Personalized self-driving travel route planning method and system

Abstract

The invention discloses a personalized self-driving travel route planning method and system, and relates to the technical field of intelligent travel, and the method comprises the steps: obtaining the historical travel track data of a user, and collecting the real-time road condition data and weather information around a destination; extracting a user preference feature matrix from the historical travel trajectory data of the user by using clustering analysis, constructing a geographic position-point-of-interest association network, and dividing the geographic position-point-of-interest association network by using a community discovery algorithm to obtain destination groups conforming to user preferences; constructing a road network cost matrix, performing path search by using an improved ant colony algorithm, and generating candidate routes in combination with real-time road conditions; and mapping the candidate route to a rasterized map, carrying out comprehensive evaluation on the candidate route, and selecting a Pareto optimal solution as a recommended route. Practicality and reliability of route planning can be improved, a reliable travel scheme is provided for a user, and travel experience and satisfaction of the user are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent travel technology, and in particular to a personalized self-driving travel route planning method and system. Background Art

[0002] With the advancement of science and technology and the in-depth development of informatization, smart travel and personalized tourism have gradually become important needs of modern society. In the field of self-driving travel, the emergence of intelligent route planning systems enables travelers to arrange their itineraries more efficiently. Traditional self-driving travel route planning methods are mainly based on static geographic information system (GIS) data, such as traffic networks, road conditions, etc., combined with simple path optimization algorithms for route recommendation. However, these traditional methods often ignore the changes in user personalized needs and real-time factors, resulting in the lack of targetedness of recommended routes, which are difficult to adapt to the travel preferences of different users and real-time changes in road conditions.

[0003] In recent years, with the rapid development of big data, artificial intelligence (AI) and machine learning technologies, intelligent recommendation systems based on user historical data and real-time dynamic information have gradually entered the field of self-driving travel route planning. Many path planning methods in the existing technology still rely on simple rules or static data, and fail to fully mine the user's historical behavior data to establish a more accurate personalized preference model. For example, many existing systems can only recommend routes based on a single user input (such as a destination or departure point), ignoring the user's diverse interests and travel habits. In addition, although community discovery algorithms and cluster analysis have been applied in some recommendation systems, how to effectively combine these algorithms with real-time road conditions, weather and other factors is still a challenge. Therefore, how to provide a reasonable and efficient self-driving travel route planning method under the premise of comprehensively considering historical data, real-time data and personalized preferences is still a problem that needs to be solved in current technology. Summary of the invention

[0004] In view of the problems existing in the existing personalized self-driving travel route planning method, the present invention is proposed. Therefore, the problem to be solved by the present invention is how to provide a personalized self-driving travel route planning method and system.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a personalized self-driving travel route planning method, which includes obtaining historical travel trajectory data of a user, collecting real-time traffic data and weather information around the destination; Using cluster analysis to extract user preference feature matrix from user historical travel trajectory data, constructing a geographic location-point of interest association network, and using community discovery algorithm to divide the geographic location-point of interest association network to obtain destination groups that meet user preferences; Construct a road network cost matrix, use an improved ant colony algorithm to search for paths, and generate candidate routes based on real-time traffic conditions; The candidate routes are mapped to a rasterized map, the candidate routes are comprehensively evaluated, and the Pareto optimal solution is selected as the recommended route.

[0006] As a preferred solution of the personalized self-driving travel route planning method of the present invention, wherein: the acquisition of the user's historical travel trajectory data includes acquiring the stored travel records by connecting with the user's smart device, and acquiring the user's travel data from the platform's database for users who use a specific travel service application; The collecting of real-time traffic data and weather information around the destination includes integrating real-time traffic data from multiple map service providers and obtaining real-time weather data around the destination from a meteorological data service provider.

[0007] As a preferred solution of the personalized self-driving travel route planning method of the present invention, wherein: the extraction of the user preference feature matrix includes: Calculate the straight-line distance between the starting point and the end point of each user's trip to obtain travel distance data, statistically analyze the travel distance data, and calculate the average travel distance to form the user's travel distance characteristics; Analyze the location information of the stopover points in the user's historical trajectory, combine the geographic information system data and the data of the online travel platform, determine the type of scenic spot to which the stopover point belongs, count the number of times and duration of users' stops at different types of scenic spots, divide the user's preference for different types of scenic spots, and form the characteristics of scenic spot types; Analyze the driving speed, acceleration and deceleration pattern data in the user's historical travel trajectory, extract the user's driving habit characteristics, and form a driving habit feature; Integrate the extracted user travel distance features, scenic spot type features, and driving habit features into a user preference feature matrix; The K-means clustering algorithm is used to perform cluster analysis on the user preference feature matrix. According to the data in the user preference feature matrix, the users are divided into different clusters, and each cluster represents a user group with similar preference characteristics.

[0008] As a preferred solution of the personalized self-driving travel route planning method of the present invention, wherein: the obtaining of the destination group that meets the user's preference includes the following steps: Integrate the tourist attraction database to obtain the location information and attraction types of the scenic area, filter out the points of interest in the area from the geographic information system data and the data of the online travel platform, and determine the initial importance weight of the points of interest based on the type, popularity and user evaluation of the points of interest; For every two points of interest, the geographical distance between the points of interest is calculated, and the geographical location association strength between the points of interest is determined according to the size of the geographical distance, and a geographical location association matrix is ​​constructed; Based on the user preference feature matrix and the attribute information of the points of interest, the similarity between every two points of interest is calculated using the following formula: ; In the formula, is the feature vector of interest point A, is the feature vector of interest point B, S is the similarity between interest point A and interest point B; Construct an association network, take points of interest as nodes, use the geographic location association strength and the similarity of points of interest as the weight of the edge, and construct a geographic location-point of interest association network; The improved community discovery algorithm is used to divide the geographic location-point of interest association network into communities, and the preference characteristics of the target user group are matched with the community division results to determine the destination group that meets the preferences of the target user group. The community division is optimized by maximizing the modularity. The formula is: ; Among them, Q is the community modularity, is the adjacency matrix of the network, which represents the edge weight between node i and node j. and is the degree of node i and node j, m is the sum of the weights of all edges in the network, γ is the resolution parameter, is the Kronecker function, which is 1 when nodes i and j belong to the same community, otherwise it is 0; The edge weight is dynamically adjusted according to the real-time traffic conditions. The adjustment formula is: ; in, is the adjusted edge weight, is the initial static edge weight, w(t) is the real-time data, is the weight coefficient; Calculate the matching probability between the target user group and each community. The formula is: ; in, is the matching probability between the target user group and community c, is the preference feature vector of the target user group, is the preference feature vector of interest point z in the community.

[0009] As a preferred solution of the personalized self-driving travel route planning method of the present invention, wherein: the generation of candidate routes includes the following steps: Collect road network data in the target area, analyze road accident records in the area, and calculate the accident rate of each road section; Construct the basic road network cost matrix according to road grade, traffic capacity and accident rate parameters; Use the traffic sensor network and the real-time traffic API of the map service provider to obtain the real-time traffic data of the current road and calculate the real-time travel time and congestion level of each road section; Combined with the basic road network cost matrix and real-time traffic data, a dynamic adjustment strategy for the road network cost is formulated. The formula is: ; in, is the road network cost at time t after dynamic adjustment, is the basic road network cost, It is the extra cost brought by real-time traffic conditions. is the weight parameter; The basic parameters of the ant colony algorithm are set, and the real-time traffic conditions and historical path experience are comprehensively considered. The pheromone update factor is introduced. The formula is: ; Where F is the pheromone renewal factor, is the real-time transit time. is the historical average travel time, It is the price of historical experience. is the weight parameter; Each ant starts from the starting point and selects the next node according to the probability rule based on the basic road network cost matrix and pheromone concentration. The formula is: ; in, is the path selection probability of selecting the next node j, is the set of next nodes that ant k can choose when it is at node i, l is the set of next nodes Any node in is the pheromone concentration on edge (i, j) at time t, is the pheromone weight, is the heuristic information of edge (i, j), β is the heuristic information weight, is the pheromone concentration on the edge (i, l) at time t, is the heuristic information of edge (i,l); After the ant completes a path exploration, the pheromone on the path is updated according to the pheromone update factor. After multiple iterations, multiple candidate routes are generated. After the path search is completed, the pheromone on the optimal path is globally enhanced. The formula is: ; in, is the enhanced pheromone concentration on the optimal path, is the amount of pheromone added to the edge (i, j) on the optimal path; is the pheromone volatility coefficient; During the movement of ants, the pheromone on the current edge is locally updated, and the formula is: ; in, is the pheromone concentration after the ants move. is the amount of pheromone added each time an ant passes through edge (i, j).

[0010] As a preferred solution of the personalized self-driving travel route planning method of the present invention, wherein: mapping the candidate routes to the rasterized map includes the following steps: Collect geographic information data of the target geographic area and determine the appropriate grid size based on the size and resolution requirements of the geographic area; Divide the geographical area into multiple grids to form a grid map, and obtain multiple candidate routes from the path planning algorithm; The geographic coordinates of each road node in the candidate route are converted into corresponding grid coordinates, and the route is represented as a connected grid according to the grid coordinates of the node to generate a grid matrix.

[0011] As a preferred solution of the personalized self-driving travel route planning method of the present invention, the comprehensive evaluation of the candidate routes includes the following steps: Collect information on scenic spots in the target geographic area, convert the latitude and longitude coordinates of the scenic spots into corresponding grid coordinates, and determine the grid where the scenic spots are located; For each candidate route, the coverage rate of scenic spots, average road condition score, and average safety factor contained in the grids passed by the route are calculated based on the grid matrix and the associated information of scenic spots grids. Taking scenic spot coverage, road condition score and safety factor as the objectives of multi-objective optimization problem, a mathematical model is constructed; A multi-objective optimization algorithm is used to comprehensively evaluate candidate routes and find the Pareto optimal solution set. The final recommended route is selected based on user preferences or specific needs, and the route recommendation results are dynamically updated based on user feedback and real-time traffic changes.

[0012] In a second aspect, the present invention provides a personalized self-driving travel route planning system, which includes: The acquisition module is used to obtain the user's historical travel trajectory data and collect real-time traffic data and weather information around the destination; An analysis module is used to extract a user preference feature matrix from the user's historical travel trajectory data using cluster analysis, construct a geographic location-point of interest association network, and divide the geographic location-point of interest association network using a community discovery algorithm to obtain a destination group that meets the user's preferences; The construction module is used to construct the road network cost matrix, use the improved ant colony algorithm to search for paths, and generate candidate routes based on real-time road conditions; The recommendation module is used to map candidate routes to a rasterized map, conduct a comprehensive evaluation of the candidate routes, and select the Pareto optimal solution as the recommended route.

[0013] The beneficial effects of the present invention are that it can deeply explore users' travel habits and preferences, ensure the smoothness and safety of travel, improve the practicality and reliability of route planning, generate multiple candidate routes that meet user needs and real-time road conditions, provide users with reliable travel plans, and improve users' travel experience and satisfaction. DETAILED DESCRIPTION

[0014] In order to make the above-mentioned purposes, features and advantages of the present invention more understandable, the specific implementation methods of the present invention are described in detail. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0015] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0016] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0017] This is the first embodiment of the present invention, which provides a personalized self-driving travel route planning method, including: S1: Obtain the user's historical travel trajectory data and collect real-time traffic data and weather information around the destination; Specifically, by connecting to the user's smart device (such as a mobile phone, in-car navigation system, etc.), the stored travel records can be obtained. These records are usually recorded by the device's positioning system (such as GPS) at a certain time interval (for example, every 10 seconds), including longitude, latitude and altitude. For some users who use specific travel service applications (such as online car-hailing platforms), the user's travel data can also be obtained from the platform's database. This data may contain more detailed travel information, such as the specific addresses of the departure and destination.

[0018] Extract the start time of each trip from the record, accurate to the minute or even the second. This helps analyze the user's travel time pattern, such as whether it is a weekday rush hour trip, or a weekend leisure trip. By analyzing the changes in location information, when the device stays at a certain location for a period of time (which can be judged based on speed and location changes, for example, the speed remains at 0 and the location remains basically unchanged for more than 5 minutes), the location is recorded as a stop point. At the same time, record the arrival and departure time of the stop point, as well as the geographic coordinates of the stop point. Calculate based on consecutive location record points. The distance can be obtained by calculating the geographic distance between two consecutive location points (using the latitude and longitude coordinate calculation formula, such as the Haversine formula).

[0019] Real-time traffic data collection, integrating real-time traffic data from multiple map service providers (such as Amap, Baidu Map, etc.). These providers analyze traffic conditions through users' map usage data (such as real-time location and speed information when users turn on navigation) and traffic camera data, and provide them to other applications in the form of APIs (application programming interfaces).

[0020] Weather information collection: obtain real-time weather data around the destination from professional meteorological data service providers (such as the National Meteorological Center, commercial meteorological companies, etc.), including temperature, humidity, precipitation probability, wind speed and direction, etc. Meteorological data is usually collected through a variety of means such as meteorological satellites and ground meteorological observation stations, and will be updated according to real-time meteorological changes.

[0021] The collected historical travel trajectory data of users is cleaned to remove abnormal data points (such as points that obviously deviate from the normal path due to positioning errors) and duplicate data points. By setting reasonable thresholds (such as speed thresholds, distance thresholds, etc.), data points that do not conform to normal travel patterns are identified and eliminated to ensure the accuracy and reliability of the data.

[0022] It can deeply explore users' travel habits and preferences, provide a basis for building a user preference feature matrix, make the subsequently planned routes more in line with users' actual needs, and provide real-time environmental factors for route planning. For example, if the traffic around the destination is congested or the weather is bad, the route can be adjusted in advance to avoid unnecessary troubles for users during travel, ensure the smoothness and safety of travel, provide important real-time constraints for path search, and improve the practicality and reliability of route planning.

[0023] S2: Use cluster analysis to extract the user preference feature matrix from the user's historical travel trajectory data, build a geographic location-point of interest association network, and use a community discovery algorithm to divide the geographic location-point of interest association network to obtain destination groups that meet user preferences; Specifically, the user travel distance feature is extracted, and the straight-line distance between the starting point and the end point of each user's trip is calculated to obtain the travel distance data. These travel distance data are statistically analyzed and the average travel distance is calculated to reflect the user's preference characteristics in terms of travel distance.

[0024] Scenic spot type feature extraction, combined with geographic information system (GIS) data and online travel platform data, classifies the stopover points in the user's historical travel trajectory and determines the type of scenic spot to which the stopover point belongs (such as natural scenery, historical and cultural sites, theme parks, etc.). The number of times and duration of users' stops at different types of scenic spots are counted as an indicator of the user's preference for different types of scenic spots. For example, if the user stops at natural scenery attractions more times and stays for a longer time, it means that the user has a higher preference for natural scenery attractions.

[0025] Driving habit feature extraction: Analyze the driving speed, acceleration and deceleration patterns in the user's historical travel trajectory to extract the user's driving habit features. For example, calculate the user's average driving speed under different road conditions, the frequency of rapid acceleration and deceleration, and other indicators. Through these indicators, the user's driving habits can be divided into different types, such as stable type, aggressive type, etc., to provide a basis for subsequent analysis.

[0026] The extracted user travel distance features, scenic spot type features, and driving habit features are integrated into a user preference feature matrix. The rows of the matrix represent different feature dimensions (travel distance, scenic spot type, driving habits), and the columns represent different users. The value of each element represents the user's preference for the corresponding feature dimension. The user's preference features are quantified in a numerical way to provide a data basis for cluster analysis.

[0027] The K-means clustering algorithm is used to perform cluster analysis on the user preference feature matrix. According to the data in the user preference feature matrix, users are divided into different clusters, and each cluster represents a user group with similar preference characteristics. Through cluster analysis, the commonalities and differences in users' travel preferences can be found, providing a basis for the subsequent construction of the geographic location-point of interest association network.

[0028] Taking scenic spots as nodes of the network, the location information (latitude and longitude coordinates) and attributes such as the type of scenic spots are obtained by integrating tourist attraction databases (such as scenic spot lists, scenic spot information on online travel platforms, etc.). Points of interest within a certain area are screened out from the data of the geographic information system (GIS) and the data of the online travel platform, and the initial importance weight of the points of interest is determined based on factors such as the type, popularity, and user evaluation of the points of interest.

[0029] For every two points of interest, the geographic distance (such as Euclidean distance or actual road distance) between the points of interest is calculated. According to the size of the geographic distance, the geographic location association strength between the points of interest is determined. The closer the geographic distance, the higher the association strength. By calculating the geographic location association strength between all points of interest, a geographic location association matrix is ​​constructed.

[0030] Based on the user preference feature matrix and the attribute information of the points of interest (such as attraction type, special items, etc.), the similarity between every two points of interest is calculated. The similarity can be calculated using methods such as cosine similarity and Jaccard similarity coefficient. The similarity value is determined based on the similarity of the points of interest in terms of user preference features and attribute information. The higher the similarity value, the more similar the two points of interest are in terms of user preferences and attributes, which provides a basis for the subsequent construction of the association network; the calculation formula is: ; In the formula, is the feature vector of interest point A, is the feature vector of interest point B, is the similarity between interest point A and interest point B; Construct an association network, take points of interest as nodes, and use the geographic location association strength and the similarity of points of interest as the weight of the edge to construct a geographic location-point of interest association network. In the network, nodes represent points of interest, and edges represent the association between points of interest. The larger the weight of the edge, the closer the association between the two points of interest.

[0031] The improved community discovery algorithm is used to divide the geographic location-point of interest association network into communities. Based on the traditional community discovery algorithm, the improved community discovery algorithm introduces a dynamic adjustment mechanism for edge weights. According to the changes in the association strength and similarity between points of interest, the edge weights are dynamically adjusted, so that the community division results can more accurately reflect the user preference characteristics.

[0032] The improved community discovery algorithm is used to divide the association network into multiple communities, each of which represents a group of points of interest with similar user preference characteristics and geographical location associations. Through community division, a large number of points of interest are classified to provide users with destination group recommendations that meet their preferences.

[0033] The target user group is determined based on the user's historical travel trajectory data and preference feature matrix. The target user group can be users with similar travel preference characteristics or specific types of users (such as users who like natural scenery, users who like historical and cultural sites, etc.).

[0034] The preference characteristics of the target user group are matched with the community division results to determine the destination group that meets the preferences of the target user group. According to the user preference feature matrix and the characteristic information of the community, the matching degree between the target user group and each community is calculated. The higher the matching degree, the more the points of interest in the community meet the preferences of the target user group. Through the matching process, users can be provided with destination group recommendations that meet their preferences, meet their personalized needs, and optimize community division by maximizing modularity.

[0035] ; Among them, Q is the community modularity, is the adjacency matrix of the network, which represents the edge weight between node i and node j. and is the degree of node i and node j (the sum of the weights of the edges connected to them), m is the sum of the weights of all edges in the network, and γ is a resolution parameter used to control the size of the community. is the Kronecker function, which is 1 when nodes i and j belong to the same community, otherwise it is 0; The edge weight is dynamically adjusted according to the real-time traffic conditions. The adjustment formula is: ; in, is the adjusted edge weight, is the initial static edge weight, w(t) is the real-time data, is the weight coefficient, which controls the degree of fusion of static and dynamic data; Calculate the matching probability between the target user group and each community. ; in, is the matching probability between the target user group and community c, is the preference feature vector of the target user group, is the preference feature vector of interest point z in the community.

[0036] Accurately quantify user travel preferences and provide users with more personalized recommendations. For example, if a user likes natural scenery, geographical locations and points of interest containing natural scenery can be grouped into the same community, thereby recommending destination groups that meet their preferences to users, improving the accuracy and satisfaction of recommendations.

[0037] S3: Construct a road network cost matrix, use the improved ant colony algorithm to search for paths, and generate candidate routes based on real-time traffic conditions; Specifically, collect and organize road network data in the target area, including road length, grade (such as expressway, first-class highway, second-class highway, etc.), traffic capacity (such as the number of vehicles passing per hour), etc. These data can be obtained from public databases of transportation departments, map service providers and other channels.

[0038] Analyze the road accident records in the area and count the number of accidents on each road section in the past period of time. Calculate the accident rate of each road section based on the number of accidents and road traffic flow. Roads with high accident rates need to be treated with caution in subsequent route planning.

[0039] The basic road network cost matrix is ​​constructed based on parameters such as road grade, traffic capacity and accident rate.

[0040] The road grade weight parameter is used to reflect the priority of roads of different grades in path planning; the capacity cost parameter represents the increased cost of roads due to capacity restrictions; and the accident rate cost parameter is used to quantify the increased cost of roads due to accident risks. Taking these three parameters into consideration, the basic road network cost matrix is ​​obtained, in which the cost value of each road segment is the weighted sum of the road grade weight, capacity cost, and accident rate cost.

[0041] Use the traffic sensor network and the map service provider's real-time traffic API to obtain data on current road traffic flow, speed, congestion, etc. The traffic sensor network includes traffic flow monitors, speedometers and other equipment distributed along the road. The real-time traffic API can provide traffic analysis results based on user feedback and real-time traffic data.

[0042] The collected real-time traffic data is processed and analyzed to calculate the real-time travel time, congestion level and other indicators of each road section. For example, the saturation of the road can be calculated based on the traffic volume and road capacity; the speed ratio and other indicators can be calculated based on the vehicle speed and road speed limit.

[0043] Combine the basic road network cost matrix and real-time traffic data to formulate a dynamic adjustment strategy for the road network cost. Data such as traffic volume and speed in real-time traffic conditions will affect the travel time of the road section, and accident rate data will affect the safety of the road section. According to real-time traffic data, adjust the cost of each road section in real time to make it closer to the actual traffic conditions. For example, when a certain section of road is congested, its travel time increases and the cost of the road section increases accordingly; when the accident rate of a certain section of road increases, its safety risk increases, and the cost of the road section also increases accordingly. Combine the basic road network cost matrix and real-time traffic data to dynamically adjust the section cost. The formula is: ; in, is the road network cost at time t after dynamic adjustment, is the basic road network cost, It is the extra cost brought by real-time traffic conditions. is the weight parameter; Set the basic parameters of the ant colony algorithm, including the number of ants, initial pheromone concentration, pheromone volatility coefficient, etc. The number of ants is determined according to the scale and complexity of the road network. The initial pheromone concentration and volatility coefficient affect the convergence speed and global search ability of the algorithm. Determine the appropriate initial parameter values ​​based on empirical formulas and experimental tests. The number of ants (N): determines the parallelism and search ability of the algorithm. The initial pheromone concentration (τ): affects the degree of ant dependence on historical information when choosing a path. The pheromone volatility coefficient ( ): Control the dissipation rate of pheromone over time to avoid premature convergence of the algorithm. Heuristic information weight (β): Adjust the sensitivity of path selection to distance or travel cost. Pheromone weight ( ): Balancing the role of pheromone and heuristic information in path selection.

[0044] The pheromone update factor is introduced, which comprehensively considers the real-time traffic conditions and historical path experience. The pheromone update factor is dynamically calculated based on the real-time travel time, accident rate and other data of the current section. For example, when the real-time travel time of a section is short and the accident rate is low, the pheromone update factor is large, indicating that the section should be given more attention in path planning; otherwise, the pheromone update factor is small.

[0045] Taking into account the real-time traffic conditions and historical path experience, the pheromone update factor (F) is introduced, and its calculation formula is: ; Where F is the pheromone renewal factor, is the real-time transit time. is the historical average travel time, It is the price of historical experience. is a weight parameter that controls the importance of real-time traffic conditions and historical experience.

[0046] Each ant starts from the starting point and selects the next node according to the probability rule based on the basic road network cost matrix and pheromone concentration. The selection probability is proportional to the pheromone concentration of the road section and inversely proportional to the cost of the road section. Specifically, the probability of an ant choosing a road section is determined by the pheromone concentration and the road section cost. This makes ants more inclined to choose roads with high pheromone concentration and low cost.

[0047] The probability of an ant choosing the next node is determined by the pheromone concentration and heuristic information, and the formula is: ; in, is the path selection probability of selecting the next node j, is the set of next nodes that ant k can choose when it is at node i, l is the set of next nodes Any node in is the pheromone concentration on edge (i, j) at time t, is the pheromone weight, is the heuristic information of edge (i, j), β is the heuristic information weight, is the pheromone concentration on the edge (i, l) at time t, is the heuristic information of edge (i,l); After the ant completes a path exploration, the pheromone on the path is updated according to the pheromone update factor. The pheromone volatilization amount is determined by the volatilization coefficient, and the pheromone increase is determined by the path cost and the pheromone update factor. For paths with lower costs, the pheromone increase is larger; while for paths with poor real-time road conditions, the pheromone increase is smaller or may even decrease. It can quickly adapt to changes in road conditions, adjust the pheromone distribution, and guide subsequent ants to choose a better path.

[0048] Global update: After the path search is completed, the pheromone on the optimal path is strengthened. The formula is: ; in, is the enhanced pheromone concentration on the optimal path, is the amount of pheromone added to the edge (i, j) on the optimal path; is the pheromone volatility coefficient; Local update: During the movement of the ant, the pheromone on the current edge is fine-tuned. The formula is:

[0049] in, is the pheromone concentration after the ants move. is the amount of pheromone added each time an ant passes through edge (i, j); After multiple iterations, the ant colony algorithm will generate multiple candidate routes. These paths are discovered by ants during the search process and have different costs and pheromone concentrations.

[0050] In the process of path search, the pros and cons of each road can be evaluated more accurately, providing an important reference for path selection. Combining the global search capability and local search capability of the ant colony algorithm, it can quickly find the optimal path from the starting point to the end point. The improved ant colony algorithm improves the convergence speed and search accuracy of the algorithm by introducing pheromone update mechanism and heuristic information, making the searched path more in line with actual needs. If a road is congested or an accident occurs, the path can be adjusted in time to avoid unnecessary delays for users during travel, and multiple candidate routes that meet user needs and real-time road conditions are generated, providing rich choices for comprehensive evaluation.

[0051] S4: Map the candidate routes to the rasterized map, conduct a comprehensive evaluation on the candidate routes, and select the Pareto optimal solution as the recommended route.

[0052] Specifically, geographic information data of the target geographical area is collected, including regional boundaries, topography, town distribution, etc., to provide basic reference data for the construction of raster maps.

[0053] Determine the appropriate grid size based on the size of the geographic area and the resolution requirements. For example, a smaller grid (such as 100 meters by 100 meters) can be selected in urban areas, and a larger grid (such as 500 meters by 500 meters) can be selected in suburban or rural areas. Give each grid a unique number to facilitate subsequent data processing and analysis.

[0054] The geographical area is divided into multiple grids to form a rasterized map. Each grid represents a certain range of geographical space, and its position is determined by latitude and longitude coordinates or plane rectangular coordinates. Rasterized maps can be stored as digital map files for easy computer processing and analysis.

[0055] Get multiple candidate routes from the path planning algorithm. Each candidate route consists of a series of road nodes and roads connecting them, representing a possible path from the starting point to the end point.

[0056] The geographic coordinates (such as longitude and latitude) of each road node in the candidate route are converted into corresponding grid coordinates. According to the numbering rules of the rasterized map, the grid to which each node belongs is determined.

[0057] According to the grid coordinates of the nodes, the entire route is represented as a series of connected grids. If two adjacent nodes are located in different grids, it is necessary to determine how the route is connected between the two grids, such as using an interpolation method or a geometric algorithm to generate a path between the grids.

[0058] A grid matrix is ​​generated for each candidate route, where the rows and columns of the matrix correspond to the grids in the rasterized map. The elements in the matrix indicate whether the grid is passed by the route, or record the specific situation of the route in the grid (such as the number of times it passes, the stay time, etc.).

[0059] Collect information about attractions in the target geographic area, including the name, location (latitude and longitude), type, etc. The location information of the attraction is used to determine its location in the rasterized map.

[0060] The latitude and longitude coordinates of each scenic spot are converted into corresponding grid coordinates to determine the grid where the scenic spot is located. In this way, each scenic spot is associated with one or more grids in the rasterized map.

[0061] For each candidate route, the number of attractions contained in the grids that the route passes through is calculated based on its grid matrix and the associated information of the attraction grids.

[0062] Attraction coverage rate = (the number of attractions included in the grids passed by the route) / (the total number of attractions in the geographical area) × 100% Obtain real-time traffic data in the target geographic area, including road traffic volume, speed, congestion level, etc. This data is obtained from channels such as the real-time traffic API of the map service provider.

[0063] Based on the road condition data, a road condition score is calculated for each section of road. For example, the road condition can be divided into three levels: smooth, slow, and congested, based on the traffic volume and speed, and the corresponding scores are assigned (e.g., smooth is 100 points, slow is 70 points, and congested is 40 points).

[0064] For each candidate route, the average road condition score of the entire route is calculated based on the road segments it passes through and the road condition scores of the segments.

[0065] Average road condition score = (sum of road condition scores of road sections) / (number of road sections passed by the route) Collect road safety data within the target geographic area, including road accident rates, road grades, and the degree of completeness of traffic facilities.

[0066] Based on safety data such as road grade and accident rate, the safety factor is calculated for each section of road.

[0067] Safety factor = road grade weight × road grade + accident rate weight × (1-accident rate) / maximum accident rate.

[0068] Among them, the road grade weight and accident rate weight are determined according to the actual situation, for example, they are set to 0.6 and 0.4 respectively.

[0069] For each candidate route, the average safety factor of the entire route is calculated based on the road segments it passes through and the safety factors of the segments.

[0070] Average safety factor = (sum of road section safety factors) / (number of road sections the route passes through) The scenic spot coverage, road condition score, and safety factor are taken as the three objectives of the multi-objective optimization problem. A mathematical model is constructed, and the objective function is to maximize the scenic spot coverage, maximize the road condition score, and maximize the safety factor. At the same time, constraints such as route length are considered to ensure the feasibility of the route.

[0071] A multi-objective optimization algorithm (such as NSGA-II algorithm, MOPSO algorithm, etc.) is used to comprehensively evaluate the candidate routes and find the Pareto optimal solution set.

[0072] The final recommended route is selected from the Pareto optimal solution set based on the user's preferences or specific needs. For example, if the user pays more attention to the coverage of scenic spots, the Pareto optimal solution with a higher coverage of scenic spots can be selected; if the user pays more attention to road conditions and safety, the Pareto optimal solution with a higher road condition score and safety factor can be selected.

[0073] The recommended routes are presented to users in a visual way. The routes are drawn on the map and marked with the locations of scenic spots, road conditions, and safety factors. Users can intuitively understand the characteristics of the recommended routes by viewing the visualization results, provide reasons for the recommended routes, and dynamically update the route recommendation results based on user feedback and real-time traffic changes. If users report that a route does not meet expectations, or if there are significant changes in real-time traffic conditions, the route can be re-planned and optimized to provide users with better travel suggestions.

[0074] The recommended routes not only meet the user's preferences, but also have good road conditions and safety, providing users with comprehensive and objective evaluation results and improving their travel experience and satisfaction.

[0075] Furthermore, this embodiment also provides a personalized self-driving travel route planning system, including: The acquisition module is used to obtain the user's historical travel trajectory data and collect real-time traffic data and weather information around the destination; An analysis module is used to extract a user preference feature matrix from the user's historical travel trajectory data using cluster analysis, construct a geographic location-point of interest association network, and divide the geographic location-point of interest association network using a community discovery algorithm to obtain a destination group that meets the user's preferences; The construction module is used to construct the road network cost matrix, use the improved ant colony algorithm to search for paths, and generate candidate routes based on real-time road conditions; The recommendation module is used to map candidate routes to a rasterized map, conduct a comprehensive evaluation of the candidate routes, and select the Pareto optimal solution as the recommended route.

[0076] This embodiment also provides a computer device, which is suitable for the case of a personalized self-driving travel route planning method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement all or part of the steps of the method described in the embodiment of the present invention as proposed in the above embodiment.

[0077] This embodiment also provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method in any optional implementation of the above embodiment is executed. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk.

[0078] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0079] In summary, the present invention can deeply explore users' travel habits and preferences, ensure the smoothness and safety of travel, improve the practicality and reliability of route planning, generate multiple candidate routes that meet user needs and real-time road conditions, provide users with reliable travel plans, and improve users' travel experience and satisfaction.

Claims

1. A personalized self-driving travel route planning method, characterized by: include, Obtain the user's historical travel trajectory data and collect real-time traffic data and weather information around the destination; Using cluster analysis to extract user preference feature matrix from user historical travel trajectory data, constructing a geographic location-point of interest association network, and using community discovery algorithm to divide the geographic location-point of interest association network to obtain destination groups that meet user preferences; Construct a road network cost matrix, use an improved ant colony algorithm to search for paths, and generate candidate routes based on real-time traffic conditions; The candidate routes are mapped to a rasterized map, the candidate routes are comprehensively evaluated, and the Pareto optimal solution is selected as the recommended route.

2. The personalized self-driving travel route planning method according to claim 1, characterized in that: The obtaining of the user's historical travel trajectory data includes obtaining the stored travel records by connecting with the user's smart device, and for users using travel service applications, obtaining the user's travel data from the platform's database; The collecting of real-time traffic data and weather information around the destination includes integrating real-time traffic data from multiple map service providers and acquiring real-time weather data around the destination from a meteorological data service provider.

3. The personalized self-driving travel route planning method according to claim 2, characterized in that: The extracting of the user preference feature matrix comprises: Calculate the straight-line distance between the starting point and the end point of each user's trip to obtain travel distance data, statistically analyze the travel distance data, calculate the average travel distance, and form the user's travel distance characteristics; Analyze the location information of the stopover points in the user's historical travel trajectory data, combine the geographic information system data and the data of the online travel platform to determine the type of scenic spot to which the stopover point belongs, count the number of times and duration of users' stops at different types of scenic spots, divide the user's preference for different types of scenic spots, and form the characteristics of scenic spot types; Analyze the driving speed, acceleration and deceleration pattern data in the user's historical travel trajectory data, extract the user's driving habit characteristics, and form a driving habit feature; Integrate the extracted user travel distance features, scenic spot type features, and driving habit features into a user preference feature matrix; The K-means clustering algorithm is used to perform cluster analysis on the user preference feature matrix. According to the data in the user preference feature matrix, the users are divided into different clusters, and each cluster represents a user group with similar preference characteristics.

4. The personalized self-driving travel route planning method according to claim 3, characterized in that: The obtaining of a destination group that meets the user's preference comprises the following steps: Integrate the tourist attraction database to obtain the location information and attraction types of the scenic area, filter out the points of interest in the area from the geographic information system data and the data of the online travel platform, and determine the initial importance weight of the points of interest based on the type, popularity and user evaluation of the points of interest; For every two points of interest, the geographical distance between the points of interest is calculated, and the geographical location association strength between the points of interest is determined according to the size of the geographical distance, and a geographical location association matrix is ​​constructed; Based on the user preference feature matrix and the attribute information of the points of interest, the similarity between every two points of interest is calculated using the following formula: In the formula, is the feature vector of interest point A, is the feature vector of interest point B, is the similarity between interest point A and interest point B; Construct an association network, take points of interest as nodes, use the geographic location association strength and the similarity of points of interest as the weight of the edge, and construct a geographic location-point of interest association network; The improved community discovery algorithm is used to divide the geographic location-point of interest association network into communities, and the preference characteristics of the target user group are matched with the community division results to determine the destination group that meets the preferences of the target user group. The community division is optimized by maximizing the modularity. The formula is: Among them, Q is the community modularity, is the adjacency matrix of the network, which represents the edge weight between node i and node j. and is the degree of node i and node j, m is the sum of the weights of all edges in the network, γ is the resolution parameter, is the Kronecker function, which is 1 when nodes i and j belong to the same community, otherwise it is 0; The edge weight is dynamically adjusted according to the real-time traffic conditions. The adjustment formula is: in, is the adjusted edge weight, is the initial static edge weight, w(t) is the real-time data, is the weight coefficient; Calculate the matching probability between the target user group and each community. The formula is: in, is the matching probability between the target user group and community c, is the preference feature vector of the target user group, is the preference feature vector of interest point z in the community.

5. The personalized self-driving travel route planning method according to claim 4, characterized in that: Generating candidate routes comprises the following steps: Collect road network data in the target area, analyze road accident records in the area, and calculate the accident rate of each road section; Construct the basic road network cost matrix according to road grade, traffic capacity and accident rate parameters; Use the traffic sensor network and the real-time traffic API of the map service provider to obtain the real-time traffic data of the current road and calculate the real-time travel time and congestion level of each road section; Combined with the basic road network cost matrix and real-time traffic data, a dynamic adjustment strategy for the road network cost is formulated. The formula is: in, is the road network cost at time t after dynamic adjustment, is the basic road network cost, It is the extra cost brought by real-time traffic conditions. is the weight parameter; The basic parameters of the ant colony algorithm are set, and the real-time traffic conditions and historical path experience are comprehensively considered. The pheromone update factor is introduced. The formula is: Where F is the pheromone renewal factor, is the real-time transit time. is the historical average travel time, It is the price of historical experience. is the weight parameter; Each ant starts from the starting point and selects the next node according to the probability rule based on the basic road network cost matrix and pheromone concentration. The formula is: in, is the path selection probability of selecting the next node j, is the set of next nodes that ant k can choose when it is at node i, l is the set of next nodes Any node in is the pheromone concentration on edge (i, j) at time t, is the pheromone weight, is the heuristic information of edge (i, j), β is the heuristic information weight, is the pheromone concentration on the edge (i, l) at time t, is the heuristic information of edge (i,l); After the ant completes a path exploration, the pheromone on the path is updated according to the pheromone update factor. After multiple iterations, multiple candidate routes are generated. After the path search is completed, the pheromone on the optimal path is globally enhanced. The formula is: in, is the enhanced pheromone concentration on the optimal path, is the amount of pheromone added to the edge (i, j) on the optimal path; is the pheromone volatility coefficient; During the movement of ants, the pheromone on the current edge is locally updated, and the formula is: in, is the pheromone concentration after the ants move. is the amount of pheromone added each time an ant passes through edge (i, j).

6. The personalized self-driving travel route planning method according to claim 5, characterized in that: Mapping the candidate routes to the rasterized map comprises the following steps: Collect geographic information data of the target geographic area and determine the appropriate grid size based on the size and resolution requirements of the geographic area; Divide the geographical area into multiple grids to form a grid map, and obtain multiple candidate routes from the path planning algorithm; The geographic coordinates of each road node in the candidate route are converted into corresponding grid coordinates, and the route is represented as a connected grid according to the grid coordinates of the node to generate a grid matrix.

7. The personalized self-driving travel route planning method according to claim 6, characterized in that: The comprehensive evaluation of the candidate routes comprises the following steps: Collect information on scenic spots in the target geographic area, convert the latitude and longitude coordinates of the scenic spots into corresponding grid coordinates, and determine the grid where the scenic spots are located; For each candidate route, the coverage rate of scenic spots, the average road condition score and the average safety factor contained in the grids passed by the route are calculated according to the grid matrix and the associated information of scenic spots grids; Taking scenic spot coverage, road condition score and safety factor as the objectives of multi-objective optimization problem, a model was constructed; A multi-objective optimization algorithm is used to comprehensively evaluate candidate routes and find the Pareto optimal solution set. The final recommended route is selected based on user preferences or specific needs, and the route recommendation results are dynamically updated based on user feedback and real-time traffic changes.

8. A personalized self-driving travel route planning system, based on the personalized self-driving travel route planning method according to any one of claims 1 to 7, characterized in that: include: The acquisition module is used to obtain the user's historical travel trajectory data and collect real-time traffic data and weather information around the destination; An analysis module is used to extract a user preference feature matrix from the user's historical travel trajectory data using cluster analysis, construct a geographic location-point of interest association network, and divide the geographic location-point of interest association network using a community discovery algorithm to obtain a destination group that meets the user's preferences; The construction module is used to construct the road network cost matrix, use the improved ant colony algorithm to search for paths, and generate candidate routes based on real-time road conditions; The recommendation module is used to map candidate routes to a rasterized map, conduct a comprehensive evaluation of the candidate routes, and select the Pareto optimal solution as the recommended route.

Images