School bus arrival time estimation method and system

Abstract

The application relates to the technical field of urban public transport, and discloses a school bus arrival time estimation method and system. The method comprises the following steps: acquiring a current latitude and longitude coordinate and a real-time traffic congestion index of a road section during school bus driving; performing driving route time sequence matching on the current latitude and longitude coordinate to obtain a current driving road section; performing dynamic correction according to the real-time traffic congestion index to obtain a real-time driving speed; calculating a real-time section travel time according to the real-time driving speed and a remaining distance and performing conduction effect analysis to obtain a cumulative effect value; calculating a subsequent station time offset according to the cumulative effect value to generate a final arrival time. The method can realize dynamic adjustment and delay accumulation control of the arrival time, and improve the estimation accuracy under complex road conditions.

Description

Technical Field

[0001] This invention relates to the field of urban public transportation technology, and in particular to a method and system for estimating school bus arrival times. Background Technology

[0002] Currently, in the field of urban public transportation technology, accurate prediction of school bus arrival times is a key aspect of optimizing students' commuting experience and improving overall operational scheduling efficiency, and has become a major research focus in the field.

[0003] In one existing technology, a one-time estimation model based on historical data or static routes is mainly used to determine the arrival time. This relies heavily on fixed scheduling rules or single time indicators for evaluation, such as using a pre-set route to perform time estimation tasks and judging the expected arrival status of the vehicle by simple historical data on time consumption. However, this approach is clearly insufficient in complex urban road traffic environments. The one-time estimation model cannot respond to dynamic changes in the road environment. Single time assessments are easily affected by sudden congestion and traffic accidents during the morning rush hour, and are affected by drastic fluctuations in real-time road conditions and changes in travel time. This makes it difficult to guarantee the accuracy of the estimation during the actual operation of the school bus. Furthermore, it lacks the ability to perceive changes in road conditions ahead and dynamically correct delays through cascading transmission. Especially in scenarios with frequent sudden congestion and cumulative time effects from station delays, it is difficult to avoid serious time deviations caused by static estimations, rendering the original station arrival times meaningless.

[0004] In summary, existing technologies are insufficient to dynamically adjust arrival times and manage accumulated delays during school bus journeys, failing to meet the core requirement of accurate time estimation for urban public transportation in the face of dynamic road condition changes. Summary of the Invention

[0005] This invention provides a method and system for estimating school bus arrival times, enabling dynamic adjustment of arrival times and cumulative delay control during school bus operation, thus meeting the core requirement of accurate time estimation for urban public transportation in the face of dynamic road condition changes.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a method for estimating school bus arrival time, comprising: Obtain the current latitude and longitude coordinates of the school bus during its journey, and perform time-series matching processing on the current latitude and longitude coordinates to obtain the current route segment; Obtain the real-time traffic congestion index corresponding to the current driving segment, and perform dynamic correction processing based on the real-time traffic congestion index to obtain the real-time driving speed; The real-time segment travel time is calculated by performing time extrapolation based on the real-time driving speed and the remaining distance of the current travel segment, and the cumulative effect value is obtained by performing transmission effect analysis based on the real-time segment travel time. Calculate the time offset of the cumulative effect value to subsequent stations; if the stop buffer margin corresponding to the time offset is less than the minimum buffer threshold, perform smoothing correction processing on the time offset to obtain the corrected arrival time, and perform alignment processing on the corrected arrival time using the school bus departure interval to generate the final arrival time; if the stop buffer margin is not less than the minimum buffer threshold, directly generate the final arrival time based on the time offset.

[0007] Secondly, the present invention provides a school bus arrival time estimation system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the school bus arrival time estimation method described above.

[0008] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the school bus arrival time estimation method described above.

[0009] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains the current location coordinates of the school bus and dynamic road condition data, compares the road topology network to determine the congestion status of the road segment ahead, integrates historical traffic data and real-time traffic density to predict the current driving speed of the road segment, and determines the sudden change in road condition when the speed is lower than the threshold and extracts the congestion length to calculate the first delay time. It breaks through the limitation of the traditional one-time prediction mode before departure, which cannot respond to the dynamic changes in the road environment. It integrates real-time road condition perception and machine learning prediction algorithms to dynamically capture the sudden congestion status during the driving process, accurately quantifies the actual time delay caused by single-point congestion, and solves the problem that static route prediction is easily affected by sudden congestion and traffic accidents during the morning rush hour, and the deviation between the predicted time and the actual arrival time is too large. It provides an accurate local delay data basis for subsequent station time correction.

[0010] (2) This invention calculates the travel time from the current road segment to the next station by the first delay time and the predicted travel speed to determine the first arrival time. It compares the original arrival time of the schedule to obtain the single-station time delay and integrates the historical operation data of subsequent stations with the distance matrix between adjacent stations to determine the delay transmission amount. When the transmission amount exceeds the threshold, the cumulative effect value is calculated and the attenuation weight is extracted. This invention breaks through the limitation of traditional prediction methods that lack the ability to quantify the transmission effect of single-station delay to subsequent stations. Starting from the single-point delay, it dynamically calculates the transmission and attenuation process of the delay on the entire line by combining the spatial distance between stations and historical travel patterns. It accurately depicts the time accumulation effect of congestion delay and makes up for the defects of existing technologies that cannot respond to the chain transmission of delays at multiple stations, resulting in the overall inaccuracy of the predicted time of subsequent stations.

[0011] (3) This invention combines the cumulative effect value to perform time offset calculation on the original arrival time of subsequent stations to obtain the stop buffer margin. When the buffer margin is insufficient, it performs smooth correction and school bus departure interval alignment processing to generate a dynamic estimated arrival time. It constructs a closed-loop dynamic estimation system from real-time traffic condition perception, single-point delay calculation, multi-station transmission quantification to global time correction. It breaks through the limitation of traditional school bus arrival time estimation methods that lack dynamic correction ability during driving. It uses real-time traffic condition changes as the driving force to perform linkage correction on the arrival time of each station on the entire line. It takes into account the accurate extraction of single-station delay and the reasonable attenuation of multi-station cumulative effect, effectively improving the accuracy and reliability of school bus arrival time estimation in complex dynamic traffic environment, and meeting the core needs of school bus operation for dynamic adjustment of arrival time and delay accumulation control. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of a school bus arrival time estimation method provided by the present invention; Figure 2 This is a schematic diagram of a real-time dynamic correction of driving speed provided by the present invention; Figure 3 This is a schematic diagram of the time offset of subsequent stations and the final arrival time provided by the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Reference Figure 1 The first embodiment of the present invention provides a method for estimating the arrival time of a school bus, comprising the following steps: S11, obtain the current latitude and longitude coordinates of the school bus during its journey, and perform time-series matching processing on the current latitude and longitude coordinates to obtain the current route segment; S12, obtain the real-time traffic congestion index corresponding to the current driving segment, and perform dynamic correction processing based on the real-time traffic congestion index to obtain the real-time driving speed; S13. Based on the real-time driving speed and the remaining distance of the current driving segment, perform time extrapolation calculation to obtain the real-time segment travel time, and perform transmission effect analysis processing based on the real-time segment travel time to obtain the cumulative effect value; S14, calculate the time offset of the cumulative effect value to subsequent stations; if the stop buffer margin corresponding to the time offset is less than the minimum buffer threshold, perform smoothing correction processing on the time offset to obtain the corrected arrival time, and perform alignment processing on the corrected arrival time using the school bus departure interval to generate the final arrival time; if the stop buffer margin is not less than the minimum buffer threshold, directly generate the final arrival time based on the time offset.

[0015] In step S11, obtaining the current latitude and longitude coordinates during the school bus's journey and performing time-series matching processing on the current latitude and longitude coordinates to obtain the current travel segment includes: The real-time location trajectory of the school bus during its journey is collected periodically, and the current latitude and longitude coordinates and the current timestamp are extracted from the real-time location trajectory. The current latitude and longitude coordinates are input into a preset electronic map system to perform projection mapping, thereby obtaining the projected coordinate points; Match the corresponding preset school bus route based on the current timestamp; Calculate the matching distance between the projected coordinate points and the center lines of each segment in the preset driving route; The matching distance is compared with a distance determination threshold. If the distance is less than the distance determination threshold, the school bus is determined to be in the corresponding candidate road segment. If the distance is not less than the distance determination threshold, the school bus is determined to be in a yaw state. Perform temporal connectivity checks on each candidate road segment to determine the current travel segment.

[0016] It should be noted that the school bus's onboard terminal has a built-in global navigation satellite signal receiving module. This module receives satellite positioning signals every second and outputs a raw positioning statement in NMEA 0183 format as the school bus's real-time positioning trajectory during operation. This statement includes fields such as UTC time, latitude, longitude, ground speed, and heading angle. The latitude and longitude values ​​are parsed from the raw positioning statement to obtain the current latitude and longitude coordinates. The latitude is rounded to six decimal places, and the longitude is also rounded to six decimal places. Simultaneously, the UTC time is parsed and converted to the current timestamp in the local time zone, accurate to the second.

[0017] Next, the current latitude and longitude coordinates are input into a preset electronic map system to perform projection mapping, obtaining projected coordinate points. The preset electronic map system is built using an open-source geographic information platform and internally stores road topology network data for the area where the school bus route is located. The road topology network uses road segments as basic units, and each road segment consists of a start node, an end node, a centerline geometric string, and attribute information. The centerline geometric string is stored as a latitude and longitude sequence in the WGS-84 coordinate system. The projection mapping uses a Hidden Markov Model (HMM) as the map matching algorithm. This algorithm takes the current latitude and longitude coordinates as the input observation state sequence and uses the road segments in the road topology network as the hidden state candidate set. The state transition probability of the HMM is calculated based on the shortest path distance between the previous matching point and the current candidate point. The shorter the shortest path distance, the higher the transition probability. Paths that are closer to a straight line connection and have no detours are preferred. The emission probability is calculated based on the vertical projection distance from the current latitude and longitude coordinates to the centerline of the candidate road segment. The closer the vertical projection distance, the higher the emission probability. The projection point on the candidate road segment with the largest product of state transition probability and emission probability is taken as the projection coordinate point. The projection coordinate point is represented by the longitude and latitude after projection, and the road segment identification code where the projection coordinate point is located is recorded.

[0018] Next, the corresponding preset school bus route is matched based on the current timestamp. The preset school bus routes are stored in the local database of the onboard terminal. Each preset route consists of four fields: route number, operating time period, driving direction, and ordered road segment sequence. The ordered road segment sequence is a list of road segment identifiers arranged sequentially by driving direction. Operating times are divided into three periods: morning shift (6:30 AM to 9:00 AM), afternoon shift (11:30 AM to 2:00 PM), and evening shift (3:30 PM to 6:30 PM). During matching, the operating time period is determined by the hour and minute of the current timestamp, and the direction parameter is determined by the driving direction manually selected after vehicle ignition. Records matching both the operating time period and driving direction are retrieved from the local database, and the ordered road segment sequence from these records is extracted as the preset route for this trip. If multiple matching records are found, the one with the smallest route number is selected as the default route. If no matching record is found, the system is considered to be in a non-operating state, and subsequent processing is paused.

[0019] Subsequently, the matching distance between the projected coordinate points and the centerlines of each segment in the preset driving route is calculated. The matching distance between the projected coordinate points and the centerlines of the road segments is defined as the spherical distance from the projected coordinate points to the nearest point on the geometric string of the road segment centerline. The spherical distance is calculated using the Haversine formula, which takes the latitude and longitude of two points as input and outputs the length of the spherical arc between the two points, in meters. During the calculation, the projected coordinate points are sequentially compared with the centerlines of each road segment in the ordered road segment sequence. For a single road segment, the minimum distance between the projected coordinate points and all segments on the centerline of that road segment is taken as the matching distance for that road segment. After traversing all road segments, the matching distance value corresponding to each candidate road segment is obtained.

[0020] Next, the matching distance is compared with the distance judgment threshold. The distance judgment threshold is set based on the physical constraints of road width and positioning error. The width of a single lane in urban roads is usually between 3.0m and 3.5m, and the half-width of a two-way four-lane road is about 7m to 10m. Adding the civilian positioning accuracy of the global navigation satellite signal, which is about 5m to 10m, the maximum sum of the half-width of the road and the positioning accuracy is taken, with 15m as the baseline distance judgment threshold. This threshold ensures that the school bus can correctly match the actual driving segment even if there is a positioning deviation under normal circumstances, while effectively distinguishing mismatches from adjacent parallel roads. In open suburban road sections with good satellite signals, the threshold can be lowered to 10m; in urban canyons where tall buildings cause severe multipath effects, it can be raised to 20m. The adjustment range is limited to 0.7 to 1.3 times the baseline value. If the matching distance is less than the distance judgment threshold, the school bus is determined to be on that road segment and the road segment is marked as a candidate road segment. If the matching distance from the projected coordinate point to all road segments is not less than the distance judgment threshold, the school bus is determined to be in a yaw state, triggering a yaw alarm and notifying the dispatcher to intervene and verify.

[0021] Finally, a temporal connectivity check is performed on each candidate road segment to determine the current driving segment. The temporal connectivity check checks whether road segments matched by adjacent positioning points are continuously reachable in the road network structure, based on the topological connections of the road segments. Specifically, it obtains the candidate road segment set for the current positioning point and the confirmed road segments for the previous positioning point, and checks one by one whether there is a direct connectivity relationship between the current candidate road segment and the previously confirmed road segment. A connectivity relationship is defined as two road segments sharing a node in the road topology network or being reachable with no more than one intermediate node. If a connectivity relationship exists, the candidate road segment is retained; if no connectivity relationship exists, the candidate road segment may be a mismatch and is eliminated. After the connectivity check, if only one candidate road segment remains, it is directly determined as the current driving segment. If multiple candidate road segments remain, the one with the smallest matching distance is selected as the current driving segment. If all candidate road segments are eliminated, it is determined that the school bus has deviated from the preset driving route, and the yaw handling process is initiated.

[0022] In step S12, obtaining the real-time traffic congestion index corresponding to the current travel segment and performing dynamic correction processing based on the real-time traffic congestion index to obtain the real-time travel speed includes: The real-time traffic congestion index of the current travel segment in the current time period is obtained through a third-party traffic data interface; The real-time traffic congestion index of the current travel segment is calculated using the first-order difference method to obtain the dynamic change rate of road conditions. High-frequency noise filtering is performed on the dynamic change rate of road conditions to obtain smooth road condition trend characteristics; The smooth road condition trend characteristics and the real-time traffic congestion index are input into a preset nonlinear mapping model to obtain the speed attenuation coefficient. The reference driving speed of the current driving segment is obtained, and the real-time driving speed is obtained by multiplying the reference driving speed with the speed attenuation coefficient.

[0023] It should be noted that, firstly, the real-time traffic congestion index of the current travel segment for the current time period is obtained through a third-party traffic data interface. This third-party traffic data interface uses Hypertext Transfer Protocol (HTTP) and Application Programming Interface (API) keys for authentication. The keys are assigned by the traffic data service provider during registration and stored in an encrypted configuration file on the vehicle terminal. The request parameters include the segment identifier code and the current timestamp. The segment identifier code is provided by the segment number obtained from map matching in step S11, and the timestamp is accurate to the minute. The interface returns a lightweight data exchange format containing four fields: segment identifier code, data update time, real-time traffic congestion index, and traffic status description. The real-time traffic congestion index is a dimensionless floating-point number between 0 and 10. A smaller value indicates smoother traffic flow; 0 represents a completely free-flowing state where vehicles can travel at the road speed limit, while 10 represents severe congestion where vehicles are almost stationary. The data update frequency is once every 2 minutes, and each request returns the most recently updated congestion index data.

[0024] Next, the first-order difference is used to calculate the dynamic rate of change of traffic conditions on the real-time traffic congestion index of the current travel segment. The first-order difference is calculated by subtracting the previously acquired real-time traffic congestion index from the current one, dividing the difference by the time interval between the two data acquisitions (in minutes). The result represents the change in the congestion index per minute. A positive rate of change indicates that road congestion is worsening, while a negative rate indicates that congestion is easing. The larger the absolute value, the more drastic the change. The time window for the first-order difference is naturally determined by the time difference between two valid data acquisitions; there is no fixed window length. When the congestion level rapidly increases from 3 to 7 within 2 minutes, the dynamic rate of change of traffic conditions can reach above 2, indicating that a sudden traffic accident or temporary traffic control may occur ahead.

[0025] A low-pass filter is used to filter high-frequency noise from the dynamic rate of change of traffic conditions, resulting in smooth traffic trend characteristics. A first-order Butterworth low-pass filter is selected, which has the flattest amplitude-frequency response characteristic within its passband, effectively preserving the true trend components of traffic condition changes while suppressing short-term spikes caused by sensor sampling jitter or data transmission delays. The filter's cutoff frequency is set based on the effective frequency band of traffic condition changes. Traffic flow theory indicates that the congestion evolution cycle with practical predictive significance is usually greater than 5 minutes, corresponding to a frequency of 0.0033Hz. Therefore, the cutoff frequency of the low-pass filter is set to 0.003Hz to filter out slow drift components below this frequency. The filter's transfer function is discretized by mapping from the analog domain to the digital domain using a bilinear transform method. The sampling frequency is consistent with the acquisition frequency of the traffic congestion index, set to once every 2 minutes, i.e., 0.0083Hz. The smoothed traffic trend characteristics output after filtering retain the effective mid-to-high frequency components of the traffic condition change rate and remove the interference of random noise, enabling the subsequent speed estimation model to accurately capture the direction of congestion evolution.

[0026] Subsequently, the smoothed traffic trend features and the real-time traffic congestion index are input into a pre-defined nonlinear mapping model to obtain the speed decay coefficient. The pre-defined nonlinear mapping model uses a three-layer feedforward neural network. The input layer has two neurons that receive the real-time traffic congestion index and the smoothed traffic trend features, respectively processed by min-max normalization. Normalization linearly maps the two features to the interval between -1 and 1. The hidden layer has one layer with 16 neurons, and the hyperbolic tangent function is used as the activation function to introduce nonlinear mapping capability. The output layer has one neuron, and the sigmoid function is used as the activation function to compress the output value to between 0 and 1. The output value represents the speed decay coefficient; a value of 1 indicates no speed decay and the vehicle is traveling at the road speed limit, while a value close to 0 indicates that the speed approaches zero and the vehicle can hardly move forward.

[0027] The training data for the network was collected by simultaneously recording the real-time traffic congestion index and the average speed of the school bus passing through each road segment on the school bus route for 30 consecutive operating days. The average speed was calculated by dividing the segment length by the time taken to pass through the segment. The segment length was extracted from the electronic map system as the total arc length of the geometric string of the segment's centerline. The average speed was divided by the legal speed limit for that segment to obtain the true attenuation coefficient, which was used as the training label. The legal speed limit was extracted from the road attribute information of the electronic map system: 30 km / h for roads within the school campus, 50 km / h for secondary urban roads, and 60 km / h for main urban roads. Approximately 2500 valid samples were collected and divided into training and validation sets in a 7:3 ratio along the time axis. The training uses the mean squared error loss function, and the optimization algorithm is adaptive moment estimation. The initial learning rate is set to 0.001, and the batch size is 32. After every 20 rounds of traversing the complete training set, the learning rate decays to 0.9 times the previous value. Training stops when the loss on the validation set fails to decrease further after 12 consecutive rounds. After training, the network's weight and bias parameters are fixed. During online inference, only forward propagation is performed. Inputting the real-time traffic congestion index and smoothed road condition trend features into the network outputs the speed decay coefficient.

[0028] Then, the real-time speed is obtained by multiplying the baseline speed by the speed attenuation coefficient output by the nonlinear mapping model. The baseline speed is the smaller of the legal speed limit for that road segment and the average speed of school buses passing through that road segment during the same period over the past 30 days. This setting ensures that even if there is no congestion on that road segment, the estimated speed will not exceed the legal speed limit. Furthermore, if the actual speed on that road segment has been consistently lower than the speed limit due to narrow roads or frequent stops during the school bus's historical operation, the lower value is used as the baseline to better reflect actual operating conditions. The unit of real-time speed is km / h, rounded to one decimal place. This real-time speed, along with the road segment identifier and the current timestamp, is stored in the speed time-series buffer of the onboard terminal. The speed time-series buffer stores all speed records from the past 30 minutes in chronological order, with each record containing three fields: road segment identifier, timestamp, and real-time speed.

[0029] Reference Figure 2 This invention demonstrates the process by which the real-time speed of a school bus is dynamically corrected and accurately estimated based on a real-time traffic congestion index and a smoothed road condition change rate during operation. By integrating the real-time traffic congestion index and the smoothed road condition change rate for comprehensive decision-making, and calculating the speed attenuation coefficient in real time according to the dynamically changing traffic environment, this invention accurately captures local delays caused by sudden changes in road conditions. This provides an accurate speed basis for subsequent calculations of real-time travel times, thus providing a solid data foundation for improving the accuracy of school bus arrival time estimation in complex and dynamic traffic environments.

[0030] In step S13, the step of calculating the real-time segment travel time based on the real-time driving speed and the remaining distance of the current travel segment includes: Extract the total physical length of the current driving segment in the electronic map, and obtain the cumulative driving distance that the school bus has entered the segment; Subtract the accumulated travel distance from the total physical length to obtain the remaining distance of the current travel segment; Divide the remaining distance by the real-time driving speed to obtain the theoretical travel time of the current road segment; Obtain the number of traffic light nodes and the average waiting time of the current travel segment; The intersection delay time is obtained by multiplying the number of traffic light nodes with the average waiting time. The real-time segment travel time is obtained by summing the theoretical time with the intersection delay time.

[0031] It should be noted that, firstly, the total physical length of the current driving segment in the electronic map is extracted, and the cumulative driving distance of the school bus in that segment is obtained. The total physical length is directly read from the preset electronic map system. This attribute is stored in the segment length field of the segment data table, recording the complete arc distance from the starting node to the ending node of the segment centerline in meters. The cumulative driving distance is obtained by relying on the positioning data output in real time by the global navigation satellite signal receiving module built into the vehicle terminal. When step S11 determines that the school bus has entered the current driving segment, the projected coordinate point at the moment the vehicle enters is recorded as the segment entrance reference point. Thereafter, a new projected coordinate point is collected every 1 second, and the arc length between the current projected coordinate point and the segment entrance reference point along the segment centerline is taken as the cumulative driving distance.

[0032] Next, the total physical length is subtracted from the accumulated travel distance to obtain the remaining distance of the current travel segment. The remaining distance represents the physical path length that the school bus still needs to travel from its current location to the end of this segment. Then, the remaining distance is divided by the real-time travel speed to obtain the theoretical travel time for the current travel segment. The theoretical travel time represents the number of seconds required for the school bus to travel the remaining distance at the current dynamically corrected travel speed, in seconds, rounded to one decimal place. For example, if the remaining distance is 450m and the real-time travel speed is 36km / h (10m / s), the theoretical travel time is 45.0 seconds.

[0033] Then, the number of traffic light nodes and the average waiting time of the current driving segment are obtained. The number of traffic light nodes is extracted from the road attribute information of the preset electronic map system. This information is marked in the electronic map as a traffic node type field. A value of 1 indicates that the node has traffic light control, and a value of 0 indicates that the node is an intersection without traffic lights or a turning point within the road segment. The traffic light marking values ​​of all traffic nodes passed through by the center line of the current driving segment are summed to obtain the number of traffic light nodes in the road segment. The average waiting time is set based on the historical waiting data recorded when the school bus passes through the road segment in the past 30 days. The on-board terminal automatically records a stopping and waiting event each time the school bus comes to a complete stop in front of a traffic light and the speed drops below 0.5 m / s. When the speed recovers to above 0.5 m / s, the waiting time is calculated and stored in the historical waiting record table. The arithmetic mean of waiting times at all traffic lights on this road segment over the past 30 days is calculated. If there are no waiting records for this road segment in the past 30 days, the global mean of waiting times at all traffic lights along the entire school bus route is used as the default value. The global mean is extracted from the historical waiting record table and updated in batches after the end of each day's operation. The unit of average waiting time is seconds, rounded to the nearest integer.

[0034] Next, the number of traffic light nodes and the average waiting time are multiplied to obtain the intersection delay time. The intersection delay time represents the total expected stopping delay of the school bus on the current route segment due to traffic light control, in seconds. The multiplication operation assumes that the waiting behavior at each traffic light node is independent and conforms to the historical average waiting time, an assumption that is generally valid under the premise that the school bus operating route is relatively fixed.

[0035] Finally, the theoretical time is summed with the intersection delay time to obtain the real-time segment travel time. The real-time segment travel time represents the total estimated time required for the school bus to travel from its current location to the end of the current road segment under the current traffic conditions, in seconds. The real-time segment travel time, along with the road segment identifier code, current timestamp, and remaining distance, is written into the segment travel cache of the onboard terminal, storing all segment travel time records from the past 60 minutes in chronological order.

[0036] In one implementation, the step of performing transmission effect analysis based on the real-time segment travel time to obtain the cumulative effect value includes: Obtain the historical average travel time of the current travel segment, calculate the deviation between the real-time travel time and the historical average travel time, and obtain the segment delay increment; If the incremental delay of the road segment is greater than the delay judgment threshold, the incremental delay of the road segment is input into a preset cascade transmission effect model to calculate the downward cumulative transmission weight of the delay to subsequent stations; if the incremental delay of the road segment is not greater than the delay judgment threshold, the cumulative transmission weight is reset to zero. Based on the cumulative transmission weights output by the cascade transmission effect model, the delay increment of the road segment is integrally aggregated to obtain the cumulative effect value.

[0037] It should be noted that, firstly, the historical average travel time of the current travel segment is obtained. The deviation between the real-time travel time and the historical average travel time is then calculated to obtain the segment delay increment. The historical average travel time is extracted from the vehicle terminal's historical travel database. This database is indexed by segment identification code and operating period, storing travel time records for each segment in each time period over the past 30 operating days. The arithmetic mean of all travel times for the current travel segment within the same operating period over the past 30 days is taken as the historical average travel time. The same operating period is defined as the corresponding time periods within three time periods: 6:30 AM to 9:00 AM for the morning shift, 11:30 AM to 2:00 PM for the afternoon shift, and 3:30 PM to 6:30 PM for the evening shift. The segment delay increment is calculated by subtracting the historical average travel time from the real-time travel time, in seconds, rounded to one decimal place. A positive value for the road segment delay increment indicates that the current passage through the road segment takes longer than the historical normal level, i.e., there is a delay. A negative value indicates that the passage through the road segment is faster than the historical normal level, i.e., there is an advance. The sign and absolute value of the road segment delay increment together characterize the degree and direction of the deviation of the current road segment's operating efficiency from the normal.

[0038] Next, the incremental delay of a road segment is compared with the delay judgment threshold. The delay judgment threshold is set based on the punctuality tolerance standard of the school bus operation timetable. The school bus operation specifications require that the arrival time deviation of a single stop should not exceed 3 minutes, which is considered normal fluctuation and will not trigger dispatch intervention. The travel time fluctuation of a single road segment is the direct source of the station deviation. Therefore, 3 minutes, or 180 seconds, is used as the baseline delay judgment threshold. When the incremental delay of a road segment exceeds 180 seconds, it indicates that the delay of the current road segment has exceeded the normal fluctuation range and will have a significant transmission impact on the arrival time of subsequent stops. In suburban road segments with wide roads and optimized traffic light timing, the threshold can be lowered to 120 seconds to improve the punctuality rate requirement. In urban road segments with normal congestion during morning and evening rush hours, the threshold can be raised to 240 seconds to avoid frequently triggering transmission calculations and increasing unnecessary computational overhead. The adjustment range is limited to 0.7 to 1.3 times the baseline value.

[0039] If the incremental delay of a road segment exceeds the delay judgment threshold, the incremental delay is input into a pre-defined cascade transmission effect model to calculate the cumulative downward transmission weight of the delay to subsequent stations. The pre-defined cascade transmission effect model uses a two-layer Long Short-Term Memory (LSTM) network specifically designed to handle road segment delay data arranged spatially along the school bus's direction of travel. The LSM network's input layer has one feature dimension, receiving the incremental delay value after min-max normalization to the 0-1 range. It also inputs a location encoding vector representing the current road segment's relative position within the entire school bus route. The location encoding uses an alternating sine and cosine encoding method with a 16-dimensional encoding dimension to ensure that information from different positions can be distinguished by the network. The LSM network has two hidden layers, each containing 32 hidden units. The hyperbolic tangent function is used as the state update function, and the sigmoid function is used within the output and forget gates. Dropout regularization is applied between layers with a dropout ratio of 0.2 to prevent overfitting on the training set. The output layer is connected to a fully connected layer, which maps the hidden state of the last layer to the cumulative propagation weights corresponding to each subsequent station. The output dimension is equal to the total number of remaining stations from the current end point of the road segment to the end point of the route. The output layer activation function uses a sigmoid function to limit the weights to between 0 and 1.

[0040] The training data for the cascading effect model was collected by gathering operation records for all trips over the past 40 operating days. The deviations between the actual travel time and the historical average travel time for each segment along the route were used as the input sequence. The arrival time deviations recorded for subsequent stops of the same trip were used as training labels. The ground truth for the arrival time deviations was determined by the onboard terminal recording an arrival timestamp via an electronic fence each time the school bus stopped at a stop. The difference between the arrival timestamp and the planned arrival time for that stop on the timetable was the arrival time deviation. The ground truth labels were arranged sequentially according to the subsequent stops to form a deviation sequence. Approximately 1800 valid samples were collected, each corresponding to an event in a trip where the delay increment on a certain segment exceeded the delay threshold. The samples were divided into training and validation sets in a 7:3 ratio along the timeline. The training uses the mean squared error loss function, and the optimization algorithm is adaptive moment estimation. The initial learning rate is set to 0.0008, the batch size is 16, and the learning rate is decayed to 0.92 times the previous value after traversing the complete training set for 15 rounds. Training stops when the loss on the validation set does not decrease further after 10 consecutive rounds.

[0041] After training, the weight and bias parameters of the Long Short-Term Memory (LSTM) network are fixed. During online inference, the normalized segment delay increment is concatenated with the hidden state calculated in the previous layer and input into the network for forward propagation. The cumulative propagation weight of each subsequent station is output. If the segment delay increment is not greater than the delay judgment threshold, the current delay will not have a significant propagation effect on the downstream. The cumulative propagation weight is directly reset to zero, and a vector with zero weight is output for all subsequent stations. In subsequent integral aggregation calculations, the corresponding delay contribution terms are all zero.

[0042] Subsequently, based on the cumulative transmission weights output by the cascaded transmission effect model, the incremental delay of the road segment is calculated using integral aggregation to obtain the cumulative effect value. The integral aggregation calculation accumulates sequentially from the current road segment to each downstream segment along the school bus's direction of travel. The original incremental delay of the current road segment is multiplied by the cumulative transmission weight of the corresponding k-th subsequent station output by the model to obtain the delay contribution value allocated to that station. The delay contribution value of each subsequent station is added to the total delay contribution of all upstream segments between that station and the current road segment. The integral aggregation starts at the next station of the current road segment and terminates at the route terminus or the 10th subsequent station, whichever is smaller. The termination condition is designed to balance the propagation attenuation characteristics of the transmission effect with computational efficiency. The cumulative effect value represents the total impact of the current road segment delay on the arrival time of downstream stations after propagation attenuation, expressed in seconds. A larger value indicates a more severe delay in arrival time due to the current road segment delay, while a smaller value indicates a milder delay impact or that the delay is absorbed by natural recovery time.

[0043] In step S14, calculating the time offset of the cumulative effect value to subsequent stations includes: Extract the original arrival times of the subsequent stations; Using the original arrival time as a reference, shift the time axis backward by the cumulative effect value by the length of time, and output the time offset of the subsequent station.

[0044] It should be noted that the original arrival times of subsequent stations are extracted first. These original arrival times are read from the operating timetable stored locally on the vehicle terminal. The operating timetable is indexed by shift number and station number. Each record contains four fields: station name, station code, planned arrival time, and planned departure time. The planned arrival time is stored as three integer values ​​in 24-hour format (hour, minute, and second). Subsequent stations are identified by traversing downstream along the ordered sequence of segments of the school bus's preset route, starting from the current driving segment determined in step S11. All stations located after the end of the current driving segment on the preset route are extracted and sorted from front to back according to the driving direction to form a list of subsequent stations. Each station in the list corresponds one-to-one with the planned arrival time in the operating timetable through its station code. During extraction, the planned arrival time corresponding to each subsequent station in the operating timetable is read sequentially and converted into an integer value in seconds for subsequent offset calculations.

[0045] Next, using the original arrival time as a baseline, the time axis is shifted backward by the cumulative effect value, and the time offset of the subsequent station is output. The cumulative effect value is a scalar value, representing the cumulative time impact of the delay already incurred in the current road segment on downstream stations after attenuation through the cascading transmission model. This impact is a global quantity, not an independent component for each station. For station k in the list of subsequent stations, its original arrival time is added to the cumulative effect value and then multiplied by a station location attenuation coefficient to obtain the adjusted arrival time of the station. The difference between the adjusted arrival time and the original arrival time is the time offset of the station. The station location attenuation coefficient decreases as the station number k increases, reflecting the principle that stations farther away from the road segment where the delay occurred are less affected by the delay.

[0046] The station location attenuation coefficient is calculated using an exponential decay method. The attenuation coefficient for the first downstream station on the current road segment is set to 1. For each subsequent station, this coefficient is multiplied by 0.85; that is, the attenuation coefficient for the second downstream station is 0.85, the attenuation coefficient for the third downstream station is approximately 0.72, and so on. The attenuation coefficient of 0.85 is determined by least-squares fitting of the prediction bias of the cascade transmission effect model on the validation set. On the validation set, the transmission weights output by the model are fitted with attenuation curves against the actual arrival time deviations. The fitting results show that an attenuation factor of 0.85 achieves a goodness-of-fit of over 0.9 on various routes.

[0047] In one implementation, if the stop buffer margin corresponding to the time offset is less than the minimum buffer threshold, then a smoothing correction process is performed on the time offset to obtain a corrected arrival time, and the corrected arrival time is aligned using the school bus departure interval to generate the final arrival time, including: Obtain the latest departure time preset for the subsequent stations; Perform a difference operation between the latest departure time and the time offset, and output the stopping buffer margin that reflects the current station's time leeway; If the stop buffer margin is less than the minimum buffer threshold, the time offset is extracted and superimposed with the preset elastic time compensation step size to obtain the corrected arrival time. The school bus departure interval is used to perform clock alignment and rounding normalization on the corrected arrival time to generate the final arrival time.

[0048] It should be noted that the first step is to obtain the latest departure time preset for subsequent stations. The latest departure time is read from the operating timetable stored locally on the onboard terminal. This timetable records both the planned arrival time and the planned departure time for each station. The difference between the planned departure time and the planned arrival time is the preset standard stop time for that station, used to facilitate passenger boarding and alighting. The latest departure time is determined by adding a flexible extension window to the planned departure time. The flexible extension window is set according to the maximum tolerable delay standard of that station in the operating specifications. The campus shuttle bus operating specifications divide stations into three levels based on passenger flow: high-passenger-flow stations such as teaching building stations and cafeteria stations have a flexible extension window of 90 seconds; medium-passenger-flow stations such as dormitory area stations have a window of 60 seconds; and low-passenger-flow stations such as the school gate station have a window of 30 seconds. The latest departure time for a school bus is obtained by summing the planned departure time with the flexible extended time window of the corresponding station. The latest departure time means the latest time that the school bus is allowed to stay at the station without affecting the normal departure of the next bus. The unit is converted to an integer value based on seconds for subsequent calculations.

[0049] Next, the difference between the latest departure time and the time offset is calculated, outputting a stop buffer margin reflecting the current station's time leeway. The difference calculation involves subtracting the sum of the original arrival time and the time offset from the latest departure time, resulting in a difference in seconds. This difference represents how many seconds remain before the latest departure time if the school bus arrives at the current station at the corrected arrival time, considering delays along the road ahead. A positive stop buffer margin indicates that the school bus can perform normal stop and boarding operations at the current station without worrying about delays affecting subsequent departures; a negative stop buffer margin indicates that the stop operation time at the current station has been squeezed to below the minimum operational safety margin by accumulated delays along the road ahead. Without compensation measures, the school bus departure will be delayed, causing a chain reaction affecting subsequent trips.

[0050] Next, the stop buffer margin is compared with the minimum buffer threshold. The minimum buffer threshold is set based on the minimum safe time standard for school bus stop operations. Stop operations include five stages: the vehicle coming to a complete stop, the doors opening, passengers boarding and alighting, the doors closing, and the vehicle starting again. The shortest safe execution times for these five stages, as measured, are 3 seconds, 2 seconds, 8 seconds, 2 seconds, and 5 seconds, respectively, totaling 20 seconds. Using 20 seconds as the baseline minimum buffer threshold, if the stop buffer margin is less than 20 seconds, the school bus cannot even complete the shortest safe boarding and alighting operation at that stop, and the arrival time at that stop must be corrected through subsequent flexible time compensation. In university campuses where passengers are mainly adults, the minimum buffer threshold can be lowered to 15 seconds to speed up turnaround. In multi-campus routes that also serve kindergartens or primary schools, the threshold can be raised to 30 seconds to ensure the safe boarding and alighting of younger passengers. The adjustment range is limited to 0.75 to 1.5 times the baseline value.

[0051] The process is then divided according to the comparison results. If the stop buffer margin is less than the minimum buffer threshold, the extracted time offset is superimposed with the preset flexible time compensation step size to obtain the corrected arrival time. The flexible time compensation step size is set according to the flexible extension time window of the corresponding station level: 45 seconds for high passenger flow stations, 30 seconds for medium passenger flow stations, and 15 seconds for low passenger flow stations. Only one step size superposition is performed each time to avoid over-compensation and artificially increasing the arrival time. The superposition process is as follows: the time offset of the station calculated in step S14 is added to the flexible time compensation step size to obtain the compensated time offset, and then the original arrival time is added to the compensated time offset to generate the corrected arrival time. After the superposition compensation is completed, the stop buffer margin is recalculated to check whether it meets the minimum buffer threshold. If it still does not meet the threshold, another compensation step size is superimposed, but the number of consecutive superpositions does not exceed 3 to prevent the arrival time from being delayed indefinitely. If the stop buffer margin is still less than the minimum buffer threshold after 3 consecutive superpositions, an operation scheduling alarm is issued to notify manual intervention.

[0052] Finally, the arrival time is clock-aligned using the school bus departure interval. The school bus departure interval is the minimum interval between two adjacent buses on the same route, and the campus shuttle bus operation specifications set this interval to 10 minutes (600 seconds). Clock alignment means rounding the arrival time (in seconds) to the nearest 600. This rounding is done by dividing the arrival time by 600, rounding to the nearest hundredth, and then multiplying by 600, resulting in a standardized arrival time with a minimum granularity of 10 minutes. The purpose of this standardization is to eliminate the inconvenience caused by minute fluctuations at the second level, and to facilitate passengers' memory and expectation of the school bus arrival time in 10-minute increments. The final output after standardization is the final arrival time, in a 24-hour format combining hours and minutes. Second-level information is discarded and stored in the onboard terminal's arrival time release cache as estimated information to be pushed to waiting passengers.

[0053] If the stop buffer margin is not less than the minimum buffer threshold, it indicates that the stop time at the current station is ample and no additional elastic compensation is required. In this case, the original arrival time is directly added to the time offset to obtain the final arrival time without any compensatory adjustment.

[0054] After obtaining the final arrival time, it also includes: Generate a standard time string based on the final arrival time, and bind it with the pre-acquired school bus equipment identification code and the pre-acquired target station code to construct a content data packet; Obtain the user subscription list associated with the target site code, and parse it to obtain push tokens for multiple target user terminals; The content data packet and the push token are structured and encapsulated to generate an asynchronous push task flow; The asynchronous push task stream is distributed to each target user terminal through a high-concurrency message push channel to complete the arrival time estimation.

[0055] It should be noted that, firstly, a standard time string is generated based on the final arrival time, and then bound to the school bus equipment identification code and the target station code to construct a content data package. The final arrival time is stored in a 24-hour format, with the hour occupying 2 digits ranging from 00 to 23, and the minute occupying 2 digits ranging from 00 to 59. The standard time string adopts the ISO 8601 format defined by the International Organization for Standardization, concatenating the current complete date and the final arrival time into a unified string. The concatenation rule is to take the complete date of the school bus operation day, occupying 8 digits followed by a capital letter T, and combine it with the final arrival time and a 2-digit zero-second value. The school bus equipment identification code is read from the internal storage chip of the on-board terminal. This identification code is a 16-bit hexadecimal string that is burned into the device at the factory and cannot be changed, and is used to uniquely identify a school bus in the school bus operation platform. The target station code is obtained from the subsequent station list in step S14. Each station is assigned an 8-digit code, with the first 3 digits representing the route number, the middle 3 digits representing the station's sequence number on that route, and the last 2 digits being a check digit. The time string, school bus equipment identification code, and target station code are concatenated in sequence and separated by vertical bars to form a text record. This text record is the main body of the content data packet. At the same time, a message serial number is attached for subsequent push tracking. The message serial number is generated by concatenating the school bus equipment identification code, the current timestamp, and the station code and then hashing it using the SHA-256 algorithm to take the first 12 bits.

[0056] Next, the user subscription list associated with the target site code is retrieved, and push tokens for multiple target user terminals are parsed. The user subscription list is stored in the subscription relationship database of the school bus operation platform. This database maintains three relationship tables: the user information table records two fields: user identifier and account status; the site information table records two fields: site code and site name; and the subscription relationship table records two fields: the correspondence between the user identifier and the site code, and the subscription effective timestamp. Users select the sites they wish to receive notifications from through the site selection interface in the mobile application. The mobile application sends the user's user identifier and the selected site code to the backend server of the school bus operation platform via a POST request using the Hypertext Transfer Protocol. The server inserts a record into the subscription relationship table to complete the subscription. Simultaneously, after the user agrees to receive push notifications in the mobile application, the mobile operating system registers a push token with the push service provider and binds this push token to the user identifier, storing it in an extended field of the user information table. The parsing process is as follows: using the target site code as the query condition, the subscription association table is retrieved to obtain a list of all user identifiers that have subscribed to the site. Then, the user information table is retrieved for each user identifier to extract the corresponding push token. At the same time, users whose account status is disabled are filtered out. Finally, a list of push tokens is generated, and each element in the list is a string of no more than 512 characters.

[0057] The content data package and push tokens are then structurally encapsulated to generate an asynchronous push task stream. The structured encapsulation uses a lightweight data exchange format. The top-level object contains three keys: `content` (corresponding to the complete text content of the content data package), `tokens` (corresponding to the string array of push tokens), and `priority` (corresponding to the priority value of `high` or `normal`). When the final arrival time is less than 5 minutes from the current time, the priority is set to `high` to trigger the push channel's emergency delivery mechanism; when the arrival time is greater than or equal to 5 minutes from the current time, it is set to `normal` to reduce push resource consumption. Each encapsulated top-level object is treated as a push task. When traversing the target site's encoded push token list, the list is divided into groups of 100 tokens to generate multiple push tasks. Each push task corresponds to a subset of `tokens` and the same content data package. These push tasks are pushed independently and in parallel to improve distribution efficiency. The multiple push tasks generated in batches are arranged sequentially to form an asynchronous push task stream, which is stored in a first-in-first-out message queue awaiting consumption and execution by the push scheduler.

[0058] Then, the high-concurrency message push channel distributes the asynchronous push task stream to various target user terminals. The high-concurrency message push channel is built on a publish-subscribe model. After the push scheduler retrieves a push task from the head of the message queue, it creates an asynchronous coroutine to handle the task. Within the coroutine, a long-lived Transmission Control Protocol (TCP) connection is established simultaneously with all push tokens, connecting to the push gateway server of the remote push service provider on port 443. The communication protocol uses HTTP / 2 to support multiplexing, allowing multiple push requests to be transmitted simultaneously over a single connection. Each push request body contains three fields: push token, content data packet, and push priority. The request header includes the school bus operation platform's service authorization key for authentication. The service authorization key is a 256-bit string issued by the push service provider and stored in the platform's encrypted configuration center. After receiving the request, the push gateway server pushes the content data packet to the corresponding mobile application through the signaling channel maintained by the mobile operating system vendor. Upon receiving the request, the mobile application displays a notification message on the user interface indicating that the school bus will arrive at a specific stop at a specific time. After the push is completed, the push gateway server returns a response status code for each push request to the push scheduler. A status code of 200 indicates that the push has been delivered, a status code of 410 indicates that the push token has expired and needs to be cleared from the subscription relationship database, and other values ​​indicate that the push is temporarily unreachable and needs to be retried after 5 seconds.

[0059] After collecting the responses to all push requests, the push scheduler marks the push task as completed and records the number of successful and failed pushes. If the number of successful pushes is less than 95% of the total number of push requests, the failed requests are reassembled into a new push task and reinserted into the tail of the message queue to await further consumption. Once all push tasks in the asynchronous push task stream have been consumed, the push process ends, the arrival time is estimated, and the waiting passengers are informed of the school bus's estimated arrival time, which has been dynamically corrected and buffered.

[0060] Reference Figure 3 This invention demonstrates the visualized results of its dynamic adjustment and buffer management of final arrival times at subsequent stations based on real-time traffic condition correction and the cascading effect of delays. The invention generates time offsets for subsequent stations by calculating cumulative effect values ​​and performs smoothing corrections and departure interval alignment when stop buffer margins are insufficient, generating final arrival times with ample safety margins. This process accurately depicts the time accumulation and attenuation patterns of congestion delays across multiple stations, making the previously abstract delay propagation clearly visible. This provides a solid and reliable data foundation for school bus operation scheduling and accurate passenger arrival reminders in complex and dynamic traffic environments.

[0061] In summary, this invention discloses a method for estimating school bus arrival times, including obtaining the current latitude and longitude coordinates of the school bus and the real-time traffic congestion index of the road segment; performing time-series matching of the current latitude and longitude coordinates to obtain the current road segment; performing dynamic correction based on the real-time traffic congestion index to obtain the real-time driving speed; calculating the real-time segment travel time based on the real-time driving speed and remaining distance and performing transmission effect analysis to obtain the cumulative effect value; and calculating the time offset of subsequent stops based on the cumulative effect value to generate the final arrival time. This method achieves dynamic adjustment of arrival time and cumulative delay control, improving the accuracy and reliability of predictions under complex road conditions.

[0062] A second embodiment of the present invention provides a school bus arrival time estimation system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0063] It should be noted that the school bus arrival time estimation system provided in this embodiment of the invention is used to execute all the process steps of the school bus arrival time estimation method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0064] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0065] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for estimating school bus arrival time, characterized in that, include: Obtain the current latitude and longitude coordinates of the school bus during its journey, and perform time-series matching processing on the current latitude and longitude coordinates to obtain the current route segment; Obtain the real-time traffic congestion index corresponding to the current driving segment, and perform dynamic correction processing based on the real-time traffic congestion index to obtain the real-time driving speed; The real-time segment travel time is calculated by performing time extrapolation based on the real-time driving speed and the remaining distance of the current travel segment, and the cumulative effect value is obtained by performing transmission effect analysis based on the real-time segment travel time. Calculate the time offset of the cumulative effect value to subsequent stations; if the stop buffer margin corresponding to the time offset is less than the minimum buffer threshold, perform smoothing correction processing on the time offset to obtain the corrected arrival time, and perform alignment processing on the corrected arrival time using the school bus departure interval to generate the final arrival time; if the stop buffer margin is not less than the minimum buffer threshold, directly generate the final arrival time based on the time offset.

2. The school bus arrival time estimation method according to claim 1, characterized in that, The process of obtaining the current latitude and longitude coordinates during the school bus's journey, and performing time-series matching processing on the current latitude and longitude coordinates to obtain the current travel segment includes: The real-time location trajectory of the school bus during its journey is collected periodically, and the current latitude and longitude coordinates and the current timestamp are extracted from the real-time location trajectory. The current latitude and longitude coordinates are input into a preset electronic map system to perform projection mapping, thereby obtaining the projected coordinate points; Match the corresponding preset school bus route based on the current timestamp; Calculate the matching distance between the projected coordinate points and the center lines of each segment in the preset driving route; The matching distance is compared with a distance determination threshold. If the distance is less than the distance determination threshold, the school bus is determined to be in the corresponding candidate road segment. If the distance is not less than the distance determination threshold, the school bus is determined to be in a yaw state. Perform temporal connectivity checks on each candidate road segment to determine the current travel segment.

3. The school bus arrival time estimation method according to claim 1, characterized in that, The step of obtaining the real-time traffic congestion index corresponding to the current travel segment and performing dynamic correction processing based on the real-time traffic congestion index to obtain the real-time travel speed includes: The real-time traffic congestion index of the current travel segment in the current time period is obtained through a third-party traffic data interface; The real-time traffic congestion index of the current travel segment is calculated using the first-order difference method to obtain the dynamic change rate of road conditions. High-frequency noise filtering is performed on the dynamic change rate of road conditions to obtain smooth road condition trend characteristics; The smooth road condition trend characteristics and the real-time traffic congestion index are input into a preset nonlinear mapping model to obtain the speed attenuation coefficient. The reference driving speed of the current driving segment is obtained, and the real-time driving speed is obtained by multiplying the reference driving speed with the speed attenuation coefficient.

4. The school bus arrival time estimation method according to claim 1, characterized in that, The step of calculating the real-time segment travel time based on the real-time driving speed and the remaining distance of the current travel segment includes: Extract the total physical length of the current driving segment in the electronic map, and obtain the cumulative driving distance that the school bus has entered the segment; Subtract the accumulated travel distance from the total physical length to obtain the remaining distance of the current travel segment; Divide the remaining distance by the real-time driving speed to obtain the theoretical travel time of the current road segment; Obtain the number of traffic light nodes and the average waiting time of the current travel segment; The intersection delay time is obtained by multiplying the number of traffic light nodes with the average waiting time. The real-time segment travel time is obtained by summing the theoretical time with the intersection delay time.

5. The school bus arrival time estimation method according to claim 1, characterized in that, The step of performing conduction effect analysis based on the real-time segment travel time to obtain the cumulative effect value includes: Obtain the historical average travel time of the current travel segment, calculate the deviation between the real-time travel time and the historical average travel time, and obtain the segment delay increment; If the incremental delay of the road segment is greater than the delay judgment threshold, the incremental delay of the road segment is input into a preset cascade transmission effect model to calculate the downward cumulative transmission weight of the delay to subsequent stations; if the incremental delay of the road segment is not greater than the delay judgment threshold, the cumulative transmission weight is reset to zero. Based on the cumulative transmission weights output by the cascade transmission effect model, the delay increment of the road segment is integrally aggregated to obtain the cumulative effect value.

6. The school bus arrival time estimation method according to claim 1, characterized in that, The calculation of the cumulative effect value with respect to the time offset of subsequent stations includes: Extract the original arrival times of the subsequent stations; Using the original arrival time as a reference, shift the time axis backward by the cumulative effect value by the length of time, and output the time offset of the subsequent station.

7. The school bus arrival time estimation method according to claim 1, characterized in that, If the stop buffer margin corresponding to the time offset is less than the minimum buffer threshold, then a smoothing correction process is performed on the time offset to obtain the corrected arrival time, and the corrected arrival time is aligned using the school bus departure interval to generate the final arrival time, including: Obtain the latest departure time preset for the subsequent stations; Perform a difference operation between the latest departure time and the time offset, and output the stopping buffer margin that reflects the current station's time leeway; If the stop buffer margin is less than the minimum buffer threshold, the time offset is extracted and superimposed with the preset elastic time compensation step size to obtain the corrected arrival time. The corrected arrival time is clock-aligned using the school bus departure interval to generate the final arrival time.

8. The school bus arrival time estimation method according to claim 1, characterized in that, After obtaining the final arrival time, the process also includes: Generate a standard time string based on the final arrival time, and bind it with the pre-acquired school bus equipment identification code and the pre-acquired target station code to construct a content data packet; Obtain the user subscription list associated with the target site code, and parse it to obtain push tokens for multiple target user terminals; The content data packet and the push token are structured and encapsulated to generate an asynchronous push task flow; The asynchronous push task stream is distributed to each target user terminal to complete the arrival time estimation.

9. A school bus arrival time estimation system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the school bus arrival time estimation method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the school bus arrival time estimation method according to any one of claims 1 to 8.

Images