A bus route optimization design method and system based on residents' travel big data

Abstract

The present invention discloses a bus route optimization design method and system based on resident travel big data, the method comprising the steps of: obtaining mobile phone signaling data of residents in a target area; obtaining the stop points of residents' travel trajectories; dividing the target area into a number of traffic areas; obtaining an OD matrix based on the stop points of residents' travel trajectories and traffic areas; constructing a bus route planning model; generating a number of bus alternative routes based on a shortest path algorithm; obtaining a target bus route from the alternative routes based on a genetic algorithm and in combination with an objective function and an OD matrix. Compared with the prior art, the present invention not only innovatively combines spatiotemporal big data and genetic algorithms in technical means, improving the accuracy and efficiency of bus route planning, but also demonstrates strong advantages in system architecture and data processing capabilities, and can provide a more scientific, dynamic and sophisticated optimization solution for the urban public transportation system, contributing to the efficient operation of urban transportation and the construction of smart cities.

Description

Technical Field

[0001] The present invention belongs to the technical field of public transportation planning and data mining, and in particular relates to a method and system for optimizing the design of public transportation routes based on residents' travel big data. Background Art

[0002] With the acceleration of urbanization, especially the expansion of large and medium-sized cities, traffic congestion and environmental pollution are becoming increasingly prominent. According to domestic and foreign traffic data and research analysis, as the size of cities expands, the number of motor vehicles continues to increase, and the travel needs of residents grow, the traffic flow of urban roads is gradually approaching or even exceeding its carrying capacity. Traditional traffic management methods and infrastructure construction are difficult to fundamentally solve these problems in the short term. In particular, the public transportation system has not been effectively optimized and developed in many cities, and the utilization rate of public transportation is relatively low, resulting in urban traffic congestion, increased carbon emissions, and rising travel costs for residents.

[0003] At present, an important issue facing urban transportation planning is how to find a balance between limited transportation resources and the increasing travel needs of residents. As a green and low-carbon mode of transportation, the route planning and design of the bus system are of great significance to improving the attractiveness of public transportation and reducing urban traffic pressure. However, traditional bus route design methods often ignore the temporal and spatial characteristics of residents' travel, the volatility of traffic flow, and the complexity of bus operation. Therefore, how to improve the efficiency of bus routes and optimize residents' travel experience through scientific planning and optimization has become a key task in smart city transportation planning.

[0004] In recent years, with the development of information technology, especially the widespread application of technologies such as big data, the Internet of Things and cloud computing, traffic flow analysis and bus route optimization based on spatiotemporal big data have become research hotspots. Big data technology can collect, process and analyze residents' travel data in real time. These data have the characteristics of high frequency and strong timeliness, and can reflect the real-time status of urban traffic and the real travel needs of residents. However, traditional bus route planning often relies on static survey data, which makes it difficult to timely reflect the changes in urban traffic flow and the dynamic characteristics of residents' travel needs.

[0005] In this context, how to rationally plan bus routes based on spatiotemporal big data through computer technology and advanced optimization algorithms has become an effective means to improve the level of bus service. Existing bus route optimization methods are mostly based on a single goal, such as the shortest route, the least transfer, etc., but ignore the comprehensive consideration of multiple optimization goals. For example, the design of bus routes not only needs to consider the residents' travel time and the total mileage of bus routes, but also needs to optimize the walking distance of residents, the passenger capacity of buses and other factors, which requires the use of multi-objective optimization methods in the planning process.

[0006] At present, multi-objective optimization algorithms, such as genetic algorithms and simulated annealing, have been widely used in the field of transportation, especially in bus route optimization. Genetic algorithms have become an important tool for solving multi-objective optimization problems due to their powerful global search capabilities and optimization performance. However, the application of multi-objective genetic algorithms in existing research has not fully combined the actual spatiotemporal big data of cities, and there is a lack of research on customized optimization for specific needs in bus route design. Therefore, how to design a bus route planning method that meets actual needs by combining spatiotemporal big data and genetic algorithms is an important challenge in the current field of bus optimization. Summary of the invention

[0007] In order to solve the above technical problems, the present invention proposes a bus route optimization design method and system based on residents' travel big data based on spatiotemporal big data analysis and genetic algorithm. The reasonable planning of bus routes is realized through intelligent optimization algorithm to improve the service level of public transportation, reduce urban traffic pressure, promote green travel, and provide theoretical support and technical solutions for smart city traffic management.

[0008] The technical solution provided by the present invention is specifically as follows:

[0009] A bus route optimization design method based on a multi-objective genetic algorithm comprises the following steps:

[0010] S1. Obtaining mobile phone signaling big data of residents in the target area;

[0011] S2, based on mobile phone signaling data, obtain residents’ travel trajectory stop points;

[0012] S3, based on Thiessen polygons, divide the target area into several traffic areas;

[0013] S4, based on the residents’ travel trajectory stop points and traffic communities, obtain the residents’ travel OD matrix;

[0014] S5. Construct a bus route planning model, including objective functions and constraints;

[0015] S6. Generate several alternative bus routes based on the shortest path algorithm;

[0016] S7. Based on the genetic algorithm and combined with the objective function and the residents' travel OD matrix, the target bus route is obtained from the alternative routes.

[0017] Furthermore, step S1 also includes data cleaning and data sorting of the acquired mobile phone signaling big data, wherein the mobile phone signaling data includes user ID, base station longitude and latitude, base station arrival time and user stay time; the data cleaning includes: filtering out mobile phone signaling data occurring in the target area according to the coordinate range, and deleting mobile phone signaling data simultaneously recorded by two different base stations; the data sorting includes: re-sorting the acquired mobile phone signaling data according to the user ID.

[0018] Furthermore, in step S2, the travel trajectory of each resident is obtained and the stop points are extracted according to the mobile phone signaling data within the set time period. The extraction of the stop points is based on the DBSCAN clustering algorithm, which traverses each travel trajectory to obtain the corresponding stop point information, including:

[0019] Traverse the trajectory points in the trajectory TRAj. If the time interval between the trajectory points in the ε-neighborhood of the trajectory point Pi and the trajectory point Pi is less than △t, then mark the trajectory point Pi as a core point;

[0020] Connect the core points that are directly accessible through density accessibility to form several clusters, which represent the stay areas;

[0021] The latitude and longitude coordinates of the core points in the same cluster are averaged to obtain the latitude and longitude coordinates of the stay points in the stay area;

[0022] The residence time of the stay point is calculated based on the maximum and minimum timestamps of the core points in the same cluster.

[0023] Furthermore, in order to improve the efficiency of the algorithm, before extracting the stop point information of each travel trajectory, each travel trajectory is divided into several sub-trajectories, and then the stop point information of each sub-trajectory is extracted; the travel trajectory segmentation step includes:

[0024] Traverse the trajectory points in the trajectory TRAj: if the time interval between the trajectory point Pi and its previous trajectory point is greater than the set threshold, it indicates that there is data missing between the trajectory point Pi and its previous trajectory point, and the trajectory point Pi is deleted; if the speed of the trajectory point Pi relative to its previous trajectory point is greater than the threshold V1, it indicates that the trajectory point Pi is a data drift point, and the trajectory point Pi is deleted; if the speed of the trajectory point Pi relative to its previous trajectory point is greater than the threshold V2 and less than the threshold V1, the trajectory point Pi is added to the judgment queue;

[0025] If the length of the judgment queue is greater than the threshold d, the trajectory formed by the current trajectory points in the judgment queue is taken as a sub-trajectory; the queue is cleared and the judgment segmentation is continued.

[0026] Furthermore, step S3 includes: based on the current bus stops, dividing the target area into several traffic zones according to the Thiessen polygon principle; merging the traffic zones in combination with the street administrative divisions, population density and actual road network of the target area; the shape of the traffic zone is a non-annular convex shape with no isolated islands inside, there is no overlap between the traffic zones, and the travel ratio in each traffic zone does not exceed the set threshold.

[0027] Further, step S4 includes:

[0028] Construct an NxN OD matrix and initialize its element values ​​to 0, where N is the number of divided traffic zones;

[0029] Mapping the trajectory stop points obtained in step S2 to each traffic zone divided in step S2;

[0030] The two stop points of the same travel trajectory with adjacent stop times are regarded as a pair of starting and ending points, the earlier one is the starting point and the later one is the end point. The PNPoly algorithm is used to determine the traffic zones where the starting and ending points are located. If the starting and ending points cannot be mapped to the traffic zones, the point is considered to be beyond the scope of the target area.

[0031] For any starting and ending point pair, if the starting point and the end point belong to different traffic zones, the starting and ending point pair forms a travel process, and the OD matrix element value corresponding to the starting and ending points is increased by one;

[0032] Traverse each travel trajectory and complete the update of the OD matrix element values.

[0033] Furthermore, in step S5, the objective function of the planning model is:

[0034]

[0035] in,

[0036] The constraints are:

[0037] Among them, T represents the travel time of residents, Q represents the rate of return of the bus operator, μ1 and μ2 are the weight coefficients of the objective function, which are used to balance the magnitude of the two parts of the objective function; N is the total number of bus lines in the region, n k represents the number of bus nodes passed by the k-th bus line, S ki represents the i-th bus node on the k-th bus route, d(S ki ,S ki+1 ) represents the distance between bus nodes i and i+1 on the kth bus line, v represents the evaluated running speed of the bus, t sis the stop time of the bus at each bus node, r ij represents the number of passengers between bus nodes i and j, e1 represents the fare, e2 represents the government subsidy amount per person for residents who choose to travel by bus, L ij represents the path distance between bus nodes i and j, G ij represents the number of buses passing between bus nodes i and j, c is the operating cost per unit mileage of each bus on the road, R′ represents the roads where buses are allowed to pass, R represents all roads within the research scope, and l min Indicates the minimum mileage of a single bus line, l max Indicates the maximum mileage of a single bus line, l k represents the mileage of the k-th bus line, d k represents the spatial straight-line distance between the starting and ending points of the kth line, η max Indicates the maximum value of the non-linear coefficient of the preset bus route, h ij represents the number of transfers required for residents to complete a trip from bus node i to j, h max Indicates the preset maximum number of transfers, Z ij represents the number of repeated routes between bus nodes i and j, Z max Indicates the maximum number of repeated routes between two preset bus nodes.

[0038] Further, step S6 includes:

[0039] Use Dijkstra algorithm to solve the shortest path between any pair of starting and ending points and use it as an alternative bus route;

[0040] The deletion path algorithm in the KSP algorithm is used to calculate the second shortest path to expand the range of the candidate route set;

[0041] The candidate routes in the candidate route set are screened and sorted based on compliance.

[0042] Furthermore, in step S7, a genetic algorithm is used in combination with an objective function to optimize the target bus route from the candidate bus routes, including the steps of:

[0043] The routes in the candidate route set are numbered, and the chromosome length is determined according to the number of routes included in the target solution. The chromosome of the population individuals corresponding to each route is encoded using a binary string to complete the population initialization of the genetic algorithm; the genetic iteration is started, and the selection operation, crossover operation and mutation operation of the population individuals are performed in each iteration round until the set iteration round or the number of population individuals or the chromosome of the population individuals no longer changes is reached; the route corresponding to the population individuals retained after the iteration is completed is used as the final target bus route;

[0044] In the selection operation, the fitness of the individuals in the population is obtained based on the objective function of the bus route planning model. The smaller the value of the objective function, the higher the fitness of the individuals in the population, and the greater the survival probability. The individuals in the population are selected and retained according to the survival probability. The survival probability P of the individual n in the population is n for: Among them, f n It represents the objective function value of individual n in the population, where N is the number of individuals in the population participating in this round of iteration selection. When calculating the objective function value, the OD quantity in the OD matrix needs to be distributed to the urban road network: the shortest path between each starting and ending point in the OD matrix is ​​calculated, and the OD quantity of each starting and ending point is distributed to the shortest path, and the total OD quantity on each road section is obtained by superposition.

[0045] In the crossover operation, two individuals of the population are first randomly selected to pair up as parent individuals, and then two crossover points with different positions are generated in each group of paired parent individuals. Then, a random number in the interval [0,1] is generated. If this random number falls within the preset crossover probability range, a crossover operation is performed to exchange the chromosome genes after the crossover point of the parent individual.

[0046] In the mutation operation, a population individual is randomly selected with a certain probability, and a random gene on its chromosome is transformed into other alleles.

[0047] A bus route optimization design system based on the above method includes: a spatiotemporal big data storage module for storing collected mobile phone signaling data; a stop point extraction module for cleaning the mobile phone signaling data and extracting track stop point information; a geographic information processing module for dividing the target area into a number of traffic areas according to GIS data; an OD matrix generation module for combining traffic areas with track stop point information to generate an OD matrix; an alternative route generation module for generating a number of alternative bus routes; a bus route optimization module for optimizing the target bus route from the alternative bus routes; and a visualization module for displaying the optimized target bus route.

[0048] Compared with the prior art, the present invention has at least the following beneficial effects:

[0049] First of all, the present invention makes full use of the combination of spatiotemporal big data and genetic algorithms. In traditional bus route planning, it often relies on relatively simplified models and manual experience, lacking accurate analysis of complex traffic flows and residents' travel patterns. The present invention uses large-scale spatiotemporal big data based on mobile phone positioning data, which can reflect the specific travel needs of residents in real time. These data are widely sourced and updated in a timely manner, and can describe in detail the travel trajectory, travel time, frequency and other information of residents, ensuring the high timeliness and accuracy of the data. Compared with traditional methods, the use of spatiotemporal big data analysis provides a more accurate basis for bus route planning, greatly improving the rationality and adaptability of planning.

[0050] Secondly, the introduction of genetic algorithms enables accurate solutions to multi-objective optimization. Traditional bus route optimization often focuses on a single goal, such as the shortest route or the lowest cost, and ignores the trade-offs and optimization of multiple goals. The multi-objective genetic algorithm used in the present invention can comprehensively consider multiple optimization goals and find the optimal balance between multiple goals, ensuring that while meeting the travel needs of residents, the operating costs of bus companies are minimized and the utilization efficiency of resources is improved. Through the intelligent search of genetic algorithms, the system can quickly iterate and find the most suitable bus route plan, avoiding the tedious manual debugging and adjustment in traditional methods, and greatly improving efficiency and accuracy.

[0051] In addition, the present invention demonstrates innovation in big data processing and system architecture. Traditional bus route planning mostly relies on data from a single source, with a small amount of data and a low update frequency, resulting in poor adaptability and flexibility of planning results. The present invention, by building a distributed data storage and processing platform, can efficiently integrate multi-source data from geographic information systems (GIS), mobile communication data, and urban resident information databases to form a comprehensive urban travel data foundation. The big data storage and processing platform is based on the Hadoop framework and can cope with the storage and real-time analysis needs of massive data. This system architecture not only improves the data processing capabilities, but also ensures the timeliness of data updates, providing continuous and reliable data support for subsequent bus route planning.

[0052] In terms of data analysis, the present invention uses an advanced trajectory data processing algorithm. The trajectory stop points are extracted and analyzed through an improved density clustering algorithm (DBSCAN). Compared with the traditional traffic demand prediction method based on hot spot analysis or static survey, the present invention can dynamically track the travel trajectory of residents based on the user's real-time mobile phone positioning data, thereby accurately extracting the key nodes of traffic flow. This method can more accurately reflect the actual travel needs of residents, avoid data distortion or sampling bias that may occur in traditional methods, and make the effect of bus route planning more in line with actual conditions.

[0053] Furthermore, the present invention can provide a visual display for the optimization of bus routes. By presenting the optimized route planning in a visual way, relevant departments and operators can view and analyze the optimization results more intuitively. This visual display not only improves the understanding of the planning results, but also facilitates later decision-making and adjustments. In particular, among multiple options, it can help decision makers quickly compare the advantages and disadvantages of different options through visualization, thereby improving decision-making efficiency.

[0054] Finally, the present invention has high adaptability and scalability. Since the system is based on big data and intelligent algorithms, it has strong adaptability and can adjust and optimize bus route planning according to the characteristics of different cities, different scales, and different population densities. In the future, with the advancement of urbanization and the continuous changes in residents' travel needs, the system of the present invention will be able to continuously optimize and adjust bus routes based on new data, providing strong support for the construction of smart cities and the sustainable development of urban transportation. At the same time, due to the flexible system architecture, it can be modularly expanded according to needs in the future, and more functions can be added, such as real-time scheduling and optimization combined with weather data, traffic accident information, etc.

[0055] In general, compared with the existing technology, the present invention not only innovatively combines spatiotemporal big data and genetic algorithms in technical means to improve the accuracy and efficiency of bus route planning, but also demonstrates strong advantages in system architecture, data processing capabilities and intelligence level. It can provide a more scientific, dynamic and sophisticated optimization solution for the urban bus system, and contribute to the efficient operation of urban transportation and the construction of smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0057] Figure 1 It is a schematic diagram of the process of a bus route optimization design method provided by an embodiment of the present invention;

[0058] Figure 2 It is a visualization processing result of mobile phone signaling data on a GIS map provided by an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of traffic zone division provided by an embodiment of the present invention;

[0060] Figure 4 This is an example diagram of bus route topology provided by an embodiment of the present invention;

[0061] Figure 5 It is a schematic diagram of the framework of a bus route optimization design system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by ordinary technicians in this field without making creative work are all within the scope of protection of the present invention.

[0063] Embodiment 1

[0064] This embodiment provides a method for optimizing bus routes based on residents' travel big data, which includes data collection and processing, traffic demand analysis, bus route planning model construction, model optimization solution, etc. Specifically, Figure 1 As shown, the method mainly includes the following steps:

[0065] 1. Obtain residents’ mobile phone signaling data

[0066] (1) Introduction to mobile phone signaling data

[0067] Mobile phone signaling data comes from mobile communication companies such as China Mobile and China Telecom. It is the communication data between mobile phone users and transmitting base stations or micro stations. As long as the mobile phone is turned on and the operator's words are displayed on the mobile phone screen, signaling data begins to be generated. When users use their mobile phones to make calls, send and receive text messages, browse the Internet, and other communication behaviors, they will send communication relationships with base stations near the mobile phone. Since the location of the communication base station is fixed and known, the location information of the base station reflects the location of the user. Therefore, the mobile phone signaling data field always contains information such as time and location, as shown in Table 1.

[0068] Table 1 Mobile phone signaling data structure

[0069]

[0070] With the increasing number of mobile phone users, signaling data based on mobile phone positioning technology has gradually attracted the attention and application of scholars. Mobile phone signaling data has the following advantages:

[0071] 1) The data volume is large and easy to obtain. Taking Beijing Mobile as an example, there are more than 18 million samples in Beijing, and 1 billion mobile phone signaling data are generated every day. These data are automatically recorded in the systems of major communication companies, and no manual data collection is required.

[0072] 2) The data is real and objective. At present, there are more than 96 smartphones per 100 people in China. With almost one smartphone per person, the real movement status of the phone holders can be objectively recorded through base stations, GPS, Bluetooth and other technologies.

[0073] By acquiring the mobile phone signaling data of residents in the target area, constructing a mobile phone signaling data set, and then cleaning and mining the data, we can obtain real and objective information on residents’ travel and traffic flow in the target area. Figure 2 The figure shows the visualization result of mobile phone signaling data of residents in Jingkou District, Zhenjiang City during a certain period of time on the GIS map.

[0074] Table 2 Mobile phone signaling dataset

[0075]

[0076] As shown in Table 2, in the mobile phone signaling data set, each user has a unique ID identification code corresponding to it. The longitude and latitude in the data set are the longitude and latitude of the base station. When the user's mobile phone enters the service range of the base station, the base station automatically records the timestamp of arrival at the base station and the user's stay time in the service range of the base station.

[0077] (2) Mobile phone signaling data preprocessing

[0078] Mobile phone signaling data preprocessing mainly includes data cleaning and data sorting.

[0079] Data cleaning: Based on the original data set, first, the mobile phone signaling data occurring in the target area (such as Jingkou District, Zhenjiang City) is filtered out within a fixed coordinate range. By observing the data, it is found that there is a situation where the user's information is recorded by two base stations at the same time, but the stay time displayed by one of the base stations is 0. It is speculated that when the user's position was recorded, the user was at the edge of another base station, resulting in a conflict between the two base stations in the process of recording information, which will interfere with the subsequent processing of the present invention, so these data need to be deleted as interference data.

[0080] Data sorting: The original data set is recorded according to the time when the mobile phone user enters the service range of the base station. In order to accurately analyze the travel characteristics and travel purposes of urban residents, the present invention re-sorts the data set according to the user's ID for use in the subsequent processing steps of the present invention.

[0081] 2. Obtain residents’ travel origin and destination data

[0082] (1) Resident travel chain extraction

[0083] After the initial acquisition and preprocessing of mobile phone signaling data, the core work step is to enter the travel chain extraction. The so-called travel chain refers to a single trip within the target area, such as a trip to work. It captures the activity range (such as a residential residence) of a mobile phone user that has not changed before leaving at a certain time in the morning. The position changes continuously at a certain speed, and a series of corresponding location information is collected until it arrives at a certain place (office). The activity range is within the specified time and the small radius specified by the specified threshold. The travel trajectory corresponding to this series of location information (staying points) from the starting point to the arrival point can be identified as a complete travel chain. The key to travel chain extraction lies in the extraction of stay points.

[0084] The main methods for extracting stay points in the prior art are time and distance threshold judgment and clustering algorithm extraction. Using time and distance thresholds to judge stay points has low time complexity, but low tolerance for noise data, and stay points are easily segmented. In contrast, the stay point recognition algorithm based on cluster analysis has better tolerance for noise. Compared with other clustering algorithms, the density-based clustering algorithm does not require the number of clusters to be set in advance, nor does it require the prototype to be initialized, and can detect clusters of any shape, but the traditional density-based spatial clustering algorithm (DBSCAN algorithm) has a high time complexity.

[0085] The main steps of the DBSCAN algorithm include:

[0086] (1) Identify core points: For each data point, calculate the number of points in its ε-neighborhood. If this number is greater than or equal to MinPts, then the point is a core point. ε is the neighborhood radius, which represents the neighborhood range around a point. MinPts is an integer parameter, indicating how many points must be within the ε radius to be considered a core point.

[0087] (2) Cluster formation: Starting from the core point, all directly density-reachable core points are connected through density reachability to form a cluster.

[0088] (3) Processing boundary points: Assign boundary points to clusters associated with core points in their neighborhood.

[0089] (4) Deleting noise points: Finally, the points that are not included in any cluster are marked as noise points.

[0090] In the scenario where only the spatial dimension is considered, the longitude and latitude data are mapped to a two-dimensional space. After visualization is achieved, the data can be manually observed to cluster in clusters. However, when the computer executes the algorithm, it has no way of knowing the relationship between vectors, so it is necessary to traverse the vectors in the entire vector space and perform calculations between them. Therefore, it takes a lot of time to use clustering algorithms to identify stop points. The worst-case time complexity of the DBSCAN algorithm is O(n 2 ). In addition, the existing technology usually adopts the method of averaging the points of the same category to obtain the cluster center, but directly uses the arithmetic mean of the coordinates of each point as the coordinate of the stop point, ignoring the information contained in the trajectory itself, and often deviates from the actual stop area of ​​the moving object. Existing research has given little consideration to the optimization processing of indoor stop with a large amount of drift noise and is not comprehensive enough, and has not considered the complexity and diversity of trajectory stop.

[0091] In view of the shortcomings of the existing methods and the abnormal situations that may occur in practice, the present invention improves the DBSCAN algorithm so that the improved algorithm can be better applied to actual situations. The main improvements are:

[0092] (1) The DBSCAN algorithm mainly clusters from the perspective of distance, but in spatiotemporal data, there is also an important constraint factor of time. Since the collected trajectory data is non-isochronous, that is, the time intervals between trajectory data are different, the density of the ε-neighborhood cannot be represented by the number of trajectory points in the ε-neighborhood. At the same time, the mobile phone positioning method is prone to missing positioning data indoors, resulting in sparse indoor stay data compared to the actual situation, which will cause misjudgment for density-based clustering algorithms such as DBSCAN.

[0093] In view of this situation, the present invention improves the original input parameters of DBSCAN from the original (ε, MinPts) to (ε, △t), and uses the time interval △t between two trajectory points to replace the threshold value MiPts of the number of samples in the ε-neighborhood in the original algorithm. An important purpose of stop point extraction is to extract the longitude and latitude coordinates of the stop point. The stop point is a representative of the entire stop area and provides important information for subsequent steps. Most of the existing algorithms use the mean longitude and latitude of the stop area as the longitude and latitude of the stop point. In the present invention, the mean longitude and latitude of the core points in the same cluster family is used as the longitude and latitude of the corresponding stop point, so that the stop point coordinates are more biased towards the area with dense trajectory points.

[0094] The specific stop point extraction algorithm in the present invention is shown in Algorithm 1.

[0095]

[0096] (2) During the process of trajectory data collection, data collection may be interrupted due to equipment problems or human factors, resulting in missing trajectories. This situation will interfere with the extraction of stop information. For example, when a mobile object X stops at a certain place A in the morning and evening, but there is a long period of data missing between the morning and evening, without considering the time factor, the improved DBSCAN algorithm based on (1) can only identify one stop point A, and the stay time is from morning to evening. In this case, the information contained in the original trajectory cannot be accurately described. Therefore, it is necessary to mark the situation where long-term data is missing.

[0097] (3) The time complexity of the DBSCAN algorithm is O(n 2 )(n is the number of data points), which will cause a lot of time consumption when facing a large amount of trajectory data. In order to improve this problem, consider cutting and dividing the trajectory data.

[0098] All existing algorithms process trajectory data within one day based on the date. Therefore, we can filter out high-speed trajectory points based on the speed indicator, and further divide the trajectory into multiple low-speed areas without data missing by combining the mark of long-term data missing in (2), and generate candidate trajectories for the stay area. Set the trajectory segmentation speed threshold to filter these trajectory points, that is, those with a speed greater than the threshold are definitely not in a stay or wandering state, and segment the trajectory after filtering. Considering the possible impact of drift data on trajectory segmentation, use the fast movement judgment queue to judge whether the moving object is in a non-staying state. Set the segmentation threshold d, and when d trajectory points with a speed greater than the trajectory segmentation speed threshold appear continuously, the trajectory is segmented.

[0099] For the specific candidate stay trajectory identification algorithm, please refer to Algorithm 2.

[0100]

[0101] After trajectory segmentation, the original data is divided into thousands of candidate areas of stay points. Different candidate areas can be clustered separately to identify specific stay points without traversing the entire trajectory space, thereby improving the efficiency of clustering.

[0102] 3. Divide traffic zones

[0103] Traffic subdivisions are the basic spatial units for traffic surveys and are divided in traffic planning to study the generation and distribution of trips. In order to analyze the distribution of residents' travel origins and destinations, the research area (target area) of traffic planning is usually divided into multiple traffic subdivisions. In order to fully understand and grasp the travel characteristics of traffic sources and the traffic characteristics between various traffic sources, it is necessary to carefully grasp the generation, attraction and distribution of urban traffic trips. However, the finer the division of traffic subdivisions, the more in-depth the research on each traffic subdivision, the greater the workload, the greater the difficulty of conducting traffic surveys, analysis, predictions, etc., and the accuracy is difficult to guarantee. Therefore, in the process of investigation and research, it is necessary to divide traffic subdivisions according to the purpose of traffic subdivisions and certain division principles.

[0104] The traffic zone division principles adopted in this embodiment are as follows:

[0105] (1) The transportation communities within the target area should maintain appropriate homogeneity in terms of area, population, and trip volume to improve the accuracy of trip generation estimates;

[0106] (2) The travel ratio within each traffic zone should not exceed 10-15% to avoid loss of travel information;

[0107] (3) The shape of the traffic zone should be as regular as possible, ensuring that it is convex;

[0108] (4) The shape of the traffic zone cannot be circular and there cannot be any “islands” inside;

[0109] (5) Traffic zones do not overlap with each other, and traffic zones within the target area can cover the entire area.

[0110] In order to ensure the feasibility of subsequent bus route optimization, it is necessary to retain most of the current bus stops in the target area. Therefore, correctly obtaining the travel flow in the area near each bus stop becomes a key issue in optimizing the route.

[0111] Based on the above principles and premises, the present invention chooses to divide the traffic area based on the current bus stops with Thiessen polygons. Thiessen polygons, also known as von Lonoy diagrams (Voronoidiagram), are a set of continuous polygons composed of perpendicular bisectors connecting two adjacent line segments. The distance from any point in a Thiessen polygon to the control point that constitutes the polygon is less than the distance to the control points of other polygons. Considering the complex and dense road network in the city, pedestrians can go to the nearest bus stop at a distance that is close to a straight line. If it is assumed that pedestrians choose the bus stop with the nearest walking distance among all routes that can reach their destination, then the division based on the bus stops with Thiessen polygons can effectively cover the traffic needs of nearby areas. After the traffic area is divided with Thiessen polygons, in order to meet the above-mentioned division principles, it is also necessary to combine the current street administrative area division, population density and actual road network to merge the partitions. Such as Figure 3 , in ArcMap of ArcGis platform, Jingkou District of Zhenjiang City is divided into 168 traffic zones.

[0112] 4. Get OD matrix

[0113] The OD matrix (Origin-Destination Matrix) is an important tool in the field of transportation to describe the travel flow of residents or vehicles between cities or regions. It is arranged in the form of source points (Origin) and destinations (Destination), recording the travel volume from one area to another, and is usually used in applications such as transportation planning, management, and control.

[0114] The principle of obtaining the OD matrix in the present invention is to map the mobile phone base station to each urban traffic community, utilize the existing mobile phone communication network facilities, and use the CO0 (Cel1ofOrigin) positioning method as the basis, and use the centroid coordinates of the mobile phone base station community to determine the position coordinates of the mobile phone within its service range; the change pattern of the continuous community switching of the mobile phone on the network communication layer is represented as a group of ordered continuous discrete points (staying points) on the plane, and these ordered stay points are regarded as the starting and ending points of the mobile phone user after preprocessing, then different ordered discrete point sets represent the point trajectory of each mobile phone user's travel (that is, the travel chain of the mobile phone user): by tracking and recording a large number of mobile phone point trajectories, the OD information between each traffic community can be counted, and the corresponding regional dynamic OD matrix can be obtained.

[0115] The idea of ​​obtaining the regional dynamic OD matrix of the present invention is as follows:

[0116] (1) Select a database to store traffic zones and mobile phone location data, and use the previously divided traffic zone object data to initialize the traffic zone class and OD matrix class.

[0117] (2) Using the mobile phone base station positioning point data (mobile phone signaling data), the mobile phone user code USER_ID field is used as the primary keyword, and the user positioning time TIME field is used as the secondary keyword to arrange them in ascending order: Then, the unique value set of the USER_ID field in the mobile phone base station positioning point data is obtained, and the values ​​in the unique value set are used as the query conditions to extract the positioning data of each mobile phone user in a specific time period in turn: By counting all the starting and ending point pairs of each mobile phone user (the method for obtaining the starting and ending points refers to the method for obtaining the travel chain stop points mentioned above), the total regional dynamic OD matrix is ​​obtained.

[0118] The specific calculation method is as follows;

[0119] Step 1: Create and initialize od[a+1][a+1] to store the OD matrix.

[0120] Step 2: In the set of starting and ending points W, traverse each starting and ending point (Spi, Epi) in order, map the starting and ending points in the traffic zone set A, and use the PNPoly algorithm to determine in which traffic zone each starting and ending point is located. The idea of ​​the PNPoly algorithm is: if a point is inside a certain traffic zone, then draw a ray from this point. If the number of intersections with the traffic zone boundary is an odd number, then the point is inside this traffic zone: if it is an even number, then the point is outside this traffic zone, and go to the next traffic zone to continue judging until all traffic zones in the set are traversed. If the starting point and the end point cannot be mapped to the traffic zone, it is considered that the point is out of the target area, and its mapping value is recorded as -1. The area ID where the starting point Spi is located is s id The region ID of the end point Epi is e id If id and e id If s is not equal, and both mapping values ​​are not -1, it can be determined that the origin and destination points form a mobile phone travel process, and the corresponding elements in the OD matrix are added by 1. id and e id If they are equal, it can be determined as a trip within the traffic zone and can be ignored.

[0121] 5. Construct bus route planning model

[0122] (1) Basic assumptions of the model

[0123] In order to simplify the model and facilitate subsequent research, the present invention makes the following assumptions about the bus route planning model:

[0124] 1) The research scope of this model is limited to urban buses, and does not consider intercity buses, urban-rural buses and other types.

[0125] 2) All bus routes operate in both directions. There are no one-way or two-way circular bus routes.

[0126] 3) All buses on each bus route stop when passing through the bus nodes on the route.

[0127] 4) The average speed of all buses on the road is a fixed value.

[0128] 5) When the starting and ending points of a bus route are the same and there are multiple routes in between, all traveling residents will choose the route with the shortest distance.

[0129] 6) When there is no direct route between two bus nodes, all traveling residents will choose a route with fewer transfers and shorter distance.

[0130] 7) All bus routes have the same departure frequency, there is no mutual interference between buses, and each route operates in normal order.

[0131] (2) Constraints and objective function

[0132] The present invention divides the research process of bus route layout into two parts. The first part: generating alternative routes by calculation, using the shortest path and K shortest path algorithms to calculate the alternative route set. The second part: combining and optimizing into a network, using a genetic algorithm to combine the routes in the alternative route set, so as to find the optimal route combination scheme.

[0133] 1) Constraints of bus routes

[0134] ① Mileage constraints of bus routes

[0135] The mileage of bus routes is related to factors such as urban structure and residents' travel distance. If the route mileage is too long, it will increase the driving pressure of bus drivers and increase operating costs; if the route mileage is too short, it will cause a waste of public resources, increase the probability of residents transferring buses, and reduce direct access. Therefore, the mileage of bus routes should be reasonably controlled within a certain range. By referring to the "Urban Road Traffic Planning and Design Specifications GB50220-95" promulgated by the Ministry of Construction in 1995, combined with the development laws of cities in recent years, the present invention takes the lower limit of bus route mileage as 10 kilometers and the upper limit as 16 kilometers.

[0136] ② Non-linear coefficient constraints of bus routes

[0137] The non-linear coefficient of a bus route refers to the ratio of the actual operating mileage of the bus route to the spatial straight-line distance between the starting and ending points. Too many turns on the route will increase the travel time of residents, increase the extra detour mileage, cause road congestion, and increase the labor intensity of bus drivers; but due to the design of the urban functional layout and the actual direction of the city's roads, the bus route cannot travel along the shortest distance in space. Therefore, the non-linear coefficient of the bus route should be controlled within a reasonable range. According to the "Urban Road Traffic Planning and Design Specifications GB50220-95" promulgated by the Ministry of Construction in 1995, the non-linear coefficient of public transportation routes should not be greater than 1.4.

[0138] ③Constraints on the number of repeated routes between bus nodes

[0139] The increase of repeated routes between bus nodes can improve the accessibility of the area and facilitate residents' travel. However, too many repeated routes will reduce the uniformity of bus route distribution and easily cause buses to queue up to enter the station, resulting in traffic congestion in the bus station and surrounding areas. The specific mathematical expression is formula Z ij ≤Z max , where Z ij represents the number of repeated routes between bus node i and bus node j, that is,

[0140]

[0141] Generally speaking, the number of bus routes running repeatedly between the same nodes should not exceed 4. However, in places where traffic is particularly concentrated, the demand for buses will be greater. Therefore, the constraint on the number of repeated routes between bus nodes needs to be analyzed in combination with actual conditions when it is applied specifically.

[0142] 2) Objective function of bus routes

[0143] The present invention takes the interests of both traveling residents and public transportation operating companies into consideration, takes minimizing the travel time of traveling residents and maximizing the daily rate of return of public transportation companies as optimization goals, integrates the two selected objective functions into a mathematical expression as a single objective function through a linear weighted method, and uses the converted single objective function as the objective function of bus route planning. The mileage of bus routes, the non-linear coefficient of bus routes, and the number of repeated routes between bus nodes are used as constraints for bus route planning. Combining the objective function with the constraints, a mathematical model is established, and the OD matrix of residents' travel obtained from mobile phone positioning data is used as the basis to study the bus route planning problem that takes into account the actual travel needs of residents, that is, to obtain the bus route that minimizes the objective function as much as possible under the constraints. The specific mathematical expression of the bus route planning model is as follows:

[0144] Objective function:

[0145] in,

[0146] Constraints:

[0147] In the above formula, T represents the travel time of residents, Q represents the rate of return of the bus operator, μ1 and μ2 are the weight coefficients of the objective function, which are used to balance the magnitude of the two parts of the objective function and cannot be 0; N is the total number of bus lines in the region, n k represents the number of bus nodes passed by the k-th bus line, S ki represents the i-th bus node on the k-th bus route, d(S ki ,S ki+1 ) represents the distance between bus nodes i and i+1 on the kth bus line, v represents the evaluated running speed of the bus, t s is the stop time of the bus at each bus node, r ij represents the number of passengers between bus nodes i and j, e1 represents the fare, e2 represents the government subsidy amount per person for residents who choose to travel by bus, L ij represents the path distance between bus nodes i and j, G ij represents the number of buses passing between bus nodes i and j, c is the operating cost per unit mileage of each bus on the road, R′ represents the roads where buses are allowed to pass, R represents all roads within the research scope, and l min represents the minimum mileage of a single bus line, l max Indicates the maximum mileage of a single bus line, l k represents the mileage of the k-th bus line, d k represents the spatial straight-line distance between the starting and ending points of the kth line, η max Indicates the maximum value of the non-linear coefficient of the preset bus route, h ij represents the number of transfers required for residents to complete a trip from bus node i to j, h max Indicates the preset maximum number of transfers, Z ij represents the number of repeated routes between bus nodes i and j, Z max Indicates the maximum number of repeated routes between two preset bus nodes.

[0148] 6. Generate alternative bus routes

[0149] (1) Generation of alternative routes

[0150] The generation of alternative routes includes two parts: on the one hand, the shortest path algorithm needs to be used to solve the shortest path between a pair of starting and ending points; on the other hand, it is also necessary to take into account the disadvantage that the shortest path does not necessarily correspond to a greater travel demand. Therefore, the K shortest path algorithm needs to be used to expand the alternative routes.

[0151] 1) Shortest path algorithm

[0152] There are two main methods for shortest path calculation. Dijkstra's algorithm is applicable to the shortest path calculation with a determined starting point. Floyd's algorithm is applicable to the calculation of the shortest path between any two points. Considering that the setting of the starting point and the end point of a bus route is related to the space of large factories, airports, railway stations, campuses and parking lots, it can be determined by government departments such as urban planning, which is relatively fixed and it is impossible to set the starting and ending points at will, so it belongs to the shortest path calculation with a determined starting point. This embodiment adopts Dijkstra's algorithm, taking the point that can be used as the starting station of the bus route as the starting point, and calculating the shortest path between the starting and ending points.

[0153] like Figure 4 As shown in the figure, the nodes in the figure represent the intersections of the accessible streets in the traffic zone of the city. The line segment between two nodes represents a section of street or lane in the traffic zone. The line segment between nodes is artificially assigned a value of 0-q, which is called the weight of this road segment. For example, in a situation where the requirements are not high, q=3 is taken, that is, the comprehensive situation of the road segments involved in the entire city suitable for bus traffic is divided into 4 levels, 0 means impassable, 1 means passable, 2 means relatively passable, and 3 means recommended passable.

[0154] First, an auxiliary vector D is introduced, each component of which represents the length of the multiple shortest paths currently found from the starting point vi to a certain end point vj. If there is a section weight greater than 0 from vi to vj, the length and weight information of the section is recorded in D, otherwise the path is blocked and returned. Use the depth-first search algorithm until the determined end point is found, and then record all the information of the path in the set of alternative routes. Obviously, in the set of alternative routes, the path with a length of D[j] = Min{D|vi,vj∈V} is the shortest path with the shortest length starting from vi.

[0155] In the process of constructing bus routes, it is necessary not only to consider the length of the route, but also the amount of passenger flow. There may be a situation where the path in a place with large passenger flow is not the shortest or the passenger flow on the shortest path is small. Therefore, it is necessary to use the K shortest path algorithm to calculate the second shortest path to expand the range of the set of alternative routes, increase the solution space of feasible solutions, and avoid missing reasonable bus routes. The KSP algorithm includes the general KSP algorithm and the restricted acyclic KSP algorithm. According to the actual situation, this embodiment uses the deletion path algorithm in the general KSP algorithm to calculate the K shortest path.

[0156] (2) Determination of compliance of alternative routes

[0157] The Dijkstra algorithm is used to obtain a set of m shortest routes. The path deletion algorithm is used to obtain a set of the second shortest routes. The two sets are combined to form an initial set of candidate bus routes. Each route is then judged in turn using the constraints described above to obtain the final set of candidate routes.

[0158] Line mileage constraints: The mileage range of the specified line is 10 to 16 km;

[0159] Non-linear coefficient constraint of the line: set to 1~1.6.

[0160] After all the initial routes have been identified, the routes with the same starting and ending points are arranged in ascending order of mileage from short to long, and the four routes with the shortest mileage between each pair of starting and ending points are selected (if there are less than four routes, all routes are selected). The set of selected routes is the final set of candidate routes.

[0161] 7. Optimize and find the best bus route

[0162] (1) Required basic data

[0163] Before using the multi-objective genetic algorithm to solve the bus route planning model, the basic data is used as a basis to prepare for the algorithm solution. The specific data includes the following aspects:

[0164] 1) Resident travel OD matrix. It is derived from the big data mobile phone signaling data of telecommunications companies.

[0165] Most residents usually carry mobile phones when they travel. Using the big data run by telecommunications companies, the system can depict the residents' travel trajectories. For example, the movement trajectory of a resident's mobile phone during the morning working hours and the movement trajectory of the mobile phone after work in the afternoon on weekdays can be used to calculate the resident's residence and workplace with a high probability. Similarly, the sections of roads covered by residents' travel trajectories for picking up children, buying groceries, etc. can be calculated with a high probability. The sum of all residents' movement trajectories passing through a certain section of road will give the traffic volume of the section. The OD matrix of residents' travel can reflect the frequency of travel between different starting and ending points.

[0166] 2) Public transport data, including bus speed, operating costs, etc. Obtained from public transport operators.

[0167] (2) Chromosome encoding

[0168] The lines in the candidate line set are numbered, and the chromosome length is determined according to the number of lines included in the solution. The encoding is represented by a binary string. For example, assuming that there are 30 lines in the candidate line set, a 5-bit binary string is required for encoding (2 5 >30), assuming that the number of lines included in the solution is 3, then each chromosome consists of 3*5=15 binary digits.

[0169] Table 3 Bus route chromosome coding

[0170]

[0171] Taking Table 3 as an example, the encoding method of chromosomes in the bus route planning scheme in this embodiment is described.

[0172] (3) Population initialization

[0173] The initialization of the population is to give the initial solution of the population according to the coding rules. The standard genetic algorithm population initialization method generally adopts the random generation method.

[0174] (4) Individual evaluation

[0175] The judgment of the quality of the solution is determined by the value of its fitness function. The present invention directly uses the objective function in the previously established bus route planning model as the fitness function.

[0176] For this objective function, the minimum value is sought. The more convergent the objective function is, the higher the fitness is. When calculating the fitness, the OD volume (traffic volume) in the calculated OD matrix of residents' travel needs to be allocated to the urban road network. The present invention adopts the shortest path allocation method, and the specific steps are as follows:

[0177] Step 1, calculate the shortest path between each starting and ending point;

[0178] Step 2: Allocate the OD of each starting and ending point to the shortest path:

[0179] Step 3: Add up and get the total OD volume (traffic volume) on each road section.

[0180] (5) Select an operation

[0181] The selection method used in this embodiment is the best individual retention method based on the roulette wheel method. The roulette wheel selection method takes the overall excellence of the population as the starting point, and also takes into account the excellent condition of the individual itself. The survival probability of individuals with high fitness is certainly high, but individuals with low fitness still have a certain survival probability. This strategy ensures the diversity of the population.

[0182] The probability P of the individual corresponding to alternative path n being selected (retained) n The specific calculation formula is:

[0183]

[0184] Among them, f n Represents the objective function value of individual n in the population, and N is the total number of individuals selected in this round of iteration.

[0185] (6) Crossover operation

[0186] Crossover is the most important operation in genetic algorithms, which embodies the idea of ​​information exchange. The crossover probability in genetic algorithms is between 0.4 and 0.99, that is, 40% to 99% of the chromosomes in the parent population undergo crossover operations. The crossover probability can be appropriately adjusted according to the results of each operation in order to achieve better solution results. Crossover operations can generally be divided into two-point crossover and single-point crossover. The former is more complicated than the latter, because the former needs to determine two intersection points to exchange genetic information between the two points, while the latter only needs a single intersection point to exchange all genetic information after the point. If the intersection point of the single-point crossover is located at the front end of the gene sequence, a large number of parent genes will be exchanged, resulting in the destruction of the excellent individuals that have been formed in the parent generation. Therefore, this embodiment selects the two-point crossover method, and the specific operating steps are as follows:

[0187] Step 1: Pair the parent individuals in groups of two according to the random principle;

[0188] Step 2: Generate two intersection points with different positions in each group of paired individuals;

[0189] Step 3: Generate a random number in the interval [0,1]. If this random number is within the preset crossover probability range, perform a crossover operation, otherwise do not perform any operation.

[0190] (7) Mutation Operation

[0191] Chromosome mutation provides an opportunity for the generation of new individuals. For the selected individual, a gene on its chromosome is randomly changed to another allele with a certain probability. In binary coding, there are only two possibilities: 0 mutates to 1 and 1 mutates to 0. The frequency of mutation operation is not common, and the mutation probability is generally between 0.0001 and 0.1.

[0192] Embodiment 2

[0193] Based on the above method, this embodiment provides a visualization system for optimizing bus routes in smart city construction. The system can complete data cleaning, stop point extraction, traffic zone division, OD matrix generation, and generation of optimized routes for the collected original spatiotemporal trajectory data, and can be applied to the field of public transportation planning.

[0194] 1. System development environment:

[0195] Python 3.9

[0196] Hadoop 2.7

[0197] MATLAB R2018b

[0198] Baidu Map Open Platform

[0199] JDK 1.8, Java Development Kit 1.8

[0200] ArcGis 10.2, a Geographic Information System (GIS) platform

[0201] This system is mainly developed using ArcGis and MATLAB. ArcGis software is mainly used for regional partitioning, combining travel trajectory data with each partition, and other operations, while genetic algorithms are used on the MATLAB platform to generate bus routes. In addition, Java and Python programming are used based on the Hadoop framework to clean and extract trajectory data. Finally, the generated bus routes are visualized using the interface provided by the Baidu Map Open Platform.

[0202] 2. The system involves the following data during experimental simulation:

[0203] Mobile phone signaling data: The signaling data is obtained by random generation, with a total of 100,000 items. The Baidu map API is called to ensure that the randomly generated data is within the specified geographical range (such as Jingkou District).

[0204] Bus stop information data: obtained from Baidu Maps through crawler technology.

[0205] Map GIS data: downloaded from OpenStreetMap.

[0206] 3. System Framework

[0207] The overall framework of this program is as follows Figure 5As shown in the figure, the original signaling data is first cleaned and the stop points are extracted by the trajectory data processing module under the Hadoop framework, and then stored in the trajectory database. At the same time, in the geographic information processing module, the geographic data such as the bus station location data and the regional GIS data are processed by ArcGis to divide the traffic area, and the traffic area is combined with the trajectory data of the trajectory database to generate the OD matrix and store it in the traffic flow database. Then, in the route generation module, Matlab software is used to use the K shortest path algorithm to generate a backup route set, and then a genetic algorithm is used to combine the obtained traffic flow data to find the optimal solution in the backup route set. Finally, the optimal bus route is displayed in the visualization module.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may also be combined, the steps may be implemented in any order, and there are many other changes in different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A bus route optimization design method based on residents' travel big data, characterized in that: Includes steps: S1. Obtain mobile phone signaling data of residents in the target area; S2, based on mobile phone signaling data, obtain residents’ travel trajectory stop points; S3, based on Thiessen polygons, divide the target area into several traffic areas; S4, based on the residents’ travel trajectory stop points and traffic communities, obtain the residents’ travel OD matrix; S5. Construct a bus route planning model, including objective functions and constraints; S6. Generate several alternative bus routes based on the shortest path algorithm; S7. Based on the genetic algorithm and combined with the objective function and the OD matrix, the target bus route is obtained from the alternative routes.

2. The bus route optimization design method according to claim 1, characterized in that: Step S1 also includes data cleaning and data sorting of the acquired mobile phone signaling data, wherein the mobile phone signaling data includes user ID, base station latitude and longitude, arrival time at the base station, and user stay time; The data cleaning includes: filtering out the mobile phone signaling data occurring in the target area according to the coordinate range, and deleting the mobile phone signaling data recorded simultaneously by two different base stations; the data sorting includes: re-sorting the acquired mobile phone signaling data according to the user ID.

3. The bus route optimization design method according to claim 1, characterized in that: In step S2, the travel trajectory of each resident is obtained and the stop points are extracted according to the mobile phone signaling data within the set time period; The extraction of the stop points is based on the DBSCAN clustering algorithm, which traverses each travel trajectory to obtain the corresponding stop point information, including: Traverse the trajectory points in the trajectory TRAj. If the time interval between the trajectory points in the ε-neighborhood of the trajectory point Pi and the trajectory point Pi is less than △t, then mark the trajectory point Pi as a core point; Connect the core points that are directly accessible by density through density accessibility to form several clusters, which represent the stay areas; The latitude and longitude coordinates of the core points in the same cluster are averaged to obtain the latitude and longitude coordinates of the stay points in the stay area; The residence time of the stay point is calculated based on the maximum and minimum timestamps of the core points in the same cluster.

4. The bus route optimization design method according to claim 3, characterized in that: In order to improve the efficiency of the algorithm, before extracting the stop point information of each travel trajectory, each travel trajectory is divided into several sub-trajectories, and then the stop point information of each sub-trajectory is extracted; the segmentation steps of the travel trajectory include: Traverse the trajectory points in the trajectory TRAj: if the time interval between the trajectory point Pi and its previous trajectory point is greater than the set threshold, it indicates that there is data missing between the trajectory point Pi and its previous trajectory point, and the trajectory point Pi is deleted; if the speed of the trajectory point Pi relative to its previous trajectory point is greater than the threshold V1, it indicates that the trajectory point Pi is a data drift point, and the trajectory point Pi is deleted; if the speed of the trajectory point Pi relative to its previous trajectory point is greater than the threshold V2 and less than the threshold V1, the trajectory point Pi is added to the judgment queue; If the length of the judgment queue is greater than the threshold d, the trajectory formed by the current trajectory points in the judgment queue is taken as a sub-trajectory; the queue is cleared and the judgment segmentation is continued.

5. The bus route optimization design method according to claim 1, characterized in that: Step S3 includes: based on the current bus stops, dividing the target area into several traffic zones according to the Thiessen polygon principle; merging the traffic zones in combination with the street administrative divisions, population density and actual road network of the target area; the shape of the traffic zone is a non-annular convex shape with no isolated islands inside, there is no overlap between the traffic zones, and the travel ratio in each traffic zone does not exceed the set threshold.

6. The bus route optimization design method according to claim 1, characterized in that: Step S4 includes: Construct an NxN resident travel matrix OD and initialize its element values ​​to 0, where N is the number of divided traffic zones; Mapping the trajectory stop points obtained in step S2 to each traffic zone divided in step S2; The two stop points of the same travel trajectory with adjacent stop times are regarded as a pair of starting and ending points, the earlier one is the starting point and the later one is the end point. The PNPoly algorithm is used to determine the traffic zones where the starting and ending points are located. If the starting and ending points cannot be mapped to the traffic zones, the point is considered to be beyond the scope of the target area. For any starting and ending point pair, if the starting point and the end point belong to different traffic zones, the starting and ending point pair forms a travel process, and the OD matrix element value corresponding to the starting and ending points is increased by one; Traverse each travel trajectory and complete the update of the OD matrix element values.

7. The bus route optimization design method according to claim 1, characterized in that: In step S5, the objective function of the planning model is: in, The constraints are: Among them, T represents the travel time of residents, Q represents the rate of return of the bus operator, μ1 and μ2 are the weight coefficients of the objective function, which are used to balance the magnitude of the two parts of the objective function; N is the total number of bus lines in the region, n k represents the number of bus nodes passed by the k-th bus line, S ki represents the i-th bus node on the k-th bus route, d(S ki ,S ki+1 ) represents the distance between bus nodes i and i+1 on the kth bus line, v represents the evaluated running speed of the bus, t s is the stop time of the bus at each bus node, r ij represents the number of passengers between bus nodes i and j, e1 represents the fare, e2 represents the government subsidy amount per person for residents who choose to travel by bus, L ij represents the path distance between bus nodes i and j, G ij represents the number of buses passing between bus nodes i and j, c is the operating cost per unit mileage of each bus on the road, R′ represents the roads where buses are allowed to pass, R represents all roads within the research scope, and l min Indicates the minimum mileage of a single bus line, l max Indicates the maximum mileage of a single bus line, l k represents the mileage of the k-th bus line, d k represents the spatial straight-line distance between the starting and ending points of the kth line, η max Indicates the maximum value of the non-linear coefficient of the preset bus route, h ij represents the number of transfers required for residents to complete a trip from bus node i to j, h max Indicates the preset maximum number of transfers, Z ij represents the number of repeated routes between bus nodes i and j, Z max Indicates the maximum number of repeated routes between two preset bus nodes.

8. The bus route optimization design method according to claim 6, characterized in that: Step S6 includes: Use Dijkstra algorithm to solve the shortest path between any pair of starting and ending points and use it as an alternative bus route; The deletion path algorithm in the KSP algorithm is used to calculate the second shortest path to expand the range of the candidate route set; The candidate routes in the candidate route set are screened and sorted based on compliance.

9. The bus route optimization design method according to claim 1, characterized in that: In step S7, a genetic algorithm is used in combination with an objective function to optimize a target bus route from candidate bus routes, including the following steps: The routes in the candidate route set are numbered, and the chromosome length is determined according to the number of routes included in the target solution. The chromosome of the population individuals corresponding to each route is encoded using a binary string to complete the population initialization of the genetic algorithm; the genetic iteration is started, and the selection operation, crossover operation and mutation operation of the population individuals are performed in each iteration round until the set iteration round or the number of population individuals or the chromosome of the population individuals no longer changes is reached; the route corresponding to the population individuals retained after the iteration is completed is used as the final target bus route; In the selection operation, the fitness of the individuals in the population is obtained based on the objective function of the bus route planning model. The smaller the value of the objective function, the higher the fitness of the individuals in the population, and the greater the survival probability. The individuals in the population are selected and retained according to the survival probability. The survival probability P of the individual n in the population is n for: Among them, f n Represents the objective function value of population individual n, where N is the number of population individuals participating in this round of iterative selection; When calculating the objective function value, the traffic flow in the resident travel matrix OD needs to be distributed to the urban road network: the shortest path between each origin and destination in the OD matrix is ​​calculated, the traffic flow of each origin and destination is distributed to the shortest path, and the total traffic flow on each road section is obtained by superposition; In the crossover operation, two individuals of the population are first randomly selected to pair up as parent individuals, and then two crossover points with different positions are generated in each group of paired parent individuals. Then, a random number in the interval [0,1] is generated. If this random number falls within the preset crossover probability range, a crossover operation is performed to exchange the chromosome genes after the crossover point of the parent individual. In the mutation operation, a population individual is randomly selected with a certain probability, and a random gene on its chromosome is transformed into other alleles.

10. A bus route optimization design system based on the method according to any one of claims 1 to 9, characterized in that: include: Spatiotemporal big data storage module, used to store collected mobile phone signaling data; The stop point extraction module is used to clean the mobile phone signaling data and extract the track stop point information; the geographic information processing module is used to divide the target area into several traffic areas according to the GIS data; the OD matrix generation module is used to generate the OD matrix by combining the traffic areas and the track stop point information; the alternative route generation module is used to generate several alternative bus routes; the bus route optimization module is used to optimize the target bus route from the alternative bus routes; The visualization module is used to display the optimized target bus routes.

Images