Highway toll robot system

Abstract

The invention discloses an expressway toll collection robot system, relates to the technical field of expressway toll collection, and constructs a full-process automatic expressway toll collection robot system through collaborative design of a toll collection control module and a special situation scheduling module. The charging control module integrates various types of hardware, realizes automatic identification of vehicle information, fee calculation and passage control, replaces manual work to complete a standardized charging process, and remarkably reduces manual operation errors and time consumption. The special situation scheduling module is used for quickly positioning the type of a special situation through object association relationship analysis for an abnormal state in a charging process, and scheduling staff to intervene; different from a traditional post response mode, the module predicts the severity level of a special situation based on flow characteristics, equipment states and environment characteristics by training a special situation prediction model and configures the number of workers in advance, so that lane congestion can be effectively avoided, and the overall traffic efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of highway toll collection, and in particular relates to a highway toll collection robot system. Background Art

[0002] With the continued growth of highway traffic, the traditional manual toll collection model has gradually exposed problems such as inefficiency and high labor costs. The existing toll collection system relies on human personnel to identify vehicle information, calculate fees, and handle special situations. This is prone to lane congestion during peak hours and is prone to high manual errors. Furthermore, special situation handling lacks foresight and typically requires a reactive response after an abnormality occurs, resulting in long processing delays and further exacerbating traffic pressure.

[0003] The present invention provides a highway toll collection robot system to solve the above technical problems. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a highway toll collection robot system, which constructs a fully automated highway toll collection robot system through the collaborative design of a toll control module and a special situation dispatching module; the toll control module integrates various types of hardware to realize automatic identification of vehicle information, fee calculation and traffic control, replacing manual completion of standardized toll collection processes, significantly reducing human operation errors and time consumption; the special situation dispatching module targets abnormal conditions in the toll collection process, quickly locates the special situation type through object association analysis, and dispatches staff to intervene; unlike the traditional "post-event response" mode, this module trains a special situation prediction model, predicts the severity level of the special situation based on traffic characteristics, equipment status and environmental characteristics, and allocates the number of staff in advance, which can effectively avoid lane congestion and improve overall traffic efficiency.

[0005] To achieve the above-mentioned object, a first aspect of the present invention provides a highway toll collection robot system, comprising a toll collection control module and a special situation dispatch module;

[0006] Toll collection control module: used to identify vehicle information and complete highway toll collection based on the vehicle information; as well as monitor abnormal conditions during the highway toll collection process; vehicle information is used to calculate tolls;

[0007] Special Situation Dispatch Module: used to identify special situations corresponding to abnormal states and dispatch staff to perform troubleshooting based on special situations; and

[0008] Used to train special situation prediction models; predict the severity level of special situations based on the special situation prediction models, and allocate staff according to the predicted severity level of special situations; among which, the special situation prediction model is built based on an artificial intelligence model.

[0009] Preferably, predicting the severity of a special situation based on a special situation prediction model includes:

[0010] Calling a special situation prediction model; wherein the special situation prediction model is trained and obtained based on historical special situation data of a single toll station;

[0011] The prediction input data is extracted from the historical special situation severity levels of the toll station, and the prediction input data is input into the special situation prediction model after being subjected to model adaptation processing to obtain the special situation severity level.

[0012] Preferably, training the special situation prediction model includes:

[0013] Obtain historical special situation information of toll stations and extract training input data and training output data from the historical special situation information; the training input data includes traffic characteristics, equipment status characteristics, and environmental characteristics, and the training output data includes the severity level of the special situation;

[0014] After the training input data and training output data are adapted to the model, the LSTM model is trained; the trained LSTM model is marked as a special situation prediction model.

[0015] Preferably, predicting the severity of a special situation based on a special situation prediction model includes:

[0016] Calling a special situation prediction model; wherein the special situation prediction model is trained based on historical special situation data of several toll stations of the same type;

[0017] Extract prediction input data from the historical severity levels of special situations at toll stations, process the prediction input data into a special situation prediction model after adaptive modeling, and obtain the severity level of the special situation;

[0018] Determine whether the toll station is an entrance; if so, assign staff based on the predicted severity level of the special situation; if not, revise the predicted severity level of the special situation.

[0019] Preferably, training the special situation prediction model includes:

[0020] Obtain historical special situation information for several similar toll booths, and extract training input data and training output data from the historical special situation information; the training input data includes traffic characteristics, equipment status characteristics, and environmental characteristics, and the training output data includes the severity level of the special situation;

[0021] After the training input data and training output data are adapted to the model, the LSTM model is trained; the trained LSTM model is marked as a special situation prediction model.

[0022] Preferably, the special situation prediction model is deployed in the cloud or at a toll station of the corresponding type; wherein the toll station type includes an entrance or an exit.

[0023] Preferably, the predicted severity level of the special situation is revised, including:

[0024] Extracting corrected input data from the associated entrance features corresponding to the toll station, performing adaptive modeling on the corrected input data and then inputting it into the special situation correction model to obtain corrected data;

[0025] The severity level of the special situation is corrected based on the correction data to obtain a corrected severity level of the special situation.

[0026] Preferably, training the special situation correction model includes:

[0027] Extract the associated entrance features corresponding to the toll station; the associated entrance features include route distance, travel time, and traffic transfer rate;

[0028] Extract the prediction residual of the special situation prediction model for the toll station and use the prediction residual as the supervision label; match the supervision label corresponding to the associated entrance feature according to the set window;

[0029] The associated entry features are used as input and the matched supervision labels are used as output to train the LSTM model and obtain the special situation correction model.

[0030] Preferably, staff should be assigned based on the predicted severity of the emergency, including:

[0031] Extract the level-number association table corresponding to the toll station; the level-number association table corresponds to the toll station one-to-one, and the severity level of the special situation in the level-number association table corresponds to the number of staff;

[0032] Match the severity level of the special situation in the severity level of the special situation with the level-number association table to determine the number of staff that need to be deployed.

[0033] Preferably, the construction of the level-number association table includes:

[0034] Extract the number of staff responsible for handling special situations at the toll station;

[0035] Extract the severity level range of special situations; divide the level range into several levels, match the number of associated staff for each level, and generate a level-number association table for toll stations.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention constructs a fully automated highway toll collection robot system through the collaborative design of the toll control module and the special situation dispatch module; the toll control module integrates various hardware to realize automatic vehicle information identification, fee calculation and traffic control, replacing manual completion of the standardized toll collection process, significantly reducing human operation errors and time consumption; the special situation dispatch module targets abnormal conditions in the toll collection process, quickly locates the special situation type through object association analysis, and dispatches staff to intervene; different from the traditional "post-event response" mode, this module trains a special situation prediction model, predicts the severity level of the special situation based on traffic characteristics, equipment status and environmental characteristics, and allocates the number of staff in advance, which can effectively avoid lane congestion and improve overall traffic efficiency.

[0038] 2. The special situation prediction model of the present invention is trained based on historical data of a single station or multiple stations of the same type. The input includes basic features such as traffic, equipment, and environment, and the timing patterns are captured through LSTM. In view of the characteristics that exit toll stations are affected by associated entrances, the special situation correction model introduces associated features such as route distance, driving time, and traffic transfer rate to dynamically calibrate the prediction results. The present invention avoids resource mismatch caused by ignoring cross-station associations through a two-layer mechanism of "basic prediction + associated correction", realizes the optimal allocation of human resources and equipment resources, and the special situation prediction model can be uniformly trained, updated and deployed, thereby improving the efficiency of model training and updating. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 Schematic diagram of the real-time processing steps of a special situation of the highway toll collection robot system according to an embodiment of the present invention;

[0041] Figure 2 Schematic diagram of the special situation prediction processing steps of the highway toll collection robot system according to an embodiment of the present invention;

[0042] Figure 3 Schematic diagram of the system principle of the highway toll collection robot system in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] See also Figure 1-Figure 3 , a first aspect of the present invention provides a highway toll collection robot system, including a toll collection control module and a special situation dispatch module;

[0045] Toll collection control module: used to identify vehicle information and complete highway toll collection based on the vehicle information; as well as monitor abnormal conditions during the highway toll collection process; vehicle information is used to calculate tolls;

[0046] Special Situation Dispatch Module: used to identify special situations corresponding to abnormal states and dispatch staff to perform troubleshooting based on special situations; and

[0047] Used to train special situation prediction models; predict the severity level of special situations based on the special situation prediction models, and allocate staff according to the predicted severity level of special situations; among which, the special situation prediction model is built based on an artificial intelligence model.

[0048] The highway toll collection robot system of the present invention is mainly used to realize the automation of highway toll collection, replacing the staff to complete simple highway toll collection work, thereby allowing the staff to devote themselves to special situation processing and improve the traffic efficiency of highway toll stations.

[0049] The highway robot is mainly controlled by the toll control module and the special situation dispatch module.

[0050] Toll collection control module: used to identify vehicle information and complete highway toll collection based on vehicle information.

[0051] The toll collection control module is connected to a number of devices, including but not limited to card readers, printers, coils, license plate recognition devices, automatic barriers, canopy lights, displays, lane lights, and ETC antennas. For example, the license plate recognition device identifies the license plate, the coil determines whether the vehicle has entered the designated location, the automatic barrier automatically lifts to release the vehicle after payment, and the display device is primarily used to alert, display fees, and display payment codes for driver interaction. To avoid content conflicts due to the angle of the display device, multiple displays can be configured, each displaying a single or non-conflicting content.

[0052] The highway robot system of the present invention is deployed at the entrance and exit of the highway. At the entrance, it mainly identifies vehicle information and releases the vehicle after the vehicle information is verified; at the exit, it identifies vehicle information, calculates the highway fee based on the vehicle information, and releases the vehicle after the highway fee is paid.

[0053] Vehicle clearance is accomplished through a coordinated effort of various devices. A coil detects the vehicle's arrival at a designated location. Once there, license plate recognition equipment (such as a high-definition camera) and an ETC antenna identify the vehicle. Vehicle information, including license plate, exterior, ETC information, and CPC / IC card information, is primarily used to record and calculate highway tolls. Upon verification at the entrance, an automatic barrier is activated to release the vehicle. At the exit, the toll is calculated based on the identified vehicle information according to built-in calculation rules. Once payment is successful, the automatic barrier is activated to release the vehicle.

[0054] Toll collection control module: monitors abnormal conditions during the high-speed toll collection process.

[0055] During the highway toll collection process, if any link has a problem, the highway fee payment cannot be completed, and it is determined to be an abnormal state. Therefore, the abnormal state means that the vehicle cannot pass through the toll station. The entrance mainly includes vehicle information identification and verification and vehicle release; if there is a problem with vehicle information identification, such as the license plate cannot be recognized, subsequent work cannot be carried out; if the vehicle information identification verification is passed, but the automatic lifting machine fails, it cannot be released smoothly. When the highway fee is calculated at the exit and collected through the payment code, if the payment code is not displayed clearly or the driver does not know how to use mobile payment, it will result in timeout and failure to complete the fee payment, and the vehicle cannot be released.

[0056] When a problem occurs in any link, the abnormal status is recorded. The occurrence of an abnormal status indicates the occurrence of a special situation, so the abnormal status is sent to the special situation dispatch module for special situation processing.

[0057] The Special Situation Dispatch Module primarily identifies special situations corresponding to abnormal conditions that occur during highway toll collection, coordinating with staff to eliminate the associated abnormal factors, thereby ensuring efficient traffic flow in each lane. For example, if the Toll Control Module cannot recognize the current vehicle's ETC information, staff will need to manually collect the toll and release the vehicle. The Special Situation Dispatch Module is also used to predict special situations. Based on historical data, it predicts possible future special situations and then allocates staff or deploys redundant equipment based on the predicted results, thereby improving vehicle traffic efficiency in extreme situations.

[0058] The special situation dispatch module primarily connects to the toll control module and various devices. After extracting abnormal status from the toll control module, it analyzes the devices associated with the abnormal status to determine the special situation. For example, if the abnormal status received by the special situation dispatch module is "payment not completed within the specified time", it may be due to a failure in generating or displaying the payment code, or it may be due to an abnormality in the vehicle's ETC device or a timeout in the driver's mobile payment. In this case, dispatch staff will need to manually perform toll collection to ensure vehicle traffic efficiency. Staff should also inspect and troubleshoot any equipment issues.

[0059] In other preferred embodiments, the special situation dispatch module can also obtain data stored in the database to assist in determining the special situation corresponding to the abnormal state, without having to connect to multiple devices to obtain their status data in real time, thereby reducing data acquisition costs.

[0060] Special situation dispatch module: used to identify special situations corresponding to abnormal states and dispatch staff to perform troubleshooting based on special situations.

[0061] During the highway toll collection process, "special situations" refer to special circumstances where payment cannot be completed normally due to abnormalities in vehicles, equipment, toll collection media or processes. The special situations corresponding to each special situation can be accurately located through internal communication data. Some special situations can be eliminated through programs. For example, if there is a problem with the Alipay payment code, a WeChat payment code can be generated through voice interaction. Some special situations require staff intervention and need to be reported to the staff once a special situation occurs. The special situations that occur during the highway toll collection process are mainly as follows:

[0062] 1. Abnormal passage medium

[0063] 1. CPC card / IC card abnormality:

[0064] 1) No card: The vehicle has not received the access card (e.g., missed the entry or lost the card), and the entry information needs to be manually confirmed (e.g., by checking the entry station through the license plate) and a new card will be issued.

[0065] 2) Bad card: The card is physically damaged (such as the chip falls off) or the data is abnormal (such as writing failure). The path information cannot be read. The bad card needs to be manually recovered and the path must be manually entered.

[0066] 2. ETC abnormality:

[0067] 1) Tag invalidation: The OBU (electronic tag) is not activated, falls off or is faulty, and automatic deduction cannot be completed.

[0068] 2) Insufficient card balance: The ETC account balance is insufficient and you need to switch to another payment method.

[0069] 3) Dispute over duplicate deductions: The system prompts duplicate deductions and requires manual verification of transaction records and processing of refunds.

[0070] 2. Abnormal license plate or vehicle model

[0071] 1. License plate recognition anomaly:

[0072] 1) No license plate or blurred license plate: The vehicle has no license plate, the license plate is blocked, or the camera fails to recognize it. The license plate must be manually entered or photographed for preservation.

[0073] 2) The license plate does not match the vehicle model: The identified vehicle model is inconsistent with the actual vehicle model, which may involve suspicion of toll evasion and requires manual review and adjustment of the vehicle model billing.

[0074] 3) If the license plate recorded on the pass card is inconsistent with the actual vehicle license plate (such as card replacement or duplicate license plate), the vehicle entry information must be verified and charges will be made based on the actual situation.

[0075] 2. Abnormal vehicle classification:

[0076] 1) If the driver objects to the vehicle classification, a manual re-determination is required based on the driving license or vehicle parameters.

[0077] 2) Abnormal vehicle weighing: If the entrance weighing data is significantly different from the actual load (such as overload), the vehicle must be directed to a designated area for re-inspection or handled in accordance with regulations.

[0078] 3. Payment anomalies (non-cash payment methods or cash payment process anomalies)

[0079] 1. Non-cash payment failed:

[0080] 1) Weak 4G / 5G signals may cause the mobile payment QR code to fail to load or the ETC transaction to time out. You need to switch to offline mode or manually generate a static code.

[0081] 2) If there are any abnormalities in third-party payment systems such as Alipay and WeChat, users need to be guided to change payment methods or temporarily store vehicle information for subsequent payment.

[0082] 3) If the lane antenna fails or the vehicle OBU communication is abnormal, the deduction cannot be completed and you need to manually trigger the ETC transaction again or switch to other payment methods.

[0083] 2. Cash payment issues:

[0084] 1) Insufficient cash / no change: If the driver does not have enough cash to pay the toll, the staff will need to negotiate a partial cash + mobile payment combination, or temporarily deposit the driver's ID and then pay the toll.

[0085] 2) If the cash change amount does not match the driver's expectations, the staff will need to retrieve the transaction records or surveillance video for verification.

[0086] 3) If a toll collection robot or a human detects a counterfeit currency, the staff must inform the driver on the spot and urge the driver to pay the toll normally as required.

[0087] Currently, when handling special situations at highway toll booths, staff intervene only after an abnormal situation (such as congestion or impassable conditions) occurs. To cope with the possibility of special situations at any time, each toll booth should be equipped with staff responsible for handling special situations. This allows for timely manual intervention when special situations occur, preventing lane congestion and affecting traffic efficiency.

[0088] Staff will first guide the driver to complete payment. If the problem isn't with the driver or vehicle, they'll be directed to a designated area to address the toll booth equipment malfunction. When an abnormality corresponds to multiple special conditions, staff can't accurately determine the specific cause. If they investigate each condition individually, significant time is wasted, potentially causing congestion in the corresponding lanes and impacting subsequent traffic efficiency.

[0089] After receiving an abnormal status, the system analyzes the abnormal status and pre-established object associations to determine the possible special situation that caused the abnormal status. Object associations primarily refer to the objects associated with the special situation. One special situation may be associated with multiple objects, and multiple special situations may be associated with the same object. Once a special situation occurs, at least one of the associated objects has a problem.

[0090] By way of example and not limitation, if the error message "Payment code generation failed" is returned when generating a payment code, the abnormal state could be a payment failure, and the corresponding special condition would be "Payment code generation failed." If the payment code is successfully generated but payment fails, the corresponding special condition could be "Display device failure" or "Driver payment failed."

[0091] Object associations refer to which objects are associated with special situations. Object associations can be constructed based on experience, knowledge graphs, and data mining techniques. They can also be constructed based on process steps. For example, at an exit, the toll collection process can be divided into information recognition, payment, and barrier release. The information recognition step can be associated with coils, card readers, license plate recognition devices, and ETC antennas; the payment step can be associated with display devices and drivers; and the barrier release step can be associated with automatic barriers.

[0092] It should be noted that the objects in the object association relationship can be devices or people. For example, in the fee payment process, the payment code is displayed normally on the display device, but the driver does not know how to scan the code to pay, and a special situation will occur due to timeout.

[0093] If an abnormality occurs, indicating that the vehicle is unable to complete payment, staff must intervene promptly, regardless of whether the issue lies with the toll booth or the driver (or their vehicle). Once staff intervene, they prioritize guiding the driver through payment to avoid lane congestion. If the driver cannot be released within a short period of time, they must guide the driver to a designated area. The staff member must then determine whether the relevant equipment indicated by the abnormality is faulty. If so, they must promptly restore or report the fault, while also implementing protective measures.

[0094] When staff intervene, their handheld smart terminals will prompt abnormal conditions and possible corresponding special situations. Staff can eliminate special situations one by one to determine the specific causes and deal with them in a timely manner to avoid affecting traffic efficiency.

[0095] In other preferred embodiments, if the matching special circumstances include both toll booth and driver-related reasons, such as a display device malfunction that prevents the payment code from being clearly displayed, or the driver failing to complete payment within a preset time, at least two staff members can be dispatched. One staff member is responsible for guiding the driver through payment and promptly releasing the driver. The other staff member is responsible for checking the display device and promptly repairing any malfunctions. If repairs are not possible, a reminder should be set.

[0096] Special situation dispatch module: used to train the special situation prediction model; predict the severity level of the special situation based on the special situation prediction model, and allocate staff according to the predicted severity level of the special situation.

[0097] Currently, highway toll booths handle special situations only after a situation has occurred. Staff then intervene according to the workflow, with handling efficiency dependent on both the effectiveness of alerts and staff response. Toll booth staffing is relatively fixed. If staff are insufficient, when a sudden surge in special situations occurs, such as multiple situations occurring within a short period of time, staff will be forced to investigate and resolve them one by one according to the order of warnings. This can even lead to staff shortages, resulting in significant time wastage and impacting vehicle traffic efficiency. If staff are overstaffed, human resources are often wasted.

[0098] The training process of the special situation prediction model can be referred to as follows:

[0099] 1. Collect historical severity levels of special situations and extract training input data and training output data from the historical severity levels using a sliding window;

[0100] 2. Use the processed training input data and training output data to train the LSTM model, and mark the trained LSTM model as a special situation prediction model.

[0101] To improve the accuracy of the special situation prediction model for toll station special situations, sufficient samples should be extracted from the historical special situation severity levels of toll stations to train the LSTM model. This will enhance the reliability of the special situation prediction model. Deploying this special situation prediction model at toll stations can accurately predict special situations, enabling dynamic staffing and ensuring timely handling. In other preferred embodiments, the special situation severity level can be replaced with a special situation severity score.

[0102] The LSTM model consists of an input layer, an LSTM layer, and an output layer. The input layer concatenates the various feature data in the training input data into a multidimensional vector. The LSTM layer can be a single-layer LSTM or a multi-layer LSTM. The output layer outputs the prediction results. It should be noted that the special situation prediction model can also be built based on other AI models, such as BiLSTM models and regression models.

[0103] The format of training input data is [number of samples, time step, feature parameters]. Training input data primarily includes the severity level of a special event. The number of samples and time step can be determined based on model training resources and special event prediction requirements. If model training resources are sufficient, as many training samples as possible can be extracted from historical special event severity levels. If hourly special events need to be predicted, the time step can be one hour, and the time step can be 24 hours.

[0104] Characteristic parameters primarily refer to parameters related to special situations, including traffic characteristics, equipment status characteristics, and environmental characteristics. Traffic characteristics include total vehicle volume, ETC vehicle ratio, truck / passenger bus ratio, and peak load factor. Equipment status characteristics include ETC antenna success rate, license plate recognition accuracy, and barrier lift success rate. Environmental characteristics include weather conditions, temperature, humidity, visibility, and time codes (e.g., whether it is a holiday).

[0105] The input data is the time series features corresponding to a sliding window, specifically the time series features of [tn, t-1], where n is the window size. The output data is the target variable to be predicted at a future time point or time period, denoted as t or [t, t+k], where k is the length of the prediction period.

[0106] It's important to note that when training the LSTM model using training input and output data, model-fitting processing, such as normalization and encoding, is required to ensure the data is recognizable by the LSTM model. Furthermore, because the output data includes the severity level of special situations, it's necessary to manually label the special situations at each time step in the historical severity level. Specifically, the severity level must be labeled according to the staff member in need, allowing for rapid staffing after the severity level is predicted.

[0107] As an example, not a limitation, when training the model, the severity level of special situations requires manual labeling. This labeling is based on the severity level of the special situation. The severity level of the special situation is a comprehensive evaluation result, that is, a comprehensive evaluation of all special situations occurring at the same time or time period. For example, if the special situation is minor, the severity level is set to level 1; if the special situation is moderate, the severity level is set to level 2; and if the special situation is severe, the severity level is set to level 3.

[0108] It's worth noting that the severity level of the emergency situation output by the emergency prediction model is primarily used to allocate staff to ensure sufficient personnel are available to handle emergency situations at the appropriate time. When an emergency situation occurs, the allocated staff will still use the emergency dispatch module to identify the emergency situation based on the abnormal state and handle it.

[0109] When predicting the severity level of a special situation at a toll station, the special situation prediction model is called; prediction input data is extracted from the historical severity levels of special situations at the toll station, and the prediction input data is input into the special situation prediction model after being subjected to model adaptation processing to obtain the severity level of the special situation.

[0110] After obtaining the severity level of the special situation at the toll station, staff will be assigned according to the severity level, including:

[0111] Extract the level-number association table corresponding to the toll station; match the special situation severity level in the level-number association table to determine the number of staff required.

[0112] As an example and not a limitation, assuming that the severity level of the special situation at a toll station is general, the staff configured when the level is general are matched from the level-number association table corresponding to the toll station, and the staff can be dispatched according to the matching results.

[0113] The construction of the level-number association table includes: extracting the number of staff responsible for handling special situations in the toll station; extracting the level range of the severity level of the special situation; dividing the level range into several levels, matching the number of associated staff for each level, and generating the level-number association table of the toll station.

[0114] As an example and not a limitation, suppose a toll station has five staff members responsible for handling special situations. If the severity of the special situation is mild, one staff member will be assigned to handle it; if the severity is average, three staff members will be assigned to handle it; if the severity is severe, five staff members will be assigned to handle it.

[0115] Example 2:

[0116] The above-mentioned special situation prediction model is trained based on the historical special situation severity levels of toll stations (entrances or exits). Therefore, it is only applicable to the corresponding toll stations. At other toll stations, it may be affected by the geographical environment, resulting in inaccurate special situation prediction results and cannot be deployed at all toll stations. This requires that each toll station needs to independently train and update the special situation prediction model, which places high demands on training resources.

[0117] To ensure that the special situation prediction model can be deployed at every toll station, its generalization capability must be guaranteed. This means that a large amount of data is required to train the LSTM model. Therefore, the historical special situation severity levels should be for all toll stations within the deployment range of the special situation prediction model. However, due to the large data coverage, the regional specificity is somewhat weakened, which can lead to deviations in the special situation prediction results for each toll station when the special situation prediction model is deployed. Considering both training efficiency and model prediction accuracy, the model is first trained with a large number of historical special situation severity levels for each toll station, and then refined based on the associated entry features of each toll station. This balances training efficiency and model prediction accuracy.

[0118] The training process of the special situation prediction model in this embodiment is as follows:

[0119] Obtain historical special situation information of several toll stations of the same type, extract training input data and training output data from the historical special situation information; train the LSTM model after modeling the training input data and training output data; mark the trained LSTM model as a special situation prediction model.

[0120] Several toll stations of the same type refer to several entry toll stations or several exit toll stations. The special situations that occur at entry and exit toll stations are different, and the corresponding influencing factors are also different. Therefore, toll stations are divided into entry and exit, and special situation prediction models are trained separately. Because the special situation prediction model in Example 1 is trained and constructed based on the historical special situation data of the toll station itself, it does not distinguish between toll station types.

[0121] The special situation prediction model is deployed in the cloud or at the corresponding toll booth. When deployed in the cloud, toll booths can upload their historical special situation data and use the model to predict the severity of special situations in the future. When deployed at toll booths, the model can be used at any time to predict historical special situation data and determine the corresponding severity level.

[0122] For entrances, since they are generally unaffected by other entrances or exits, staff can be dispatched directly based on their corresponding severity level. However, for exits, they are affected to some extent by the associated entrances, so the severity level of the exit needs to be adjusted.

[0123] The predicted severity level of the special situation has been revised, including:

[0124] Corrected input data is extracted from the associated entrance features corresponding to the toll station, and the corrected input data is processed into a suitable model and then input into the special situation correction model to obtain corrected data; the special situation severity level is corrected based on the corrected data to obtain a corrected special situation severity level.

[0125] The characteristics of associated entrances mainly include route distance, travel time, and traffic flow transfer rate. Route distance refers to the driving distance between the entrance and exit, which is used to reflect the spatial correlation strength of traffic flow. Travel time refers to the average driving time from the entrance to the exit, which is mainly affected by factors such as road conditions and speed limits. Traffic flow transfer rate refers to the proportion of traffic flow from an entrance to exit through the exit, which is used to reflect the direct connection between the entrance and exit.

[0126] The associated entry refers to the entry that will affect the special situation of the toll station. If there are multiple associated entries, the corrected input data includes multiple samples.

[0127] After the special situation prediction model is deployed at a toll station, it is used to analyze the station's historical special situation severity levels to determine the special situation severity level. The predicted special situation severity level is compared with the actual special situation severity level to calculate the special situation level error. The special situation level errors are then arranged by time to form a level error sequence.

[0128] At the same time, the associated entrance features corresponding to the toll station are obtained. The associated entrance features mainly include route distance, travel time, traffic transfer rate, etc. The associated entrance features and the corresponding special situation level error are used to train the LSTM model to obtain the special situation correction model.

[0129] For example, the special situation prediction model predicts the severity levels of special situations at toll stations as [YD1, YD2, …, YDn], the manually labeled actual special situation severity levels are [ZD1, ZD2, …, ZDn], and the level error sequence of special situation severity levels is [PD1, PD2, …, PDn]. The level error PDn (n represents the time step) is used as the output of the LSTM model, and the associated entry features of the several time steps before the level error are used as the input of the LSTM model. The trained LSTM model is labeled as the special situation correction model, and the special situation correction model is deployed at the corresponding toll station.

[0130] When predicting the special situation at a toll station, the severity level of the special situation at the toll station is first predicted using the special situation prediction model, and then the special situation correction model is used to predict the special situation level error of the toll station. The special situation severity level is corrected using the special situation level error to obtain the final special situation severity level.

[0131] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A highway toll collection robot system, characterized in that: Including charging control module and special situation dispatch module; Toll collection control module: used to identify vehicle information and complete highway toll collection based on the vehicle information; and to monitor abnormal conditions during the highway toll collection process; wherein the vehicle information is used to calculate the toll; Special situation dispatch module: used to identify the special situation corresponding to the abnormal state and dispatch staff to perform troubleshooting based on the special situation; and Used to train a special situation prediction model; predict the severity level of the special situation based on the special situation prediction model, and allocate staff according to the predicted severity level of the special situation; wherein the special situation prediction model is constructed based on an artificial intelligence model.

2. A highway toll collection robot system according to claim 1, characterized in that: Predicting the severity of a special situation based on the special situation prediction model includes: Calling the special situation prediction model; wherein the special situation prediction model is trained and obtained based on historical special situation data of a single toll station; Extract prediction input data from the historical severity levels of special situations at toll stations, perform model-adaptive processing on the prediction input data, and then input it into the special situation prediction model to obtain the severity level of the special situation.

3. A highway toll collection robot system according to claim 2, characterized in that: Training the special situation prediction model includes: Obtaining historical special situation information of a toll station, and extracting training input data and training output data from the historical special situation information; wherein the training input data includes traffic characteristics, equipment status characteristics, and environmental characteristics, and the training output data includes the severity level of the special situation; After the training input data and the training output data are subjected to model-adaptive processing, an LSTM model is trained; and the trained LSTM model is marked as a special situation prediction model.

4. A highway toll collection robot system according to claim 3, characterized in that: Predicting the severity of a special situation based on the special situation prediction model includes: Calling the special situation prediction model; wherein the special situation prediction model is trained and obtained based on historical special situation data of several toll stations of the same type; Extracting prediction input data from the historical severity levels of special situations at toll stations, performing model-based processing on the prediction input data and then inputting it into the special situation prediction model to obtain the severity level of the special situation; Determine whether the toll station is an entrance; if yes, assign staff according to the predicted severity level of the special situation; if no, revise the predicted severity level of the special situation.

5. The highway toll collection robot system according to claim 2, characterized in that: Training the special situation prediction model includes: Obtaining historical special situation information of a number of similar toll stations, and extracting training input data and training output data from the historical special situation information; wherein the training input data includes traffic characteristics, equipment status characteristics, and environmental characteristics, and the training output data includes the severity level of the special situation; After the training input data and the training output data are subjected to model-adaptive processing, an LSTM model is trained; and the trained LSTM model is marked as a special situation prediction model.

6. A highway toll collection robot system according to claim 5, characterized in that: The special situation prediction model is deployed in the cloud or in a toll station of a corresponding type; wherein the toll station type includes an entrance or an exit.

7. The highway toll collection robot system according to claim 5, characterized in that: The predicted severity level of the special situation has been revised, including: Extracting corrected input data from the associated entrance features corresponding to the toll station, performing adaptive modeling on the corrected input data and then inputting it into the special situation correction model to obtain corrected data; The special situation severity level is corrected based on the correction data to obtain a corrected special situation severity level.

8. A highway toll collection robot system according to claim 7, characterized in that: Training the special situation correction model includes: Extract the associated entrance features corresponding to the toll station; the associated entrance features include route distance, travel time, and traffic transfer rate; Extracting the prediction residual of the special situation prediction model for the toll station and using the prediction residual as a supervision label; matching the supervision label corresponding to the associated entrance feature according to the set window; The associated entry features are used as input and the matched supervision labels are used as output to train the LSTM model to obtain a special situation correction model.

9. A highway toll collection robot system according to claim 4 or 8, characterized in that: Staffing is based on the predicted severity of the situation, including: Extract the level-number association table corresponding to the toll station; the level-number association table corresponds to the toll station one-to-one, and the severity level of the special situation in the level-number association table corresponds to the number of staff; The special situation severity levels in the special situation severity level are matched in the level-number association table to determine the number of staff that need to be deployed.

10. A highway toll collection robot system according to claim 9, characterized in that: The construction of the level-number association table includes: Extract the number of staff responsible for handling special situations at the toll station; Extract the level range of the severity level of the special situation; divide the level range into several levels, match the number of associated staff for each level, and generate a level-number association table for the toll station.

Images