Intelligent AI highway traffic safety early warning system

Abstract

The invention, which relates to the technical field of traffic safety, discloses an intelligent AI highway traffic safety early warning system comprising the following components: a data acquisition module, a data analysis and prediction module, an optimal route planning module, a scene simulation module and a coping strategy generation module. According to the invention, through integrating data prediction and scene simulation functions, the early warning capability of highway traffic safety is significantly improved, and the data acquisition module widely collects historical traffic data, real-time traffic flow data and meteorological condition data; the data analysis and prediction module performs accurate analysis and prediction on the multi-source data by using a deep learning algorithm, and can output congestion probabilities and accident risk probabilities of a specific road section in different time periods in real time, and the scene simulation module performs scene simulation according to hypothetical conditions of a traffic management department in combination with current traffic data and prediction results. A traffic flow simulation model is constructed for scene simulation, and the simulation result truly reflects the influence of different traffic events on the road traffic capacity.

Description

Technical Field

[0001] The present invention relates to the field of traffic safety technology, and in particular to an intelligent AI highway traffic safety early warning system. Background Art

[0002] With the continuous increase in traffic volume on expressways, traffic safety issues have increasingly become the focus of attention from all walks of life. As an important part of the modern transportation network, the smooth flow and safety of expressways are directly related to the travel safety of the people and the development of the economy and society. However, due to the complexity, variability and uncertainty of traffic flow, expressway traffic safety faces many challenges.

[0003] Traditional technologies have shortcomings. On the one hand, traditional prediction methods mainly rely on statistical analysis of historical traffic data and lack the full utilization of real-time traffic data, resulting in the prediction results being unable to accurately reflect changes in current traffic conditions. On the other hand, scenario simulation technology is often limited by the complexity of the model and the limitations of computing power, making it difficult to truly reflect the impact of different traffic events on road capacity.

[0004] To sum up, traditional intelligent road condition prediction and scenario simulation technologies have many shortcomings in responding to highway traffic safety warnings. In order to overcome these shortcomings, it is particularly important to develop an intelligent AI highway traffic safety warning system. Summary of the Invention

[0005] The purpose of this invention is to remedy the shortcomings of existing technologies and provide an intelligent AI highway traffic safety warning system. It can accurately process and analyze multi-source data by integrating advanced data collection, analysis and prediction technologies. At the same time, it also introduces a more realistic scenario simulation model to reflect the impact of different traffic events on road capacity.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an intelligent AI highway traffic safety warning system, which includes the following components: a data acquisition module, a data analysis and prediction module, an optimal route planning module, a scenario simulation module, and a response strategy generation module;

[0007] The data acquisition module collects multi-source data of the expressway, which includes historical traffic data, real-time traffic flow data and meteorological condition data. The historical traffic data covers the traffic volume, speed, accident occurrence and vehicle type information of different time periods and different sections, and is regularly acquired by establishing a data connection with the data storage system of the traffic management department. The real-time traffic flow data is collected by a variety of sensors installed on the expressway, including but not limited to geomagnetic sensors, radar sensors and video surveillance cameras. The sensors monitor the passing information of vehicles in real time and transmit the data to the system's data acquisition server through a high-speed communication network. The meteorological condition data is acquired by connecting to the data interface of the meteorological department. At the same time, meteorological monitoring stations are set up along key sections of the expressway to collect temperature, humidity, air pressure, wind speed, wind direction, precipitation intensity and visibility information. The collected meteorological data is stored in a temporally and spatially associated manner with the traffic data.

[0008] The data analysis and prediction module: This module analyzes and predicts the collected multi-source data based on a customized deep learning algorithm, pre-processes the collected data, and converts the original data into a format suitable for algorithm input, including normalizing the data so that data of different dimensions are at the same order of magnitude, and filling in missing values. It uses a time series-based interpolation method to fill in missing data points based on the data characteristics of the previous and next moments, and constructs a deep neural network model. The architecture of this model includes an input layer, multiple hidden layers, and an output layer, wherein the hidden layer adopts a hierarchical structure, and the number of neurons in each layer is dynamically adjusted according to the characteristic dimensions of the input data and the key information dimensions obtained from the historical data analysis. The input of the model is pre-processed historical traffic data, real-time traffic flow data, and meteorological condition data. The input data undergoes a series of linear and nonlinear transformations. The specific calculation process is as follows:

[0009]

[0010] Among them, x i is the i-th feature of the input data, w i is the weight of the i-th feature, b is the bias term, n is the number of input features, and f is the activation function, using a custom activation function This activation function combines linear and nonlinear parts and can better handle the complex nonlinear relationships in traffic data;

[0011] Weight w iThe determination process first assigns initial weights based on the historical impact of different data types. The historical traffic data are weighted according to the accident rate and congestion frequency of different road sections. Then, a hybrid optimization algorithm based on particle swarm optimization and genetic algorithm is used to optimize the weights. The weights are iteratively updated with the goal of minimizing the prediction error. The formula is as follows:

[0012]

[0013] in, is the weight of the i-th feature at the k-th iteration, is the velocity of the i-th feature at the k-th iteration, ω is the inertia weight that is dynamically adjusted according to the data update frequency and the degree of fluctuation of traffic flow, c1 and c2 are learning factors, r1 and r2 are random numbers in the range of [0, 1], is the local optimal weight of the i-th feature, gbest k is the global optimal weight, which is obtained by minimizing the prediction error function OK, y j is the actual observed value, is the model prediction value, m is the number of samples;

[0014] The final output is the congestion probability P of a specific road section at different time periods c and the accident risk probability P a , where the congestion probability P c By analyzing the vehicle density, average speed, lane occupancy rate, and traffic flow change rate of different road sections, combined with the congestion threshold in historical data, and using a calculation model based on fuzzy logic, the formula is obtained:

[0015] P c =F(ρ,v,l,q,T)

[0016] Where ρ is the vehicle density, v is the average speed, l is the lane occupancy rate, q is the traffic flow change rate, T is the time variable, and F is a custom fuzzy logic function. This function is trained based on historical data of different road sections and maps the input of different parameters to congestion probability values between [0, 1]. The accident risk probability P a By considering the influence of vehicle speed, distance between vehicles, vehicle type, weather conditions and road curvature, as well as historical accident data, a multi-factor weighted model is used for calculation. The formula is:

[0017] P a =G(v,d,m,w,c,H)

[0018] Where v is vehicle speed, d is vehicle distance, m is vehicle type, w is weather conditions, c is road curvature, H is historical accident data, and G is a custom multi-factor weighted function. This function is weighted and summed based on the contribution of each factor to the accident, and is trained and adjusted using historical accident data.

[0019] The optimal route planning module: uses the path planning algorithm to calculate the congestion probability P output by the data analysis and prediction module. c and the accident risk probability P a To plan the optimal route for travelers, we first construct a graph structure based on the road network topology and geographic information. The nodes represent the intersections or key locations of the road sections, the edges represent the road sections, and the weights of the edges are calculated based on the congestion probability P. c and the accident risk probability P a Set up, the weight calculation formula is W=k1P c +k2P a , where k1 and k2 are weight coefficients determined according to the importance of the road and user preferences. For major roads and users who prioritize safety, k1 is relatively small and k2 is relatively large. Then, an improved A* algorithm is used to search for the optimal path in the graph structure. During the search process, the heuristic function h(n) of the node is considered for path search. The heuristic function combines distance and risk factors. The formula is h(n) = k3d(n,goal) + k4R(n), where d(n,goal) is the distance from the current node n to the target node, R(n) is the comprehensive risk assessment on the path from the current node n to the target node, and k3 and k4 are coefficients determined according to user preferences and road characteristics. The algorithm finds the optimal route from the departure point to the destination and outputs it to the user's navigation device or mobile application;

[0020] The scenario simulation module receives hypothetical conditions input by the traffic management department, including but not limited to traffic accidents, road construction, severe weather, the type, location, start time, duration and severity of emergencies. Based on the input hypothetical conditions, combined with the current traffic data and the output of the data analysis and prediction module, the scenario simulation is performed by constructing a traffic flow simulation model. The simulation model considers the interaction between the vehicle's dynamic characteristics, driver behavior and road environment. The vehicle dynamic characteristics are modeled based on the vehicle's physical parameters and traffic regulations. The driver's behavior is simulated based on the driving habits in the force history data and the current road conditions. The road environment takes into account the road geometry, number of lanes, speed limit and road surface conditions. The specific simulation calculation process is as follows:

[0021]

[0022] Where v is the vehicle speed, t is time, a(t) is the vehicle acceleration, d is the distance between vehicles, w is the vehicle weight, s is the road slope, e is the road surface friction coefficient, and f is a function that comprehensively considers various factors. During the simulation process, the corresponding parameters are updated according to the assumptions. When a traffic accident occurs, the vehicle speed and distance between vehicles on the corresponding road section are adjusted accordingly. During road construction, the number of lanes and road surface conditions change, thereby simulating possible traffic impacts, including the spread and duration of traffic congestion, changes in traffic flow, and traffic delays.

[0023] The response strategy generation module generates corresponding response strategies based on the simulation results of the scenario simulation module. For different simulated traffic impact situations, response strategies are generated through a rule-based expert system. The rules are formulated based on traffic engineering theory and historical experience. When traffic congestion occurs, the scope and severity of the congestion determine whether it is necessary to adjust the signal light duration, open emergency lanes, and guide vehicle diversion strategies. When a traffic accident occurs, the dispatch routes and numbers of rescue vehicles and tow vehicles are determined based on the location and severity of the accident. When road construction occurs, the location and time of the construction determine whether temporary traffic signs, detour routes, and traffic control measures in the construction area need to be set up. The generated response strategies are output to the traffic management department and stored in the system's strategy database.

[0024] Furthermore, the storage method of historical traffic data in the data acquisition module adopts a distributed database system, which performs data shard storage according to different road sections and different time ranges, and stores the data on multiple nodes to improve the efficiency of data storage and query. Each data shard is encoded according to the geographical location and time range of the road section to facilitate rapid location and retrieval of data. At the same time, during the data storage process, the data is encrypted and an encryption algorithm based on blockchain technology is used to divide the data into multiple data blocks. Each data block contains a hash value and encryption information of the data. The chain structure of the blockchain ensures the integrity and security of the data and prevents data tampering. When querying data, the system automatically locates the corresponding data shard according to the road section and time range of the query request, and reads data from multiple nodes in parallel, reducing data query time and improving the response speed of the system. In addition, the data acquisition module also has a data backup function, which regularly backs up data to a remote data storage center. The backup data adopts a combination of incremental backup and full backup. The backup cycle and backup volume are determined according to the frequency and importance of data updates to ensure that data can be quickly restored in the event of hardware failure or system crash.

[0025] Furthermore, during the training process of the deep neural network model in the data analysis and prediction module, an adaptive learning rate adjustment strategy is introduced. The initial learning rate is set according to the scale and complexity of the data. When the training error of the model no longer decreases or decreases slowly, the learning rate is automatically adjusted. The specific adjustment formula is:

[0026]

[0027] Among them, α k is the learning rate at the kth iteration, α k+1 is the learning rate at the k+1th iteration, E k and E k+1 are the training errors at the kth and k+1th iterations, ∈ is the set error threshold, γ is the decrease factor, which is used when the error decreases less than the threshold, and β is the growth factor, which is used when the error increases or does not decrease. At the same time, in order to avoid overfitting, a new regularization method is introduced in the hidden layer, and a regularization term based on information entropy is added to the loss function. The formula is:

[0028] L=E+λH

[0029] Among them, L is the total loss function, E is the original prediction error loss function, λ is the regularization parameter, which is adjusted according to the noise level of the data and the complexity of the model, and H is the information entropy term, which is calculated as H = -∑ i p i log(p i ), p i It is the probability distribution of the model prediction output. Through this adaptive learning rate adjustment and new regularization method, the training efficiency and generalization ability of the model are improved, so that it can achieve better prediction performance under different traffic data distributions.

[0030] Furthermore, the weight optimization algorithm in the data analysis and prediction module is also combined with a simulated annealing algorithm to improve the optimization effect. During the optimization process, when the particle swarm optimization and genetic algorithm fall into the local optimum, the simulated annealing algorithm is introduced to search. The acceptance criterion of the simulated annealing algorithm is:

[0031]

[0032] Where ΔE is the energy difference, that is, the difference between the error of the new solution and the error of the current solution, T is the simulated annealing temperature, the initial temperature is determined according to the complexity of the data and the size of the search space, and as the number of iterations increases, the temperature decreases according to the cooling function T k+1 =αT kThrough this hybrid optimization algorithm, we can better avoid local optimality and find the global optimal weight, so that the model has stronger adaptability and accuracy when processing complex traffic data.

[0033] Furthermore, the improved A* algorithm in the optimal route planning module introduces a dynamic weight adjustment mechanism when searching for the optimal path. It dynamically adjusts the weight coefficients k3 and k4 in the heuristic function of the node based on real-time traffic information and user feedback. When the user finds new congestion or risk in the recommended route during driving, the system receives user feedback and updates the risk assessment of the node based on the current traffic information. If the user feedback shows that the actual congestion level of a certain section is more serious than the prediction, the system increases the risk assessment weight k4 of the section, recalculates the heuristic function, and re-searches the optimal path. At the same time, during the search process, different road section types, such as tunnels, bridges, and ordinary sections, are given different additional weights according to their safety risk levels. Tunnels and bridges have higher weights to ensure that the safety risks of special sections are fully considered when planning routes. The algorithm also takes into account the tidal phenomenon of traffic flow and adjusts the weight of the edge according to the traffic flow direction at different times. During peak traffic hours, the weight of sections in the main traffic flow direction is reduced to guide users to avoid congested sections, thereby improving the real-time and flexibility of route planning.

[0034] Furthermore, the traffic flow simulation model in the scenario simulation module also takes into account the characteristics of vehicle platooning. For platoons or large transport vehicle platoons, platoon dynamics are introduced into the model based on the communication and collaboration relationships between vehicles. The model considers the safe spacing, speed synchronization, and collaborative driving behavior between vehicles. When simulating platoon driving, the spacing and speed relationship between vehicles are adjusted according to the length, number, and type of vehicles in the platoon. The formula is:

[0035] d ij =f(n,m,v)

[0036] Among them, d ij is the distance between the i-th and j-th vehicles, n is the number of vehicles in the platoon, m is the vehicle type, v is the vehicle speed, and f is a spacing function determined according to vehicle performance and traffic regulations. The impact of information sharing on traffic flow is also considered. When the lead vehicle in the platoon receives road condition information, it transmits the information to subsequent vehicles through vehicle-to-vehicle communication, simulating the propagation speed and impact range of information within the platoon to more realistically simulate the behavior of traffic flow in the case of platooning and improve the authenticity and accuracy of scenario simulation.

[0037] Furthermore, the scenario simulation module models traffic flow noise during the simulation process. Traffic flow noise takes into account the random behavior of drivers, fluctuations in vehicle performance, and environmental uncertainty. This is simulated by adding random terms to the vehicle dynamics equations, specifically:

[0038]

[0039] Among them, ξ(t) is a random noise term, whose distribution is determined by statistical analysis of vehicle speed fluctuations and environmental changes in historical data. The distribution parameters of ξ(t) are different for different types of vehicles and different environmental conditions. In severe weather conditions, the variance of ξ(t) is larger to reflect the uncertainty of driver operation and the instability of vehicle performance. By adding the noise term, the scenario simulation results are made closer to the actual traffic conditions, avoiding over-idealization of the simulation results and improving the simulation results' ability to reflect real traffic.

[0040] Furthermore, the rule-based expert system in the response strategy generation module has a dynamic update function in its rule base. In addition to the initial rules formulated according to traffic engineering theory and historical experience, the system automatically updates and optimizes the rules based on actual traffic management decision-making results and scenario simulation results. When the traffic management department implements a response strategy and achieves certain results, the system adjusts the corresponding rules based on the implementation effect and traffic conditions. For example, the rules for adjusting the duration of traffic lights are updated according to the actual traffic flow changes and congestion relief, and the rules for dispatching rescue vehicles are adjusted according to the rescue efficiency and rescue time. At the same time, the expert system also has self-learning capabilities. By evaluating and analyzing the implementation effects of different response strategies, learning the optimal response strategies in different situations, and using machine learning algorithms, such as decision tree algorithms, new rules are automatically generated and added to the rule base, so that the system can continuously adapt to new traffic conditions and management needs, and improve the scientificity and effectiveness of response strategies.

[0041] Furthermore, the system also includes a user feedback module, which collects user feedback information on the system, including user evaluation of the accuracy of road condition prediction, satisfaction with the optimal route planning, rationality of scenario simulation results, and feedback on the effectiveness of response strategies. Users can input feedback information through mobile applications or web interfaces. After classification and analysis, the feedback information is stored in the system's feedback database. The system optimizes various modules of the system based on user feedback information. If the user feedback shows that the road condition prediction is inaccurate, the data analysis and prediction module will make targeted adjustments to the prediction model and re-evaluate the relevant weights and parameters based on the user's travel time and road section information. If the user is not satisfied with the optimal route planning, the optimal route planning module will adjust the weight coefficient and heuristic function in the path planning algorithm. For feedback on scenario simulation and response strategies, the corresponding modules will also be improved based on user feedback to improve the user experience and performance of the system and achieve continuous optimization and improvement of the system.

[0042] Compared with existing technologies, this intelligent AI highway traffic safety warning system has the following beneficial effects:

[0043] 1. This system significantly improves the early warning capability of highway traffic safety by integrating advanced data prediction and scenario simulation functions. The data acquisition module widely collects historical traffic data, real-time traffic flow data and meteorological conditions data. The data analysis and prediction module uses deep learning algorithms to accurately analyze and predict these multi-source data, and can output the congestion probability and accident risk probability of specific road sections at different time periods in real time. The scenario simulation module constructs a traffic flow simulation model based on the assumptions of the traffic management department, combined with current traffic data and prediction results, and conducts scenario simulation. The simulation results truly reflect the impact of different traffic events on road capacity. This combination of prediction and simulation not only improves the accuracy and timeliness of early warnings, but also provides a scientific decision-making basis for traffic management departments, effectively preventing the occurrence of traffic accidents and ensuring the smooth flow and safety of highways.

[0044] 2. This system uses a distributed database system to store historical traffic data and introduces an encryption algorithm based on blockchain technology to encrypt the data, ensuring the security and integrity of the data. At the same time, the deep neural network model in the data analysis and prediction module introduces an adaptive learning rate adjustment strategy and a simulated annealing algorithm during the training process, which optimizes the weight determination process and improves the prediction performance. In addition, the improved A* algorithm in the optimal route planning module takes into account real-time road conditions and user feedback, dynamically adjusts the path planning, and provides users with safer and more efficient driving routes. These technological innovations not only improve the overall performance of the system, but also enhance the user experience, making this system have higher practical value and promotion significance in actual applications.

[0045] Other advantages, objects and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be learned from the practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0047] Figure 1 This is a process operation diagram of an intelligent AI highway traffic safety warning system. DETAILED DESCRIPTION

[0048] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0049] Example 1

[0050] This example describes a rear-end collision between two vehicles on a straight section of a highway. The driver of a following vehicle was not paying attention and failed to react promptly to the deceleration of the vehicle in front. The accident occurred at 10:00 AM, during peak travel time on a weekday, with heavy traffic. The section already had dense traffic and relatively high speeds. The speed limit on this section was 120 km / h, and there were four lanes in both directions. The road was flat and had good visibility, and traffic was generally smooth before the accident.

[0051] Rapid response: The geomagnetic sensor monitors the passing information of vehicles in real time, including the number of vehicles and passing time, by sensing the geomagnetic changes caused by the passing of vehicles, and transmits these data to the data acquisition server. The radar sensor uses the principle of electromagnetic wave reflection to accurately measure the speed and distance parameters of the vehicle, and also transmits the data to the server through the high-speed communication network. The video surveillance camera captures the real-time picture of the road section, and the image recognition technology extracts the type and location information of the vehicle from it. At the same time, the data acquisition module regularly obtains the historical traffic data of the road section through the data connection established with the data storage system of the traffic management department, covering the traffic volume, speed, and accident situation at different times. The data of meteorological conditions is obtained through the data interface connection with the meteorological department. The meteorological monitoring stations set up along the road section synchronously collect temperature, humidity, air pressure, wind speed, wind direction, precipitation intensity and visibility information. The historical traffic data is stored in a distributed database system, and the data is sharded according to different road sections and time ranges. Each data shard is encoded according to the geographical location and time range of the road section and encrypted using an encryption algorithm based on blockchain technology. In addition, the data acquisition module regularly backs up the data to a remote data storage center, using a combination of incremental backup and full backup, and determines the backup cycle and backup volume according to the frequency and importance of data updates.

[0052] First, the collected multi-source data is preprocessed, the raw data is converted into a format suitable for algorithm input, and the data of different dimensions are normalized so that they are in the same order of magnitude. For missing values in the data, a time series-based interpolation method is used to fill in the missing data points according to the data characteristics of the previous and next moments. The processed data is input into the deep neural network model. The model architecture includes an input layer, multiple hidden layers and an output layer. The hidden layer adopts a hierarchical structure. The number of neurons is dynamically adjusted according to the characteristic dimension of the input data and the key information dimension obtained by historical data analysis. The activation function adopts a custom This function can better handle the complex nonlinear relationships in traffic data. In the weight determination process, the initial weights are first assigned according to the historical influence of different data types, and the historical traffic data are weighted according to the accident rate and congestion frequency of different road sections. Then, the hybrid optimization algorithm of particle swarm optimization and genetic algorithm is used to minimize the prediction error. The weights are updated iteratively for the target. The model is used to analyze the vehicle density, average speed, lane occupancy rate and traffic flow change rate of different road sections. Combined with the congestion threshold in historical data, the fuzzy logic calculation model is used to obtain the congestion probability P of the road section. c =F(ρ,v,l,q,T). Since the accident has occurred, the accident risk probability is directly set to a high value.

[0053] According to the congestion probability and accident risk probability output by the data analysis and prediction module, a graph structure is first constructed based on the road network topology and geographic information. The nodes represent intersections or key locations of road sections, and the edges represent road sections. The weights of the edges are set according to the congestion probability and accident risk probability. The weight calculation formula is W = k1P c +k2P a For main roads and users who prefer safety, k1 is relatively small and k2 is relatively large. Then, the improved A* algorithm is used to search for the optimal path in the graph structure. During the search process, the heuristic function of the node is considered: h(n) = k3d(n,goal) + k4R(n), where d(n,goal) is the distance from the current node to the target node, and R(n) is the comprehensive risk assessment on the path from the current node to the target node. k3 and k4 are determined according to user preferences and road characteristics. The algorithm is used to plan the optimal route for travelers to avoid accident sections and output the route to the user's navigation device or mobile application.

[0054] Receive traffic accident information input from the traffic management department, including accident type, location, start time, duration and severity. Based on this information, combine the current traffic data and the output of the data analysis and prediction module to build a traffic flow simulation model. During the simulation process, due to the occurrence of traffic accidents, the vehicle speed and distance on the corresponding road section are adjusted accordingly, and the traffic conditions under the interaction of vehicle dynamic characteristics, driver behavior and road environment are simulated and calculated, thereby simulating the spread, duration, changes in traffic flow and traffic delays of traffic congestion.

[0055] According to the simulation results of the scenario simulation module, a response strategy is generated through a rule-based expert system. Since the simulation results show traffic congestion, the expert system determines the need to adjust the signal light duration based on the scope and severity of the congestion, extend the green light time at intersections around the accident section, so as to speed up vehicle evacuation, open emergency lanes, guide vehicle diversion, and alleviate congestion in the main lanes. At the same time, according to the location and severity of the accident, the dispatch routes and numbers of rescue vehicles and tow vehicles are determined, and priority is given to ensuring the smooth flow of rescue channels. The generated response strategy is output to the traffic management department and stored in the system's strategy database.

[0056] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent AI highway traffic safety warning system, characterized by: The system includes the following components: data acquisition module, data analysis and prediction module, optimal route planning module and scenario simulation module, and response strategy generation module; The data acquisition module collects multi-source data on the expressway and regularly acquires data by establishing a data connection with the data storage system of the traffic management department. Real-time traffic flow data is collected by various sensors installed on the expressway. The sensors monitor vehicle passing information in real time and transmit the data to the system's data acquisition server via a high-speed communication network. Meteorological condition data is obtained by connecting to a data interface with the meteorological department. At the same time, meteorological monitoring stations are set up along key sections of the expressway to collect temperature, humidity, air pressure, wind speed, wind direction, precipitation intensity and visibility information. The collected meteorological data is stored in a temporally and spatially associated manner with traffic data. The data analysis and prediction module analyzes and predicts the collected multi-source data based on a customized deep learning algorithm, pre-processes the collected data, and converts the original data into a format suitable for algorithm input, including normalizing the data so that data of different dimensions are at the same order of magnitude, and filling in missing values. A time series-based interpolation method is used to fill in missing data points based on the data characteristics of the previous and next moments, and to construct a deep neural network model. The architecture of the model includes an input layer, multiple hidden layers, and an output layer. The hidden layer adopts a hierarchical structure, and the number of neurons in each layer is dynamically adjusted according to the characteristic dimensions of the input data and the key information dimensions obtained from the historical data analysis. The input of the model is pre-processed historical traffic data, real-time traffic flow data, and meteorological condition data. The input data undergoes a series of linear and nonlinear transformations. The specific calculation process is as follows: Among them, x i is the i-th feature of the input data, w i is the weight of the i-th feature, b is the bias term, n is the number of input features, and f is the activation function, using a custom activation function This activation function combines linear and nonlinear parts and can better handle the complex nonlinear relationships in traffic data; Weight w i The determination process first assigns initial weights based on the historical impact of different data types. The historical traffic data are weighted according to the accident rate and congestion frequency of different road sections. Then, a hybrid optimization algorithm based on particle swarm optimization and genetic algorithm is used to optimize the weights. The weights are iteratively updated with the goal of minimizing the prediction error. The formula is as follows: in, is the weight of the i-th feature at the k-th iteration, is the velocity of the i-th feature at the k-th iteration, ω is the inertia weight that is dynamically adjusted according to the data update frequency and the degree of fluctuation of traffic flow, c1 and c2 are learning factors, r1 and r2 are random numbers in the range of [0, 1], is the local optimal weight of the i-th feature, gbest k is the global optimal weight, which is obtained by minimizing the prediction error function OK, y j is the actual observed value, is the model prediction value, m is the number of samples; The final output is the congestion probability P of a specific road section at different time periods c and the accident risk probability P a , where the congestion probability P c The calculation model based on fuzzy logic is derived as follows: P c =F(ρ,v,l,q,T) Where ρ is the vehicle density, v is the average speed, l is the lane occupancy rate, q is the traffic flow change rate, T is the time variable, F is the custom fuzzy logic function, and the accident risk probability P is a Calculated by multi-factor weighted model, the formula is: P a =G(v,d,m,w,c,H) Where v is vehicle speed, d is vehicle distance, m is vehicle type, w is weather conditions, c is road curvature, H is historical accident data, and G is a custom multi-factor weighted function. This function is weighted and summed based on the contribution of each factor to the accident, and is trained and adjusted using historical accident data. The optimal route planning module: uses the path planning algorithm to calculate the congestion probability P output by the data analysis and prediction module. c and the accident risk probability P a To plan the optimal route for travelers, we first construct a graph structure based on the road network topology and geographic information. The nodes represent the intersections or key locations of the road sections, the edges represent the road sections, and the weights of the edges are calculated based on the congestion probability P. c and the accident risk probability P a Set up, the weight calculation formula is W=k1P c +k2P a , where k1 and k2 are weight coefficients determined according to the importance of the road and user preferences. For major roads and users who prioritize safety, k1 is relatively small and k2 is relatively large. Then, an improved A* algorithm is used to search for the optimal path in the graph structure. During the search process, the heuristic function h(n) of the node is considered for path search. The heuristic function combines distance and risk factors. The formula is h(n) = k3d(n,goal) + k4R(n), where d(n,goal) is the distance from the current node n to the target node, R(n) is the comprehensive risk assessment on the path from the current node n to the target node, and k3 and k4 are coefficients determined according to user preferences and road characteristics. The algorithm finds the optimal route from the departure point to the destination and outputs it to the user's navigation device or mobile application; The scenario simulation module receives hypothetical conditions input by the traffic management department, including but not limited to traffic accidents, road construction, severe weather, the type, location, start time, duration and severity of emergencies. Based on the input hypothetical conditions, combined with the current traffic data and the output of the data analysis and prediction module, the scenario simulation is performed by constructing a traffic flow simulation model. The simulation model considers the interaction between the vehicle's dynamic characteristics, driver behavior and road environment. The vehicle dynamic characteristics are modeled based on the vehicle's physical parameters and traffic regulations. The driver's behavior is simulated based on the driving habits in the force history data and the current road conditions. The road environment takes into account the road geometry, number of lanes, speed limit and road surface conditions. The specific simulation calculation process is as follows: Where v is the vehicle speed, t is time, a(t) is the vehicle acceleration, d is the distance between vehicles, w is the vehicle weight, s is the road slope, e is the road surface friction coefficient, and f is a function that comprehensively considers various factors. During the simulation process, the corresponding parameters are updated according to the assumptions. When a traffic accident occurs, the vehicle speed and distance between vehicles on the corresponding road section are adjusted accordingly. During road construction, the number of lanes and road surface conditions change, thereby simulating possible traffic impacts. The response strategy generation module generates corresponding response strategies based on the simulation results of the scenario simulation module. For different simulated traffic impact situations, response strategies are generated through a rule-based expert system. The rules are formulated based on traffic engineering theory and historical experience. When traffic congestion occurs, the scope and severity of the congestion determine whether it is necessary to adjust the signal light duration, open emergency lanes, and guide vehicle diversion strategies. When a traffic accident occurs, the dispatch routes and numbers of rescue vehicles and tow vehicles are determined based on the location and severity of the accident. When road construction occurs, the location and time of the construction determine whether temporary traffic signs, detour routes, and traffic control measures in the construction area need to be set up. The generated response strategies are output to the traffic management department and stored in the system's strategy database.

2. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The storage method of the historical traffic data in the data acquisition module adopts a distributed database system. The distributed database system performs data shard storage according to different road sections and different time ranges, and stores the data on multiple nodes. Each data shard is encoded according to the geographical location and time range of the road section. At the same time, during the data storage process, the data is encrypted and an encryption algorithm based on blockchain technology is used to divide the data into multiple data blocks. Each data block contains the hash value and encryption information of the data. The integrity and security of the data are guaranteed by the chain structure of the blockchain to prevent the data from being tampered with. When querying data, the system automatically locates the corresponding data shard according to the road section and time range of the query request, and reads data from multiple nodes in parallel. In addition, the data acquisition module also has a data backup function, which regularly backs up data to a remote data storage center. The backup data adopts a combination of incremental backup and full backup, and the backup period and backup volume are determined according to the frequency and importance of data updates.

3. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: During the training process of the deep neural network model in the data analysis and prediction module, an adaptive learning rate adjustment strategy is introduced. The initial learning rate is set according to the scale and complexity of the data. When the training error of the model no longer decreases or decreases slowly, the learning rate is automatically adjusted. The specific adjustment formula is: Among them, α k is the learning rate at the kth iteration, α k+1 is the learning rate at the k+1th iteration, E k and E k+1 are the training errors at the kth and k+1th iterations, ∈ is the set error threshold, γ is the decrease factor, which is used when the error decreases less than the threshold, and β is the growth factor, which is used when the error increases or does not decrease. At the same time, in order to avoid overfitting, a new regularization method is introduced in the hidden layer, and a regularization term based on information entropy is added to the loss function. The formula is: L=E+λH Among them, L is the total loss function, E is the original prediction error loss function, λ is the regularization parameter, which is adjusted according to the noise level of the data and the complexity of the model, and H is the information entropy term, which is calculated as H = -∑ i p i log(p i ), p i is the probability distribution of the model's predicted output.

4. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The weight optimization algorithm in the data analysis and prediction module is also combined with a simulated annealing algorithm. During the optimization process, when the particle swarm optimization and genetic algorithm fall into a local optimum, the simulated annealing algorithm is introduced to search. The acceptance criteria of the simulated annealing algorithm are: Where ΔE is the energy difference, that is, the difference between the error of the new solution and the error of the current solution, T is the simulated annealing temperature, the initial temperature is determined according to the complexity of the data and the size of the search space, and as the number of iterations increases, the temperature decreases according to the cooling function T k+1 =αT k To reduce, α is the cooling coefficient.

5. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The improved A* algorithm in the optimal route planning module introduces a dynamic weight adjustment mechanism when searching for the optimal path. According to real-time traffic information and user feedback, the weight coefficients k3 and k4 in the heuristic function of the node are dynamically adjusted. When the user finds new congestion or risk on the recommended route during driving, the system receives user feedback and updates the risk assessment of the node based on the current traffic information. At the same time, during the search process, for different road section types, the algorithm also takes into account the tidal phenomenon of traffic flow and adjusts the weight of the edge according to the traffic flow direction at different time periods. During peak traffic hours, the weight of the road section in the main traffic flow direction is reduced to guide users to avoid congested sections.

6. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The traffic flow simulation model in the scenario simulation module also takes into account the characteristics of vehicle platooning. For platoons or large transport vehicle platoons, platoon dynamics are introduced into the model based on the communication and collaboration relationships between vehicles. The model considers the safe spacing, speed synchronization, and collaborative driving behavior between vehicles. When simulating platoon driving, the spacing and speed relationship between vehicles are adjusted according to the length, number, and type of the platoon. The formula is: d ij =f(n,m,v) Among them, d ij is the distance between the i-th and j-th vehicles, n is the number of vehicles in the platoon, m is the vehicle type, v is the vehicle speed, and f is the spacing function determined according to vehicle performance and traffic regulations. At the same time, the impact of information sharing in the platoon on traffic flow is considered. When the lead vehicle in the platoon receives road condition information, it transmits the information to subsequent vehicles through vehicle-to-vehicle communication, simulating the propagation speed and impact range of information within the platoon.

7. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: During the simulation process, the scenario simulation module models the traffic flow noise. Traffic flow noise takes into account the random behavior of drivers, fluctuations in vehicle performance, and environmental uncertainty. It is simulated by adding random terms to the vehicle dynamics equation. Specifically, Here, ξ(t) is a random noise term whose distribution is determined by statistical analysis of vehicle speed fluctuations and environmental changes in historical data. The distribution parameters of ξ(t) are different for different types of vehicles and different environmental conditions. In severe weather conditions, the variance of ξ(t) is larger to reflect the uncertainty of driver operation and the instability of vehicle performance.

8. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The rule-based expert system in the response strategy generation module has a dynamic update function in its rule base. In addition to the initial rules formulated based on traffic engineering theory and historical experience, the system automatically updates and optimizes the rules based on actual traffic management decision-making results and scenario simulation results. When the traffic management department implements a certain response strategy and achieves certain results, the system adjusts the corresponding rules based on the implementation effect and traffic conditions feedback. At the same time, the expert system also has self-learning capabilities. By evaluating and analyzing the implementation effects of different response strategies, learning the optimal response strategies in different situations, and using machine learning algorithms, the system can continuously adapt to new traffic conditions and management needs.

9. The intelligent AI highway traffic safety warning system according to claim 1 is characterized in that: The system also includes a user feedback module that collects user feedback on the system, including user evaluation of the accuracy of road condition predictions, satisfaction with optimal route planning, rationality of scenario simulation results, and effectiveness of response strategies. Users can input feedback information through a mobile application or web interface. After classification and analysis, the feedback information is stored in the system's feedback database. The system optimizes various modules of the system based on user feedback information. For feedback on scenario simulation and response strategies, the corresponding modules will also be improved based on user feedback.

Images