it is necessary to enhance their safe driving cognition and self-control ability through systematic education and training. Such training should focus on improving the self-discipline of drivers when faced with various temptations, enabling them to more firmly adhere to traffic rules.

[0124] The punishment mechanism, as an important external constraint means, plays a significant role in regulating drivers' yielding behavior. The analysis results show that the punishment mechanism has a significant positive effect on travel habits, subjective norms, and behavioral intentions, which provides strong support for hypotheses H5, H6, and H7. Specifically, positive reward measures can encourage drivers to repeatedly adopt yielding behavior, while negative punishment can effectively curb the occurrence of illegal behavior.

[0125] The perception of the external environment has a significant impact on subjective norms, perceived behavioral control, and behavioral intentions. Specifically, when drivers' perception of the external environment decreases, their willingness to yield also decreases, which provides strong support for hypotheses H8, H9, and H10. Given the key role of environmental perception in shaping drivers' yielding behavior, improving the traffic environment is an important way to enhance drivers' willingness to yield and improve traffic safety. Specifically, measures such as adding traffic signs and optimizing road facilities can be taken to achieve this goal. For example, setting up road barriers in areas with dense traffic to improve drivers' alertness and protection for pedestrians; in areas with sparse traffic, natural barriers can be used for protection, and dedicated pedestrian crosswalks and clear stop-and-yield signs can be set up.

[0126] The decrease of the pedestrian crossing intention can significantly weaken the driver's yielding intention, and further affect their behavior attitude and subjective norms, which provides strong support for the hypotheses H11 and H12. Among them, the distraction behavior of pedestrians is a key factor that cannot be ignored, which often comes from the neglect of the traffic environment or excessive self-confidence in their own safety. According to the attention allocation theory, the reaction speed of pedestrians to key information such as traffic signals and vehicle approach will be significantly weakened in the state of distraction, which undoubtedly increases the risk of traffic accidents. Therefore, in order to improve the overall safety level of road traffic, it is necessary to pay high attention to the cultivation of pedestrian traffic safety awareness and the improvement of concentration. Specifically, we should strengthen propaganda and education to make pedestrians fully realize the danger of distraction crossing, and guide them to develop good habits of concentrating on crossing. At the same time, we should also standardize the crossing behavior of pedestrians to ensure that they strictly follow the traffic rules when crossing the road, so as to reduce the probability of traffic accidents.

[0127] The decision output module is configured to generate dynamic warning instructions based on the key influence factors and send the dynamic warning instructions to a roadside interaction device of the unsignalized intersection.

[0128] Further, the decision output module comprises:

[0129] A first decision unit is configured to trigger a roadside LED warning screen to display dynamic text prompts when detecting pedestrian distraction behavior.

[0130] A second decision unit is configured to control the flashing frequency of the self-luminous pedestrian crossing based on the pedestrian crossing density.

[0131] A third decision unit is configured to push voice warning information to a driver's vehicle terminal through V2X communication.

[0132] In this embodiment, the system further comprises an edge computing node and a cloud management platform.

[0133] The edge computing node is deployed near the intersection of the village road and is configured to process sensor data in real time and run a lightweight structural equation model.

[0134] Further, the edge computing node comprises:

[0135] A first optimization unit is configured to adjust the weight of the latent variable based on real-time pedestrian crossing behavior data.

[0136] A second optimization unit is configured to update the path coefficient of the penalty mechanism through an incremental learning algorithm.

[0137] The cloud management platform is configured to store historical data, optimize model parameters, and generate a safety evaluation report.

[0138] In this embodiment, the system further comprises a self-luminous pedestrian crossing device and a trapezoidal speed hump;

[0139] The self-luminous pedestrian crossing device has a luminous intensity dynamically adjusted according to the driver's yielding intention prediction result;

[0140] The trapezoidal speed hump, in linkage with the decision output module, raises the hump height to force speed reduction when detecting a low yielding intention.

[0141] The closed-loop control of the system comprises:

[0142] The coefficient correction unit is configured to correct the path coefficient of the structural equation model according to the actual execution rate of the driver's yielding behavior;

[0143] The dynamic optimization unit is configured to optimize the dynamic warning strategy by using a reinforcement learning algorithm.

[0144] The embodiment has the following beneficial effects:

[0145] The embodiment constructs a structural equation model of the driver's yielding intention at a non-signalized intersection on a village road. Through reliability, validity test, confirmatory factor analysis, model fitting degree evaluation and path analysis, the system verifies the relationship between each variable in the model and the overall effectiveness. In addition, subjective norms, travel habits, behavior attitudes, perceived behavior control and punishment mechanism all significantly affect the yielding intention. In addition, external environmental perception and pedestrian crossing intention are also important factors, especially pedestrian distraction behavior, punishment mechanism and behavior attitude.

[0146] The improvements in the embodiment comprehensively consider the driver's behavior attitude, subjective norms, perceived behavior control and external environmental perception and other factors, and strive to optimize the traffic environment, strengthen publicity and education, improve the awareness and concentration of pedestrian traffic safety and other multi-dimensional means, to comprehensively promote the yielding behavior of the driver and create a safer and more orderly road traffic environment. The embodiment expects to provide valuable reference and guidance for improving the traffic safety level of the village road.

[0147] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent traffic management system based on driver yielding intention analysis, characterized in that, Comprising: a data acquisition module for obtaining original data affecting the driver's yielding intention through a vehicle-mounted terminal, a mobile terminal, and a roadside sensing device, the original data including driver behavior data, pedestrian dynamic information, and environmental parameters; the data acquisition module comprising: a vehicle-mounted OBD interface for real-time acquisition of driver behavior data, the driver behavior data including vehicle speed, brake frequency, and steering angle data; a roadside infrared camera for monitoring pedestrian dynamic information, the pedestrian dynamic information including street crossing density, distraction behavior, and the number of pedestrians; an environmental sensor for acquiring environmental parameters, the environmental parameters including road grade, weather conditions, and traffic density; a data processing module for cleaning and standardizing the original data to generate a structured data set; the data processing module comprising: a normalization processing unit for normalizing the original data to eliminate dimensional differences; an abnormal data elimination unit for detecting and eliminating abnormal driving behavior data through an isolation forest algorithm to obtain cleaned data; a data mapping unit for mapping the cleaned data to a pre-set latent variable dimension, the latent variable dimension including pedestrian crossing intention, external environment perception, behavior intention, subjective norm, travel experience, punishment mechanism, perceived behavior control, and behavior attitude; a model construction module for constructing a structural equation model of the driver's yielding intention based on the latent variable and the structured data set; an analysis and verification module for performing reliability testing, validity testing, and fitting degree verification on the structural equation model, and outputting key influencing factors affecting the driver's yielding intention; a decision output module for generating dynamic warning instructions based on the key influencing factors and sending them to roadside interaction devices at unsignalized intersections; a self-luminous pedestrian crossing device with luminous intensity dynamically adjusted according to the driver's yielding intention prediction results; a trapezoidal speed hump linked with the decision output module to increase the hump height to force speed reduction when low yielding intention is detected; the closed-loop control of the system comprising: a coefficient correction unit for correcting the path coefficients of the structural equation model based on the actual execution rate of the driver's yielding behavior; a dynamic optimization unit for optimizing the dynamic warning strategy through a reinforcement learning algorithm.

2. The system according to claim 1, wherein the model construction module comprises: a measurement model unit for defining the correspondence between observed variables and latent variables; a structural model unit for defining path assumptions between latent variables and calculating path coefficients through maximum likelihood estimation.

3. The system according to claim 1, wherein the analysis and verification module comprises: a reliability verification unit for calculating Cronbach's α coefficient and combined reliability to complete reliability testing; a confirmatory factor analysis unit for obtaining discriminant validity and convergent validity through confirmatory factor analysis; a fitting degree analysis unit for comparing GFI, RMSEA, and CFI indexes with pre-set threshold values to verify model adaptability.

4. The system according to claim 1, wherein the decision output module comprises: A first decision unit is configured to trigger a roadside LED warning screen to display a dynamic text prompt when a distracted pedestrian behavior is detected. A second decision unit is configured to control the flashing frequency of the self-luminous pedestrian crossing LED tiles according to the pedestrian crossing density. A third decision unit is configured to push voice warning information to the driver's vehicle terminal through V2X communication.

5. The system of claim 1, further comprising: An edge computing node deployed near the intersection of the rural road, configured to process sensor data in real time and run a lightweight structural equation model; A cloud management platform for storing historical data, optimizing model parameters, and generating safety evaluation reports.

6. The system of claim 5, wherein the edge computing node comprises: A first optimization unit configured to adjust the latent variable weight according to real-time pedestrian crossing behavior data; A second optimization unit configured to update the path coefficient of the penalty mechanism through an incremental learning algorithm. ​ ​

Images