A city intelligent public transport dispatching method and system based on internet of things

Abstract

The application discloses a kind of city intelligent public transport scheduling method and system based on Internet of Things, including the traffic influence coefficient of gathering city multi-source traffic information calculation, construct city public transport network time-varying diagram carries out flow propagation prediction each future state feature of public transport station, calculates the dynamic adjustment vehicle station time of vehicle head time interval, calculates the support demand index to determine the station to be supported, the support weighted demand index of corresponding neighborhood station of the station to be supported is selected to support vehicle.The method not only can improve the efficiency and accuracy of city public transport scheduling, but also has good interpretability, can be directly applied in city intelligent public transport scheduling system.

Description

Technical Field

[0001] This invention relates to the field of intelligent public transportation scheduling technology, and in particular to an urban intelligent bus scheduling method and system based on the Internet of Things. Background Technology

[0002] As a core component of the urban transportation system, urban public transportation undertakes the basic service function of large-scale resident travel. With the increasingly prominent problem of urban traffic congestion, prioritizing the development of public transportation has become an important strategic choice to alleviate urban traffic pressure. Against this background, improving the operational efficiency, service level and resource utilization of the public transportation system through intelligent means has become a research hotspot in the field of intelligent transportation.

[0003] However, existing bus dispatching technologies still have significant shortcomings: First, data from various modalities are scattered across different management departments, resulting in severe data silos and making it difficult to achieve a holistic understanding of the bus network's operational status. Second, traditional systems often make post-event adjustments based on current headway distances, lacking the ability to predict traffic flow evolution patterns. This leads to dispatching instructions often lagging behind demand changes, failing to eliminate systemic supply-demand imbalances. Third, existing methods primarily optimize timetables for single routes, neglecting the spatial interconnectivity of the bus network. In scenarios such as large-scale event dispersal or extreme weather, cross-route capacity coordination cannot be achieved, resulting in passenger congestion and low resource utilization at transfer nodes. Therefore, this invention proposes an IoT-based urban intelligent bus dispatching method and system. By constructing a time-varying structural bus network map, it achieves spatiotemporal coupling of multi-source information, introduces a physically constrained flow propagation mechanism for accurate prediction, and implements forward-looking balanced dispatching and cross-route coordination based on predicted headway distances and support demand indices. This has significant theoretical and practical value for improving the operational efficiency of the bus system and enhancing the passenger travel experience. Summary of the Invention

[0004] The purpose of this invention is to provide a smart urban bus dispatching method and system based on the Internet of Things.

[0005] To achieve the above objectives, the present invention is implemented according to the following technical solution: This invention includes the following steps: Collect multi-source traffic information in the city and calculate the traffic impact coefficient based on the multi-source traffic information. The state characteristics of each bus station are composed of urban multi-source traffic information and traffic impact coefficients. A time-varying graph of the urban public transport network is constructed with individual bus stations as graph nodes. Current urban multi-source traffic information is input to predict the future state characteristics of each bus station through traffic propagation. The headway of the bus is calculated based on the future state characteristics of each bus stop, and the vehicle parking time is dynamically adjusted according to the headway. The support demand index is calculated based on the future state characteristics and corresponding traffic impact coefficient of each bus station. The stations to be supported are determined, and the support weighted demand index of the neighboring stations corresponding to the stations to be supported is calculated to select support vehicles. The multi-source traffic information includes GPS vehicle flow, passenger flow data, meteorological data, information on major events, and social media sentiment; the passenger flow data includes passenger flow at bus stops and passenger flow on buses. The GPS traffic flow includes average vehicle speed, road congestion index, queue length at the entrance lane, and arrival time of each vehicle. The passenger flow data includes the number of people waiting on the platform, the number of people boarding the train on the platform, the number of people alighting on the platform, and the passenger load factor of vehicles arriving at downstream stations; The traffic impact coefficient includes the social feedback coefficient, the meteorological impact coefficient, and the event impact coefficient.

[0006] Furthermore, the method for calculating the traffic impact coefficient includes: Multi-source traffic information is collected from GPS vehicle flow, passenger flow, weather data, information on major events, and social media sentiment. The meteorological impact coefficient for each station's corresponding area is calculated based on meteorological data, and the expression is as follows: ; ;

[0007] in The meteorological impact coefficient. For comfort deviation, Punishment for rainfall For wind speed, For visibility, Outdoor temperature Relative humidity, For comfortable average temperature, The radius of the comfortable temperature range. Rainfall intensity; The event impact coefficient for each site is calculated based on information about large-scale events. The expression is as follows: ; ; ; in For large-scale events The impact coefficient of events at bus stops. Time-phase influencing factors Due to the scale of the event, For security level, Spatial distance attenuation factor, for Distance between bus stops and large events For the current event time, For the entire duration of the event, Near-field decay rate, Spatial threshold; To capture nearby social media sentiment and calculate the social feedback coefficient for each site's corresponding region, the expression is: ; in for Social feedback coefficient near the time station The time decay coefficient, For the intensity of emotion, For sentiment weight, The number of related terms. This serves as the historical average base for discussion of the corresponding road segment. For traffic surge rate, For smoothing parameters, As an indicator function, when the traffic surge rate Less than the traffic surge rate limit If the condition is normal, take 1; otherwise, take 0.

[0008] Furthermore, the method for predicting the future state characteristics of each bus stop includes: The GPS traffic flow, passenger flow data, traffic impact coefficient, and station location code of each bus stop are used to form the bus stop features. The set of all bus stops in the city is used as the graph nodes, and the corresponding bus stop features are associated as the node state features. The edge types are divided according to business logic, and the edge weights of each graph node edge are calculated based on the edge type and GPS traffic flow to construct a time-varying graph of the urban bus network. The edge types include physical connection edges within the same line and neighborhood association edges across lines. The edge weight expression is: ; ; in for time Stations and the same line The weight of the physical connection edges of the site. , The time value coefficient, for Site and The physical distance between the stations The average speed of vehicles on the road section between the two stations. The number of intersections on the road section between the two stations. The average intersection delay time, The status of the road section between the two stations. for time Stations and cross-line The weight of the neighboring edges of a site The walking decay rate, for Station to Historical passenger flow at the station The largest passenger transfer volume on the entire network. This is the transfer sensitivity coefficient; Aggregate the neighbor information of each graph node according to the edge type, as shown in the expression: ; ; in for time The corresponding nodes in the site map and their corresponding nodes in the neighboring area map are located at the same route. Layer aggregation information, for The set of nodes in the station map and the set of nodes in the line neighborhood map. for time The site graph node corresponds to the cross-line neighborhood graph node in Layer aggregation information, for Station graph node cross-line neighborhood graph node set, for Node state characteristics of nodes in a layer neighborhood graph. This is a learnable transformation matrix across the line edges; Given physical constraints, a gated graph convolutional network is used for multi-step traffic propagation to predict the future state characteristics of each bus stop. The expression is as follows: ; ; ; in for time layer The candidate state vector of the site, For the updated hidden vector, for The future state characteristics of the site To predict the time step, The candidate state weight matrix is... To reset the gating vector, for Layer The site's hidden vector, This is the bias for the candidate state. To update the gate vector, This is a physical constraint correction term. This is the output layer weight matrix. For output layer bias; The physical constraints include hard constraints on site capacity, flow conservation constraints, and lower limits on service frequency.

[0009] Furthermore, the method for dynamically adjusting vehicle parking time includes: Calculate the prediction time domain of two adjacent trains at the same station on the same line based on future state features. The headway within the vehicle is expressed as: ; in for Site Time-domain prediction The headway inside the car, For the previous site The actual headway at any given moment For the previous site The actual traffic speed at any given moment for Site Time-domain prediction Predicted traffic speed within the area , These represent the boarding rate and alighting rate for the corresponding routes. , These represent the number of people waiting at the station and the passenger capacity for the corresponding line. This represents the average time for each person to board and alight the bus. The predicted headway deviation is calculated based on the headway and ideal headway at the same station on the same route within the predicted time domain, and the vehicle parking time is adjusted accordingly.

[0010] Furthermore, the method for selecting support vehicles includes: The support demand index for each bus stop is calculated based on future state characteristics and corresponding traffic impact coefficients, expressed as follows: ; in for time Site support demand index , As a weight for congestion, Design the number of lines for each station. , , for time The number of passengers boarding and alighting at each stop on the bus route, and the passenger load factor. for The rated passenger capacity of the line vehicles, for The number of people waiting at the station at that time. The station's rated capacity, This is the time scaling factor. for time Train headway at the station As the weight of the traffic impact coefficient, for Traffic impact coefficient of the station Index for traffic impact coefficient types; Calculate the support demand index for all stations. Stations with a support demand index greater than the demand threshold are designated as stations awaiting support. The routes with the highest occupancy rates are designated as support routes. Calculate the weighted support demand index for the neighboring stations corresponding to the stations awaiting support. The expression is as follows: ; in for time The weighted demand index for support from neighboring sites, This is the distance attenuation coefficient. To support the site To neighboring sites distance, For line switching costs, for Site supports lines and neighboring sites The similarity of the route to the next arriving vehicle; Select the next arriving vehicle from the neighborhood station with the lowest weighted demand index as the support vehicle.

[0011] Secondly, an Internet of Things-based intelligent urban public transport dispatching system includes: Traffic Impact Module: Used to collect multi-source traffic information in the city and calculate traffic impact coefficients based on the multi-source traffic information. Traffic flow propagation prediction module: This module combines multi-source traffic information and traffic impact coefficients to form the state characteristics of each bus stop. It constructs a time-varying graph of the urban public transport network using individual bus stops as graph nodes. It inputs current multi-source traffic information to predict the future state characteristics of each bus stop. The station parking adjustment module is used to calculate the headway of the bus based on the future state characteristics of each bus stop, and dynamically adjust the vehicle parking time according to the headway. Support adjustment module: used to calculate the support demand index based on the future state characteristics and corresponding traffic impact coefficient of each bus station, determine the stations to be supported, calculate the support weighted demand index of the neighboring stations corresponding to the stations to be supported, and select support vehicles.

[0012] The beneficial effects of this invention are: This invention is an intelligent urban public transport dispatching method and system based on the Internet of Things. Compared with the prior art, this invention has the following technical advantages: This invention constructs a traffic impact coefficient by collecting multi-source traffic information in the city, breaking through the limitations of traditional single data sources. It realizes the deep spatiotemporal coupling and quantitative impact assessment of multi-source heterogeneous traffic data, providing a holographic situational awareness basis for scheduling decisions. This invention constructs a constrained time-varying graph neural network to ensure that the prediction results conform to the basic laws of public transportation operation, avoids the illusion problem of pure data-driven models, and significantly improves the accuracy of short- and medium-term passenger flow prediction within 45 minutes. This invention calculates the predicted headway based on the predicted future state characteristics, realizing proactive balanced scheduling from reactive response to proactive prevention, effectively eliminating common service quality problems such as train back-to-back and large intervals. This invention identifies network bottleneck nodes by calculating the support demand index of each station and selects support vehicles based on the weighted demand index of neighboring stations. It breaks the traditional fixed capacity barrier of a single route and realizes dynamic support across routes, improving the service resilience and resource utilization efficiency of the public transportation system under events such as large-scale events and extreme weather. This invention achieves local cleaning of raw data, edge inference of AI models, and real-time issuance of dispatch instructions through the distributed deployment of on-board OBU, roadside RSU, station RFID / visual sensors, and edge computing gateways. It forms a closed-loop IoT application with "end-edge-cloud" collaboration, meets the real-time requirement of millisecond-level response in bus dispatching, and has good engineering feasibility and scalability. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of an IoT-based intelligent urban bus dispatching method according to the present invention. Detailed Implementation

[0014] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0015] The present invention discloses an intelligent urban public transport dispatching method and system based on the Internet of Things, comprising the following steps: like Figure 1 As shown, this embodiment includes the following steps: Collect multi-source traffic information in the city and calculate the traffic impact coefficient based on the multi-source traffic information. The state characteristics of each bus station are composed of urban multi-source traffic information and traffic impact coefficients. A time-varying graph of the urban public transport network is constructed with individual bus stations as graph nodes. Current urban multi-source traffic information is input to predict the future state characteristics of each bus station through traffic propagation. The headway of the bus is calculated based on the future state characteristics of each bus stop, and the vehicle parking time is dynamically adjusted according to the headway. The support demand index is calculated based on the future state characteristics and corresponding traffic impact coefficient of each bus station. The stations to be supported are determined, and the support weighted demand index of the neighboring stations corresponding to the stations to be supported is calculated to select support vehicles. The multi-source traffic information includes GPS vehicle flow, passenger flow data, meteorological data, information on major events, and social media sentiment; the passenger flow data includes passenger flow at bus stops and passenger flow on buses. The GPS traffic flow includes average vehicle speed, road congestion index, queue length at the entrance lane, and arrival time of each vehicle. The passenger flow data includes the number of people waiting on the platform, the number of people boarding the train on the platform, the number of people alighting on the platform, and the passenger load factor of vehicles arriving at downstream stations; The traffic impact coefficient includes the social feedback coefficient, the meteorological impact coefficient, and the event impact coefficient.

[0016] In this embodiment, the method for calculating the traffic impact coefficient includes: Multi-source traffic information is collected from GPS vehicle flow, passenger flow, weather data, information on major events, and social media sentiment. The meteorological impact coefficient for each station's corresponding area is calculated based on meteorological data, and the expression is as follows: ; ;

[0017] in The meteorological impact coefficient. For comfort deviation, Punishment for rainfall For wind speed, For visibility, Outdoor temperature Relative humidity, For comfortable average temperature, The radius of the comfortable temperature range. Rainfall intensity; The event impact coefficient for each site is calculated based on information about large-scale events. The expression is as follows: ; ; ; in For large-scale events The impact coefficient of events at bus stops. Time-phase influencing factors Due to the scale of the event, For security level, Spatial distance attenuation factor, for Distance between bus stops and large events For the current event time, For the entire duration of the event, Near-field decay rate, Spatial threshold; To capture nearby social media sentiment and calculate the social feedback coefficient for each site's corresponding region, the expression is: ; in for Social feedback coefficient near the time station The time decay coefficient, For the intensity of emotion, For sentiment weight, The number of related terms. This serves as the historical average base for discussion of the corresponding road segment. For traffic surge rate, For smoothing parameters, As an indicator function, when the traffic surge rate Less than the traffic surge rate limit Take 1 if the condition is met, otherwise take 0; In actual assessment, the collection of GPS traffic flow data includes: collecting bus location through the bus on-board OBU (which includes a GPS positioning chip and a 4G communication module), collecting vehicle CAN bus data to identify buses that have not stopped at the station or are idling in congestion, and collecting and reporting road segment ID, average speed, and queue length through the roadside RSU (millimeter-wave radar, video capture camera, and 5G communication module). Passenger flow data collection includes: collecting passenger flow data at bus stops through binocular stereo vision cameras (accessed via edge gateway using lightweight YOLOv5n algorithm to identify passenger flow within the station), infrared beam sensors (to assist in counting passenger flow within the station) (via 5G communication module), and collecting the number of passengers getting on and off the bus through TOF laser sensors on the bus doors; Meteorological data collection includes: collecting meteorological data (temperature, humidity, rainfall intensity, wind speed, and visibility) through miniature weather stations at bus stops (with built-in NB-IoT modules) and reporting it to the cloud as a meteorological baseline value; combining the meteorological baseline value with roadside meteorological sensing piles on key road sections (accessed via LoRaWAN to the nearest gateway) and miniature meteorological sensors on the roof of buses to obtain meteorological data; when calculating the meteorological influence coefficient, the temperature comfort range is [18,26]℃, the average temperature comfort value is 22, the radius of the temperature comfort range is 4, and the comprehensive value is compressed to the (1,2) range through the tanh function; The collection of information on large-scale events includes: the regional dispatch center securely accessing the government event registration system and ticketing software API through firewalls to obtain event lists; collecting event status data through electronic fences / WIFI probes near important venues / buildings (a surge in crowd density and a surge in mobile phone signal detection frequency correspond to entry, a high difference between crowd density and mobile phone signal strength corresponds to the event being underway, and a surge in crowd density and a sharp drop in mobile phone signal detection frequency correspond to the event being over); obtaining traffic control instructions for roads surrounding the event (such as temporary cancellation of bus lanes or closure of certain road sections) through the traffic police signal IoT platform as constraints for calculating the event impact coefficient; when calculating the event impact coefficient, the spatial distance attenuation factor is calculated in stages according to the event's progress status, and the security level is determined by the intensity of the event security control over traffic on surrounding roads; The collection of social media sentiment data includes: collecting traffic information from map software traffic APIs (collecting road congestion indexes and user-reported events), social software geofence APIs (collecting tweets with geographic tags within a 500-meter radius of each station), and passenger feedback from buses (passengers collect social media sentiment data via phone, questionnaires, or pre-set mini-programs). The data is then processed using a lightweight BERT model to analyze the sentiment polarity and obtain the social media sentiment data. Calculate the social feedback coefficient, meteorological impact coefficient, and event impact coefficient using the formulas described above.

[0018] In this embodiment, the method for predicting the future state characteristics of each bus stop includes: The GPS traffic flow, passenger flow data, traffic impact coefficient, and station location code of each bus stop are used to form the bus stop features. The set of all bus stops in the city is used as the graph nodes, and the corresponding bus stop features are associated as the node state features. The edge types are divided according to business logic, and the edge weights of each graph node edge are calculated based on the edge type and GPS traffic flow to construct a time-varying graph of the urban bus network. The edge types include physical connection edges within the same line and neighborhood association edges across lines. The edge weight expression is: ; ; in for time Stations and the same line The weight of the physical connection edges of the site. , The time value coefficient, for Site and The physical distance between the stations The average speed of vehicles on the road section between the two stations. The number of intersections on the road section between the two stations. The average intersection delay time, The status of the road section between the two stations. for time Stations and cross-line The weight of the neighboring edges of a site The walking decay rate, for Station to Historical passenger flow at the station The largest passenger transfer volume on the entire network. This is the transfer sensitivity coefficient; Aggregate the neighbor information of each graph node according to the edge type, as shown in the expression: ; ; in for time The corresponding nodes in the site map and their corresponding nodes in the neighboring area map are located at the same route. Layer aggregation information, for The set of nodes in the station map and the set of nodes in the line neighborhood map. for time The site graph node corresponds to the cross-line neighborhood graph node in Layer aggregation information, for Station graph node cross-line neighborhood graph node set, for Node state characteristics of nodes in a layer neighborhood graph. This is a learnable transformation matrix across the line edges; Given physical constraints, a gated graph convolutional network is used for multi-step traffic propagation to predict the future state characteristics of each bus stop. The expression is as follows: ; ; ; in for time layer The candidate state vector of the site, For the updated hidden vector, for The future state characteristics of the site To predict the time step, The candidate state weight matrix is... To reset the gating vector, for Layer The site's hidden vector, This is the bias for the candidate state. To update the gate vector, This is a physical constraint correction term. This is the output layer weight matrix. For output layer bias; The physical constraints include hard constraints on site capacity, traffic conservation constraints, and lower limits on service frequency. In actual evaluation, graph nodes and edges are divided into two categories according to business logic. Two stations belonging to the same bus route and physically adjacent stations have a physical connection edge on the same route (a directed edge, representing the physical travel path of the bus; a station will only contain two edges, corresponding to the previous station and the next station respectively). Two stations belonging to different routes and with a spatial Euclidean distance of less than 500m have a cross-line neighborhood association edge (an undirected edge, representing the relationship between passenger transfer choices and cross-line capacity coordination; a station can contain multiple edges). When calculating the weight of the physical connection edge on the same route, the average intersection delay is calibrated based on historical traffic flow data, and the road segment status between the two stations is set to 1 during operation and 0 otherwise. The larger the weight of the physical connection edge, the easier the traffic propagation. When calculating the weight of neighboring related edges across lines, historical transfer passenger flow is directly statistically analyzed based on bus IC card data, and the transfer sensitivity coefficient reflects that passengers within short distances tend to choose station combinations with high transfer passenger flow. When any station in any related edge is within 500 of a large event, the weight of the edge is adjusted in real time by increasing the historical transfer passenger flow. Determine physical constraints, including hard constraints on station capacity (the number of people waiting on the platform should meet the physical capacity of the platform), flow conservation constraints (flow conservation between platform transport flow, number of people waiting and number of people boarding), and lower limit constraints on service frequency (forced correction when the predicted headway is too short (exceeding the capacity limit). Reset the gating vector during multi-step traffic propagation. and updating the gating vector The expression used to control the fusion ratio of historical state and neighbor information is: ; ; in To update the gating weight matrix, To update the gating bias. To reset the gating weight matrix, To reset the gating bias.

[0019] In this embodiment, the method for dynamically adjusting vehicle parking time includes: Calculate the prediction time domain of two adjacent trains at the same station on the same line based on future state features. The headway within the vehicle is expressed as: ; in for Site Time-domain prediction The headway inside the car, For the previous site The actual headway at any given moment For the previous site The actual traffic speed at any given moment for Site Time-domain prediction Predicted traffic speed within the area , These represent the boarding rate and alighting rate for the corresponding routes. , These represent the number of people waiting at the station and the passenger capacity for the corresponding line. This represents the average time for each person to board and alight the bus. Calculate the predicted headway deviation based on the headway and ideal headway at the same station on the same route in the predicted time domain, and adjust the vehicle parking time accordingly. In actual assessment, the headway deviation at the current moment is first calculated. If the headway deviation at the current moment is outside the headway deviation range and less than 0, the vehicle's parking time at the next station is extended. If the headway deviation at the current moment is outside the headway deviation range and greater than 0, the vehicle's parking time at the next station is shortened. Predicted headway deviation is mainly used for advance scheduling of traffic flow during large-scale events or morning and evening peak hours. It adjusts the ideal headway in different time periods (5 minutes during peak hours and 10 minutes during off-peak hours), calculates the predicted headway deviation for each station, and pre-sets the stationing time of vehicles on each line at the corresponding station to ensure smooth traffic flow.

[0020] In this embodiment, the method for selecting support vehicles includes: The support demand index for each bus stop is calculated based on future state characteristics and corresponding traffic impact coefficients, expressed as follows: ; in for time Site support demand index , As a weight for congestion, Design the number of lines for each station. , , for time The number of passengers boarding and alighting at each stop on the bus route, and the passenger load factor. for The rated passenger capacity of the line vehicles, for The number of people waiting at the station at that time. The station's rated capacity, This is the time scaling factor. for time Train headway at the station As the weight of the traffic impact coefficient, for Traffic impact coefficient of the station Index for traffic impact coefficient types; Calculate the support demand index for all stations. Stations with a support demand index greater than the demand threshold are designated as stations awaiting support. The routes with the highest occupancy rates are designated as support routes. Calculate the weighted support demand index for the neighboring stations corresponding to the stations awaiting support. The expression is as follows: ; in for time The weighted demand index for support from neighboring sites, This is the distance attenuation coefficient. To support the site To neighboring sites distance, For line switching costs, for Site supports lines and neighboring sites The similarity of the route to the next arriving vehicle; Select the next arriving vehicle from the neighborhood station with the lowest weighted demand index as the support vehicle. In the actual assessment, the A Stadium bus stop (stop number HZ-201) in XX City was taken as the object. This stop is adjacent to the A Stadium. At present (21:30), the Z Star Concert is in progress, so intelligent bus dispatching is implemented. Current large-scale event information: Z Star Concert, total event duration 180 minutes (19:00-22:00), current event duration 150 minutes (event in progress), event scale 50,000 people, security level 4 (high-level security), this site is 300 meters away from the event location, spatial threshold 500m; The meteorological impact coefficient, event impact coefficient, and social feedback coefficient were calculated to be 1.15, 2.35, and 1.68 respectively based on multi-source traffic information. Based on multi-source traffic information, the time-varying graph node features of Stadium A are constructed. The previous station A1 and the next station A2 of the same route (Route 21) are associated to form physical connection edges within the same route. The stations L1, L2, and L3 within a 500-meter range are associated to form cross-route neighborhood association edges. The edge weights are calculated based on multi-source traffic information. A three-layer flow propagation method is used through a gated graph convolutional network. Considering the hard constraint of station capacity (platform capacity 150 people) and the flow conservation constraint, the state characteristics predicted 30 minutes later (22:00, the end of the concert) are as follows (predicted number of people waiting is 135, predicted boarding rate is 0.95, and predicted road speed is km / h). The current headway of bus route 21 is calculated to be 5.625 min. During the concert, the ideal headway is 4 min, resulting in a headway deviation of 1.625 min, which is outside the deviation threshold range [1-1]. Therefore, the vehicle's parking time at the next stop, A2, is shortened by 30 seconds. Based on future state characteristics, the headway of bus route 21 after 30 minutes is calculated to be 10.16 min. During the concert's end, the ideal headway is adjusted to 3 min, predicting a headway deviation of 7.16 min, which is outside the deviation threshold range [1-1] and is considered serious. Therefore, the current parking time at the A Stadium bus stop is extended to 90 seconds (to prepare for waiting passengers in advance, while also considering the large number of people and the long boarding time, as well as the presence of mobile people around the bus before departure). The congestion weight is set to 0.6 / 0.4, the time scaling factor to 0.5, the traffic impact factor weight to 0.2 / 0.5 / 0.3, the distance attenuation factor to 0.5, and the line exchange cost to 1.2. When the concert ends (22:00), the support demand index is calculated to be 2.81 based on the status characteristics of the A Stadium bus station and the corresponding traffic impact factor, which exceeds the demand threshold of 2. The station is set as a station to be supported. Route 21, which has the highest passenger load among the three routes of the station, is selected as the support route. The support weighted demand index of the neighboring stations (L1 station / next arrival vehicle corresponding to route 15, L2 station / next arrival vehicle corresponding to route 37, L3 station / next arrival vehicle corresponding to route 89) is calculated to be 1.53 (route similarity 0.65), 1.75 (route similarity 0.72), and 2.02 (route similarity 0.58). Route 15, the next arrival vehicle of L1 station, is selected as the support vehicle to support the route 21 of the A Stadium bus station.

[0021] Secondly, an Internet of Things-based intelligent urban public transport dispatching system includes: Traffic Impact Module: Used to collect multi-source traffic information in the city and calculate traffic impact coefficients based on the multi-source traffic information. Traffic flow propagation prediction module: This module combines multi-source traffic information and traffic impact coefficients to form the state characteristics of each bus stop. It constructs a time-varying graph of the urban public transport network using individual bus stops as graph nodes. It inputs current multi-source traffic information to predict the future state characteristics of each bus stop. The station parking adjustment module is used to calculate the headway of the bus based on the future state characteristics of each bus stop, and dynamically adjust the vehicle parking time according to the headway. Support adjustment module: used to calculate the support demand index based on the future state characteristics and corresponding traffic impact coefficient of each bus station, determine the stations to be supported, calculate the support weighted demand index of the neighboring stations corresponding to the stations to be supported, and select support vehicles.

[0022] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart urban bus dispatching method based on the Internet of Things, characterized in that, Includes the following steps: S1. Collect multi-source traffic information in the city and calculate the traffic impact coefficient based on the multi-source traffic information in the city; S2. The state characteristics of each bus station are composed of urban multi-source traffic information and traffic impact coefficients. A time-varying graph of the urban bus network is constructed with a single bus station as a graph node. The current urban multi-source traffic information is input to predict the future state characteristics of each bus station by traffic propagation. S3. Calculate the headway from each bus stop based on its future state characteristics, and dynamically adjust the vehicle's parking time according to the headway. S4. Calculate the support demand index based on the future state characteristics and corresponding traffic impact coefficient of each bus station, determine the stations to be supported, calculate the support weighted demand index of the neighboring stations corresponding to the stations to be supported, and select support vehicles. The multi-source traffic information includes GPS vehicle flow, passenger flow data, meteorological data, information on major events, and social media sentiment; the passenger flow data includes passenger flow at bus stops and passenger flow on buses. The GPS traffic flow includes average vehicle speed, road congestion index, queue length at the entrance lane, and arrival time of each vehicle. The passenger flow data includes the number of people waiting on the platform, the number of people boarding the train on the platform, the number of people alighting on the platform, and the passenger load factor of vehicles arriving at downstream stations; The traffic impact coefficient includes the social feedback coefficient, the meteorological impact coefficient, and the event impact coefficient.

2. The urban intelligent bus dispatching method based on the Internet of Things according to claim 1, characterized in that, The method for calculating the traffic impact coefficient includes: Multi-source traffic information is collected from GPS vehicle flow, passenger flow, weather data, information on major events, and social media sentiment. The meteorological impact coefficient for each station's corresponding area is calculated based on meteorological data, and the expression is as follows: ； ； ; in This is the meteorological impact coefficient. For comfort deviation, Punishment for rainfall For wind speed, For visibility, Outdoor temperature Relative humidity, For comfortable average temperature, The radius of the comfortable temperature range. Rainfall intensity; The event impact coefficient for each site is calculated based on information about large-scale events. The expression is as follows: ； ； ； in For large-scale events The impact coefficient of events at bus stops. Time-phase influencing factors Due to the scale of the event, For security level, Spatial distance attenuation factor, for Distance between bus stops and large events For the current event time, For the entire duration of the event, Near-field decay rate, Spatial threshold; To capture nearby social media sentiment and calculate the social feedback coefficient for each site's corresponding region, the expression is: ； in for Social feedback coefficient near the time station The time decay coefficient, For the intensity of emotion, For sentiment weight, The number of related terms. This serves as the historical average base for discussion of the corresponding road segment. For traffic surge rate, For smoothing parameters, As an indicator function, when the traffic surge rate Less than the traffic surge rate limit If the condition is normal, take 1; otherwise, take 0.

3. The urban intelligent bus dispatching method based on the Internet of Things according to claim 1, characterized in that, The method for predicting the future state characteristics of each bus stop includes: The GPS traffic flow, passenger flow data, traffic impact coefficient, and station location code of each bus stop are used to form the bus stop features. The set of all bus stops in the city is used as the graph nodes, and the corresponding bus stop features are associated as the node state features. The edge types are divided according to business logic, and the edge weights of each graph node edge are calculated based on the edge type and GPS traffic flow to construct a time-varying graph of the urban bus network. The edge types include physical connection edges within the same line and neighborhood association edges across lines. The edge weight expression is: ； ； in for time Stations and the same line The physical connection edge weights of the site. , The time value coefficient, for Site and The physical distance of the stations The average speed of vehicles on the road section between the two stations. The number of intersections on the road section between the two stations. The average intersection delay time, The status of the road section between the two stations. for time Stations and cross-line The weight of the neighboring edges of a site The walking decay rate, for Station to Historical passenger flow at the station The largest passenger transfer volume on the entire network. This is the transfer sensitivity coefficient; Aggregate the neighbor information of each graph node according to the edge type, as shown in the expression: ； ； in for time The corresponding nodes in the site map and their corresponding nodes in the neighborhood map are located at the same route. Layer aggregation information, for The set of nodes in the station map and the set of nodes in the line's neighborhood map. for time The site map node corresponds to the cross-line neighborhood map node in Layer aggregation information, for Station graph node cross-line neighborhood graph node set, for Node state characteristics of nodes in a layer neighborhood graph. This is a learnable transformation matrix across the line edges; Given physical constraints, a gated graph convolutional network is used for multi-step traffic propagation to predict the future state characteristics of each bus stop. The expression is as follows: ； ； ； in for time layer The candidate state vector of the site, For the updated hidden vector, for The future state characteristics of the site To predict the time step, The candidate state weight matrix is... To reset the gate vector, for Layer The site's hidden vector, This is the bias for the candidate state. To update the gate vector, This is a physical constraint correction term. This is the output layer weight matrix. For output layer bias; The physical constraints include hard constraints on site capacity, flow conservation constraints, and lower limits on service frequency.

4. The urban intelligent bus dispatching method based on the Internet of Things according to claim 1, characterized in that, The method for dynamically adjusting vehicle parking time includes: Calculate the prediction time domain of two adjacent trains at the same station on the same line based on future state features. The headway within the vehicle is expressed as: ； in for Site Time-domain prediction The headway inside the car, For the previous site The actual headway at any given moment For the previous site The actual traffic speed at any given moment for Site When predicting time Predicted traffic speed within the area , These represent the boarding rate and alighting rate for the corresponding routes. , These represent the number of people waiting at the station and the passenger capacity for the corresponding line. This represents the average time for each person to board and alight the bus. The predicted headway deviation is calculated based on the headway and ideal headway at the same station on the same route in the predicted time domain, and the vehicle parking time is adjusted accordingly.

5. The urban intelligent bus dispatching method based on the Internet of Things according to claim 1, characterized in that, The method for selecting support vehicles includes: The support demand index for each bus stop is calculated based on future state characteristics and corresponding traffic impact coefficients, expressed as follows: ； in for time Site support demand index , As a weight for congestion, Design the number of lines for each station. , , for time The number of passengers boarding and alighting at each stop on the bus route, and the passenger load factor. for The rated passenger capacity of the line vehicles, for The number of people waiting at the station at any given time. The station's rated capacity, This is the time scaling factor. for time Train headway at the station As the weight of the traffic impact coefficient, for Traffic impact coefficient of the station Index for traffic impact coefficient types; Calculate the support demand index for all stations. Stations with a support demand index greater than the demand threshold are designated as stations awaiting support. The routes with the highest occupancy rates are designated as support routes. Calculate the weighted support demand index for the neighboring stations corresponding to the stations awaiting support. The expression is as follows: ； in for time The weighted demand index for support from neighboring sites, This is the distance attenuation coefficient. To support the site To neighboring sites distance, For line switching costs, for Site supports lines and neighboring sites The similarity of the route to the next arriving vehicle; Select the next arriving vehicle from the neighborhood station with the lowest weighted demand index as the support vehicle.

6. An Internet of Things-based intelligent urban public transport dispatching system, used to perform the method according to any one of claims 1-5, characterized in that, include: Traffic Impact Module: Used to collect multi-source traffic information in the city and calculate traffic impact coefficients based on the multi-source traffic information. Traffic flow propagation prediction module: This module combines multi-source traffic information and traffic impact coefficients to form the state characteristics of each bus stop. It constructs a time-varying graph of the urban public transport network using individual bus stops as graph nodes. It inputs current multi-source traffic information to predict the future state characteristics of each bus stop. The station parking adjustment module is used to calculate the headway of the bus based on the future state characteristics of each bus stop, and dynamically adjust the vehicle parking time according to the headway. Support adjustment module: used to calculate the support demand index based on the future state characteristics and corresponding traffic impact coefficient of each bus station, determine the stations to be supported, calculate the support weighted demand index of the neighboring stations corresponding to the stations to be supported, and select support vehicles.

Images