Automatic highway toll deduction method and system based on credit pre-authorization

Abstract

The invention relates to the technical field of expressway electronic toll collection, and discloses an expressway toll automatic deduction method and system based on credit pre-authorization, and the method comprises the following steps: carrying out the sub-region division of an expressway network topological graph based on a graph partitioning algorithm, carrying out equivalence class aggregation on reachable paths starting from each entrance portal according to sub-regions through a sequence, calculating an expected cost value and a cost variance for each path cluster, and generating a compressed path cluster graph; and performing type affiliation analysis on the candidate formation cluster, and when no single type proportion in the cluster reaches an effective formation confirmation threshold and a mixed formation condition is satisfied, confirming the candidate formation cluster as a mixed formation, and generating a mixed formation record. According to the invention, independent grouping pre-authorization freezing, independent portal adjustment and independent exit settlement processes of each type enable each type to enjoy formation-level optimization and maintain independent expense accounting and timely fund release at the same time.

Description

Technical Field

[0001] This invention relates to the field of electronic toll collection technology for highways, and more specifically, to a method and system for automatic toll collection for highways based on credit pre-authorization. Background Technology

[0002] In dense highway networks within urban agglomerations, vehicles of different types, due to their similar entry times, are grouped into the same candidate platoon cluster in temporal density clustering. However, no single type of vehicle within the cluster reaches the effective platoon confirmation threshold. Existing technology treats such candidate platoon clusters as non-platooning states, and all vehicles revert to an independent pre-authorization mode: each vehicle independently enumerates all reachable paths at the entrance gantry and freezes the pre-authorization amount according to the maximum possible cost per vehicle. At each intermediate gantry, all reachable paths are traversed for each vehicle, Bayesian posterior updates are performed, and the frozen amount is adjusted. At the exit, the actual cost is calculated for each vehicle, and the pre-authorization amount is deducted and unfrozen.

[0003] The above approach presents two technical problems: First, vehicles in a mixed platoon share the same entrance gantry. Existing technology, failing to meet the threshold for the proportion of a single type, reverts all vehicles to single-vehicle mode, thus missing the opportunity to optimize platoon-level cost uncertainty through averaging. Even when allocating the estimated total cost of the platoon based on the proportion of vehicles, the differences in route selection preferences among different types are ignored, leading to a systematic deviation between the frozen amount and the actual cost. Second, in dense road networks, the number of reachable paths originating from the same entrance can reach thousands or even tens of thousands. If each vehicle in the platoon independently performs full path probability calculations and gantry updates, it will cause severe computational delays and a large amount of redundant computation during peak hours. Summary of the Invention

[0004] This invention provides a method and system for automatic deduction of highway tolls based on credit pre-authorization, which solves the technical problems in related technologies such as the inability to calculate differentiated pre-authorization amounts for different types of subgroups in mixed formation scenarios, insufficient accuracy of path probability estimation, and the lack of reasonable modeling of the impact of formation size on the pre-authorization safety factor.

[0005] This invention provides a method for automatic deduction of highway tolls based on credit pre-authorization, comprising the following steps: Based on the graph partitioning algorithm, the topology of the highway network is divided into sub-regions. The reachable paths starting from each entrance gantry are aggregated into equivalence classes according to the sub-regions. The expected cost and cost variance of each path cluster are calculated to generate a compressed path cluster graph. Perform type classification analysis on candidate formation clusters. When the proportion of no single type in a cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. Based on the entry gantry identifier in the hybrid formation record, the corresponding path cluster set and cost statistics are extracted from the compressed path cluster map to generate basic data of the shared path cluster in the formation. Calculate the path cluster-level probability preference vector for each type of subgroup within the mixed formation, and perform Bayesian fusion with the statistical prior probability distribution vector of the entire network to generate the path cluster-level posterior probability distribution vector for each type. Based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics for each type of path cluster, the differentiated pre-authorization target amount for each type of subgroup is calculated, where the safety factor for each type is reduced by the formation size correction. The differentiated pre-authorization target amounts of each type are aggregated into a total pre-authorization amount at the group level. Group freeze requests are then sent to the payment accounts bound to each type, generating a mixed group-level combined pre-authorization certificate.

[0006] Furthermore, the aggregation of reachable paths originating from each entrance gantry by sub-region through sequence equivalence classes includes: For each entrance gantry in the road network, traverse all reachable paths originating from that entrance gantry, extract the sub-region identifier sequence that each reachable path passes through in sequence, and group paths that pass through the same sub-region identifier sequence into the same path cluster; the sub-region passing sequence is an ordered list of identifiers of each sub-region that a reachable path passes through in sequence from the entrance gantry. Using the sub-region identifier sequence as the path cluster identifier, the expected cost and cost variance of all paths within each path cluster are calculated, and the set of exit gantry reached by the paths within the cluster is calculated. The above information is used as the path cluster attribute. Wherein, the expected cost of the path cluster is the arithmetic mean of the historical tolls of all paths within the path cluster, and the cost variance of the path cluster is a measure of the dispersion of the historical tolls of each path within the path cluster relative to the expected cost.

[0007] Furthermore, the step of confirming the candidate formation cluster as a mixed formation when no single type within the cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met includes: Obtain the type affiliation information of each vehicle within the cluster, group the vehicles according to their type affiliation, and extract the number of vehicles and the vehicle identifier list for each type subgroup; Determine whether a candidate formation cluster simultaneously meets the following two conditions: the cluster contains at least two different types, and the total number of vehicles in the cluster exceeds the minimum size threshold for mixed formations. When both conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. The mixed formation record includes a formation identifier, a list of members of each type of subgroup, an entrance gantry identifier, and an entry time range. The effective formation confirmation threshold is the lower limit that the ratio of the number of vehicles of a single type in a candidate formation cluster to the total number of vehicles in the cluster must reach, and the minimum size threshold of the mixed formation is the minimum total number of vehicles that the mixed formation must contain.

[0008] Further, the step of calculating path cluster-level probability preference vectors for each type of subgroup within the mixed formation and performing Bayesian fusion with the network-wide statistical prior probability distribution vector to generate path cluster-level posterior probability distribution vectors for each type includes: For each type subgroup within the mixed formation, historical exit distribution data of vehicles of that type is obtained from the historical passage database. The frequency of vehicles of that type hitting each path cluster in the basic data of the shared path cluster in the historical passage is counted, and a path cluster-level probability preference vector of that type is generated. The frequency is the ratio of the number of historical records of that type actually passing through a certain path cluster to the total number of historical passage records of that type. Obtain the path cluster selection statistics of all vehicles in the entire network that depart from this entrance gantry, and generate the network-wide statistical prior probability distribution vector; where the network-wide prior probability is the ratio of the number of records of actual passage through a certain path cluster in the entire network to the total number of historical passage records of the entire network that depart from this entrance gantry; Bayesian fusion is performed on the path cluster-level probability preference vectors of each type and the statistical prior probability distribution vector of the entire network. The path cluster-level posterior probability of each type is the product of the historical data confidence weight of that type and the path cluster frequency of that type, plus the product of the complementary value of the confidence weight and the prior probability of the entire network, and then normalized. The historical data confidence weight is determined based on the sufficiency of the number of historical access records of this type. A larger value is taken when the number of historical access records exceeds the sufficiency threshold, and a smaller value is taken when the number of historical access records is less than the sufficiency threshold.

[0009] Furthermore, the calculation of differentiated pre-authorization target amounts for each type of subgroup based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics includes: For each type subgroup, calculate the expected cost of a single vehicle of that type. The expected cost is the weighted sum of the posterior probability of each path cluster and the expected cost of the corresponding path cluster. The cost variance of this type of bicycle is calculated based on the total probability variance expansion theorem. It is a weighted sum of the posterior probability of each path cluster, the cost variance of the corresponding path cluster, and the square of the expected cost value, minus the square of the estimated cost value. Obtain the number of vehicles in this type of subgroup, and calculate the differentiated pre-authorization target amount for this type as the product of the number of vehicles in this type of subgroup, the expected value of the estimated cost per vehicle for this type, plus the square root of the safety factor and the variance of the cost per vehicle for this type.

[0010] Furthermore, the safety factor is reduced through formation size correction, including: Obtain the total number of vehicles in the mixed formation, and calculate the safety factor for each type as the base safety factor in the single-vehicle mode divided by the sum of the product of the size reduction factor and the natural logarithm of the total number of vehicles in the formation; Wherein, the size reduction factor is a positive value and satisfies the constraint that the sum of the product of the size reduction factor and the natural logarithm of the total number of vehicles in the formation is positive under any effective formation size. The upper bound of the size reduction factor is determined by the maximum number of vehicles allowed in the formation in the system. When the total number of vehicles in the formation increases, the safety factor is appropriately reduced to reflect the partial hedging effect of cross-type cost deviations when the formation size increases.

[0011] Furthermore, after generating the hybrid formation-level combined pre-authorization certificate, the process also includes: When a convoy member vehicle passes through an intermediate gantry, the sub-region identifier to which the gantry belongs is obtained. Sub-region prefix matching is performed on the path cluster set in the convoy shared path cluster basic data. The probability of path clusters whose sub-region passing sequence does not match the prefix of the already passed sub-region sequence is set to zero, and the probability of the remaining valid path clusters is normalized. The filtering operation is performed only once on the convoy shared path cluster basic data, and the filtering results are shared by all types of subgroups. On the set of effective path clusters, the probability weights of the corresponding effective path clusters in the posterior probability distribution vectors of each type of path cluster are applied and normalized to obtain the path cluster-level probability distribution vectors after the gantry update for each type. For each type, a probability concentration check is performed. The cumulative probability of the top few path clusters in each type, ranked in descending order of probability, is calculated to determine whether the cumulative probability exceeds the concentration threshold. When the cumulative probability exceeds the concentration threshold, fine-grained cost calculations are performed on the intra-cluster paths for high-probability path clusters. The probability-weighted cost of the intra-cluster paths is used to approximate the cluster-level cost expectation value. For the remaining low-probability path clusters, the path cluster-level cost expectation value is used for approximation. When the cumulative probability does not exceed the concentration threshold, the path cluster-level cost expectation value is used for approximation on all valid path clusters. Based on the above processing results, the expected value of the revised estimated cost and the revised differential pre-authorization target amount for each type are calculated respectively. The revised differential pre-authorization target amount is compared with the current frozen amount to generate differential pre-authorization adjustment instructions for each type of gantry update.

[0012] Furthermore, it also includes: Based on the differentiated pre-authorization adjustment instructions after the gantry update for each type, additional freeze requests or partial unfreeze requests are sent to the corresponding type of payment accounts to ensure that the current frozen amount for each type is consistent with the revised differentiated pre-authorization target amount. When vehicles from each type of subgroup exit the highway, the actual exit gantry identification and actual travel route information for each vehicle are obtained. The actual toll is calculated for each vehicle, and the actual toll is written into the corresponding toll details table according to the vehicle type. After all vehicles in a certain type of subgroup have completed the exit toll calculation, the actual total toll for that type of subgroup is summarized, and a precise deduction instruction and a remaining frozen amount unfreezing instruction are sent to the payment account for that type, generating an independent deduction completion record for that type.

[0013] Furthermore, it also includes: Once all types of subgroups within the mixed formation have completed their billing settlements, the cost details for each type of subgroup are summarized, the total formation cost is calculated, and the deviation between the actual cost of each type and the differentiated pre-authorization target amount is calculated, generating a complete settlement record at the mixed formation level. Obtain the actual travel route information of vehicles in each type of subgroup, count the actual route cluster identifiers of each vehicle, and feed back the actual route cluster travel distribution data to the route cluster-level probability preference vector of each type according to the type affiliation. Adjust the historical travel frequency of each route cluster by incremental update. The updated route cluster-level historical travel frequency is the ratio of the cumulative number of records of a certain type of vehicle actually passing through a certain route cluster to the total number of cumulative historical records of that type. Based on the deviation between the actual cost of each type and the differentiated pre-authorization target amount, adjust the scale reduction factor; when the sum of the absolute values ​​of the deviations of each type exceeds the preset deviation tolerance threshold, calculate the signed deviation summary, which is the sum of the differences between the differentiated pre-authorization target amount and the corresponding actual cost of each type; when the signed deviation summary is positive, increase the scale reduction factor; when the signed deviation summary is negative, decrease the scale reduction factor; generate model parameter update records.

[0014] This invention also proposes an automatic toll collection system for highways based on credit pre-authorization, comprising: The path cluster compression module is used to divide the highway network topology map into sub-regions based on the graph partitioning algorithm. It aggregates the reachable paths from each entrance gantry into sub-regions by sequence through equivalence classes, calculates the expected cost and cost variance of each path cluster, and generates a compressed path cluster graph. The mixed formation identification module is used to perform type classification analysis on candidate formation clusters. When the proportion of no single type in the cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. The shared data extraction module is used to extract the corresponding path cluster set and cost statistics from the compressed path cluster map based on the entry gantry identifier in the mixed formation record, and generate basic data of the formation shared path cluster. The probability fusion module is used to calculate the path cluster-level probability preference vector for each type of subgroup in the mixed formation, and perform Bayesian fusion with the statistical prior probability distribution vector of the whole network to generate the path cluster-level posterior probability distribution vector for each type. The differentiated pre-authorization calculation module is used to calculate the differentiated pre-authorization target amount for each type of subgroup based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics. The safety factor for each type is reduced by the formation size correction. The pre-authorization certificate generation module is used to aggregate the differentiated pre-authorization target amounts of various types into a total pre-authorization amount at the group level, send group freeze requests to the payment accounts bound to each type respectively, and generate mixed group-level combined pre-authorization certificates.

[0015] The beneficial effects of this invention are as follows: This invention reduces the data dimension of platoon-level cost probability calculation from the original number of paths to the number of path clusters by dividing the highway network topology map into sub-regions and aggregating reachable paths into path clusters. Furthermore, the path cluster filtering operation only needs to be performed once on the shared entry data, which reduces the redundant calculations caused by performing full path probability calculations independently for each vehicle in mixed platoons.

[0016] This invention calculates the path cluster-level posterior probability distribution vector for each type of subgroup in a mixed formation, so that the frozen amount for each type matches its actual cost distribution characteristics, avoiding the freezing deviation between types caused by coarse-grained allocation based on the proportion of vehicle numbers. Meanwhile, by introducing a safety factor to correct the formation size, the total pre-authorized frozen amount of the mixed formation is reduced relative to the sum of the amounts frozen by each vehicle independently in single-vehicle mode. This allows the mixed formation, which was originally excluded from the formation optimization scope because it did not meet the single-type proportion threshold, to achieve the optimization effect of averaging the uncertainty of formation-level costs. In addition, the independent pre-authorization freeze for each type of group, the independent gantry adjustment, and the independent export settlement process enable each type to enjoy fleet-level optimization while maintaining independent cost accounting and timely fund release. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for automatic deduction of highway tolls based on credit pre-authorization according to the present invention; Figure 2 This is a diagram showing the composition and distribution of vehicle types in a mixed formation, as described in an example of the present invention. Figure 3This is a statistical graph showing the expected value and variance of path cluster costs in an example of the present invention; Figure 4 This is a posterior probability distribution diagram of various path clusters after Bayesian fusion in an example of the present invention; Figure 5 This is a comparison chart of the different types of differentiated pre-authorization target amounts S500 and S700 in the examples of the present invention; Figure 6 This is a comparison diagram of the probability distribution of path cluster of type A before and after gantry update in an example of the present invention; Figure 7 This is a scatter plot comparing the expected value and variance of various types of costs before and after gantry replacement in an example of the present invention; Figure 8 This is a structural block diagram of a highway toll automatic deduction system based on credit pre-authorization according to the present invention. Detailed Implementation

[0018] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0019] At least one embodiment of the present invention discloses a method for automatic deduction of highway tolls based on credit pre-authorization, such as... Figure 1 As shown, it includes the following steps: S100: Based on the graph partitioning algorithm, the topology graph of the highway network is divided into sub-regions, and reachable paths are aggregated by sequence through equivalence class aggregation of sub-regions to generate a compressed path cluster graph; Offline preprocessing is performed on the highway network topology map, and a graph partitioning algorithm is used to divide the network into multiple sub-regions, where each sub-region contains a set of geographically adjacent gantry nodes. The input of the graph partitioning algorithm is the highway network topology map, and the output is the sub-region affiliation label of each gantry node.

[0020] For each entrance gantry in the road network, traverse all reachable paths originating from that gantry, extract the sub-region identifier sequence that each reachable path sequentially passes through, and group paths passing through the same sub-region identifier sequence into the same path cluster. Using the sub-region identifier sequence as the path cluster identifier, calculate the expected cost and cost variance of all paths within each path cluster, and statistically analyze the set of exit gantries reached by the paths within the cluster. Use this information as the path cluster attribute. Organize all path clusters corresponding to all entrance gantries and their path cluster attributes into a compressed path cluster graph.

[0021] It should be noted that the above-mentioned sub-region sequence refers to an ordered list of identifiers of each sub-region that a reachable path passes through sequentially from the entrance gantry.

[0022] For example, a line from the entrance gantry Start, and pass through the sub-regions in sequence. , , Arrival at the exit gantry The path, whose subregions are sequenced as Through the same sub-region sequence All paths, regardless of whether they pass through the same gantry nodes within each sub-region, are grouped into the same path cluster. For the entrance gantry, , , For sub-region identification, For export gantry.

[0023] It should be noted that the expected cost of the above path clusters and cost variance The calculation method is as follows: for path clusters All paths contained within Obtain historical toll records for each route, and calculate the mean and variance of the toll. For path cluster identification, For path clusters The number of paths contained within. , , Path clusters The first Article, No. Article, No. A path. Specifically, define the path. The historical average toll fee is ,in If the path is the index within the cluster, then the path cluster... Expected cost and cost variance They are respectively: ; ; in, For path clusters Expected cost For path clusters The cost variance, For path clusters The number of paths contained within. For path clusters Inner The historical average toll for each route.

[0024] S200: Perform type classification analysis on candidate formation clusters, identify candidate formation clusters that meet the mixed formation conditions as mixed formations, and generate mixed formation records; For candidate formation clusters generated by temporal density clustering, it is determined whether a single type of vehicle within the cluster reaches the valid formation confirmation threshold. When no single type within the cluster reaches the valid formation confirmation threshold, the type affiliation information of each vehicle within the cluster is obtained, and the vehicles are grouped according to type affiliation. The number of vehicles and vehicle identifiers for each type subgroup are extracted. It is then determined whether the candidate formation cluster simultaneously meets the following two conditions: the cluster contains at least two different types, and the total number of vehicles within the cluster exceeds the minimum size threshold for mixed formation. When both conditions are met, the candidate formation cluster is confirmed as a mixed formation, and a mixed formation record is generated. The mixed formation record includes the formation identifier, the list of members of each type subgroup, the entrance gantry identifier, and the entry time range.

[0025] It should be noted that the aforementioned effective formation confirmation threshold refers to the lower limit that the ratio of the number of vehicles of a single type to the total number of vehicles in a candidate formation cluster must reach. This ratio ranges from [value missing]. When the proportion of all types within a cluster is lower than the effective formation confirmation threshold, it indicates that the candidate formation cluster does not meet the confirmation conditions of a traditional single-type formation. The aforementioned minimum size threshold for mixed formations refers to the minimum total number of vehicles required in a mixed formation, used to exclude candidate formation clusters with too few vehicles to have formation-level statistical significance.

[0026] S300: Based on the entry gantry identifier in the hybrid formation record, extract the corresponding path cluster set and cost statistics from the compressed path cluster map to generate basic data of the shared path cluster in the formation. Retrieve the entrance gantry identifier from the mixed platoon record. Using the entrance gantry identifier as an index, retrieve all path clusters corresponding to that entrance gantry from the compressed path cluster map. Extract the path cluster identifier, expected cost, cost variance, and exit gantry set for each path cluster, and organize them into a path cluster-level cost statistics table. Use the path cluster set and path cluster-level cost statistics table as basic path cluster data shared by all vehicles in the platoon, generate platoon-shared path cluster basic data, and associate it with the platoon identifier in the mixed platoon record.

[0027] S400: Calculate the path cluster-level probability preference vector for each type of subgroup in the mixed formation, and perform Bayesian fusion with the statistical prior of the whole network to generate the path cluster-level posterior probability distribution vector for each type. For each type subgroup within the mixed formation, historical exit distribution data for that type of vehicle is retrieved from the historical traffic database, and this historical exit distribution is mapped to the path cluster space. Specifically, the frequency with which that type of vehicle hits each path cluster in the formation's shared path cluster base data during historical traffic is counted, the type-specific probability weight of each path cluster is calculated, and a path cluster-level probability preference vector for that type is generated. ,in For type identification, The total number of path clusters, For type Vehicles select route clusters in historical traffic. frequency, The path cluster identifier is calculated as follows: Let the type... The total number of historical access records is The actual path cluster The number of records is ,but: ; in, For type Vehicles select route clusters in historical traffic. frequency, For type Actual path cluster The number of historical records, For type Total number of historical access records.

[0028] Obtain statistical data on path cluster selections for all vehicles departing from this entrance gantry across the entire network, and generate a network-wide statistical prior probability distribution vector. ,in Select route clusters for all vehicles on the network The prior probability is calculated as follows: Let the total number of historical passage records originating from this entrance gantry be . The actual path cluster The number of records is ,but: ; in, Select route clusters for all vehicles on the network The prior probability, For the actual path clusters passed through the entire network The number of records, This represents the total number of historical passage records originating from this entrance gantry across the entire network.

[0029] For each type of path cluster-level probability preference vector Compared with the prior probability distribution vector of the entire network Perform Bayesian fusion to calculate the posterior probability distribution vector of each path cluster level. The fusion formula is: ; in, For type For path clusters The posterior probability, For type Historical data confidence weights , To sum the index variables of the traversal path cluster, This represents the total number of path clusters.

[0030] It should be noted that the confidence weights of the historical data mentioned above... The value is determined by the type. The sufficiency of the number of historical records. When type When the number of historical passage records originating from this entrance gantry exceeds the sufficiency threshold. Take a larger value; in this case, the path cluster-level posterior probability distribution vector is dominated by the path cluster-level probability preference vector. When the number of historical passage records is lower than the sufficiency threshold... Take the smaller value, at which point the path cluster-level posterior probability distribution vector is dominated by the network-wide statistical prior probability distribution vector.

[0031] S500: Based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics table for each type of subgroup, calculate the differentiated pre-authorization target amount for each type of subgroup; For each type subgroup Based on the path cluster-level posterior probability distribution vector of this type By sharing the path cluster-level cost statistics table with the formation's basic path cluster data, the expected value of the estimated cost for this type of vehicle can be calculated. and cost variance : ; ; in, For type The expected cost per bicycle. For type The cost variance of a single bicycle For type For path clusters The posterior probability, For path clusters Expected cost For path clusters The cost variance, The total number of path clusters, This serves as the identifier for the path cluster.

[0032] Furthermore, the above The calculation formula is based on the law of total probability variance expansion: for a discrete mixture distribution, the total variance equals the weighted sum of the variances of each component plus the weighted second moment of the mean of each component minus the square of the total mean, i.e. The first two items combined are: Finally subtract By eliminating the squared term of the overall mean, we obtain the type. The total variance of single-vehicle costs under a mixed distribution of path clusters.

[0033] Get the number of vehicles in this type of subgroup Calculation type Estimated total cost and differentiated pre-authorization target amount : ; in, For type Differentiated pre-authorization target amount, For type Number of vehicles in the subgroup For type The expected cost per bicycle. For type The cost variance of a single bicycle For type The safety factor is used to add a safety margin for cost fluctuations to the estimated expected cost value. .

[0034] Furthermore, to leverage the scale effect of mixed formations to reduce the total pre-authorized frozen amount, a formation size correction is introduced when calculating the safety factor for each type. Specifically, the total number of vehicles in the mixed formation is obtained. The safety factor for each type is adjusted according to the following formula: ; in, For type The safety factor, This serves as the baseline safety factor in single-vehicle mode. For size reduction factor, This refers to the total number of vehicles in the formation. , This represents the natural logarithm function. When the total number of vehicles in the platoon... When it increases, The safety factor should be appropriately reduced to reflect the partial offsetting effect of cross-type cost deviations as the formation size increases. Since all types share the same entrance gantry and there is spatial correlation in path selection, there is a probability that the cost deviation directions of different types will partially offset each other. The larger the formation size, the more significant this offsetting effect will be. Therefore, the reduction of the safety factor has a statistical basis.

[0035] Furthermore, the aforementioned size reduction factor Must meet and To ensure At any effective formation size All values ​​below are positive, thus ensuring differentiated pre-authorization target amounts. The estimated cost should never be lower than the expected value for each type of cost. .

[0036] Furthermore, the above The constraints are Time is equivalent to ,in The maximum number of vehicles allowed in a platoon in the system, i.e., the size reduction factor. The upper bound of the value is determined by the maximum formation size in the actual deployment scenario, to ensure that under any legal formation size... All remain positive.

[0037] S600: Aggregate the differentiated pre-authorization target amounts of each type into a total pre-authorization amount at the group level, send group freeze requests to the payment accounts bound to each type, and generate mixed group-level combined pre-authorization vouchers; Differentiated pre-authorization target amounts for each type of subgroup Summarize and calculate the total pre-authorization amount at the formation level. ,in For the total pre-authorized amount at the formation level, For type Differentiated pre-authorization target amount, This is a type identifier. Pre-authorization freeze requests for the corresponding amount are sent to the payment accounts bound to each type, where the type... The frozen amount is Receive freeze credentials returned by various types of payment accounts, associate the grouped freeze credentials of each type with the formation identifier in the mixed formation record, and generate a mixed formation-level combined pre-authorization credential.

[0038] It should be noted that the payment accounts linked to the above-mentioned types refer to the unified payment account associated with vehicles in each type of subgroup within the mixed formation, or the payment account of the entity to which the type belongs. The frozen amount for each type is independent of the others, and the freezing operation is performed in the corresponding payment account.

[0039] S700: When a convoy member vehicle passes through an intermediate gantry, it performs a sub-region prefix matching filter based on the convoy shared path cluster basic data, updates the path cluster-level probability distribution vectors of each type, and generates differentiated pre-authorization adjustment instructions after each type of gantry update. When a convoy member vehicle passes through an intermediate gantry, the sub-region identifier to which that gantry belongs is obtained. Using the sub-region identifier as the matching condition, a sub-region prefix matching filter is performed on the path cluster set in the convoy's shared path cluster base data. This involves checking whether the sub-region passage sequence of each path cluster has a consistent prefix with the sequence of already passed sub-regions, setting the probability of mismatched path clusters to zero, and normalizing the probability of the remaining valid path clusters to obtain an updated set of valid path clusters. Since all members within the convoy share the same entrance gantry, the above filtering operation only needs to be performed once on the convoy's shared path cluster base data, and the filtering results are shared across different types of subgroups.

[0040] On the set of effective path clusters, the probability weights of the corresponding effective path clusters in the posterior probability distribution vectors of each type of path cluster are applied to normalize each type, so as to obtain the path cluster-level probability distribution vectors after the gantry update for each type.

[0041] Check the probability concentration for each type separately: Calculate the top [number] values ​​in each type, ranked in descending order of probability. The cumulative probability of each path cluster is used to determine whether the cumulative probability exceeds the concentration threshold. ,in The number of path clusters used for concentration checks. .

[0042] When the cumulative probability exceeds the concentration threshold At that time, towards the front For each high-probability path cluster, a detailed cost calculation is performed within the cluster—the precise cost value of each path within the high-probability path cluster is obtained, and the probability-weighted cost of the paths within the cluster is used as an approximation to replace the expected cost of the cluster-level cost; for the remaining low-probability path clusters, the expected cost of the path cluster level is used for approximation. This is done when the cumulative probability does not exceed the concentration threshold. At that time, the expected cost of the path cluster is used to approximate all valid path clusters.

[0043] Based on the above processing results, the expected value of the revised estimated cost and the revised differential pre-authorization target amount for each type are calculated respectively. The revised differential pre-authorization target amount is compared with the current frozen amount to generate differential pre-authorization adjustment instructions for each type of gantry update.

[0044] It should be noted that the above concentration thresholds This is used to determine whether the probability distribution of a path cluster has converged to at least a few path clusters. When the probability distribution is highly concentrated, performing fine-grained calculations within high-probability path clusters can yield more accurate cost estimates; when the probability distribution is relatively dispersed, the expected cost at the path cluster level already has sufficient estimation accuracy, and there is no need to perform fine-grained calculations within the cluster. This is the parameter for the number of path clusters used for concentration checks.

[0045] S800: Executes frozen amount adjustment according to the differentiated pre-authorization adjustment instructions of each type, and calculates the actual cost for each vehicle when the vehicles of each type of subgroup leave the highway, generating independent deduction completion records for each type; Based on the differentiated pre-authorization adjustment instructions after the gantry update for each type, additional freeze requests or partial unfreeze requests are sent to the corresponding type of payment accounts to ensure that the current frozen amount for each type is consistent with the revised differentiated pre-authorization target amount.

[0046] As vehicles from different subgroups exit the highway, the system obtains the actual exit gantry identification and actual travel route information for each vehicle. The actual toll for each vehicle is then precisely calculated, and the toll is recorded in the corresponding toll details table based on the vehicle type. Once all vehicles in a certain subgroup have completed their exit toll calculations, the system aggregates the total actual toll for that subgroup, sends a precise deduction instruction and a remaining frozen amount unfreezing instruction to the payment account for that type, and generates an independent toll completion record for that type.

[0047] S900: After all types of subgroups in the mixed formation have completed the toll settlement, a complete settlement record at the mixed formation level is generated, and the actual passage data is fed back to update the path cluster-level probability preference vector parameters and scale reduction factor parameters, generating a model parameter update record; Once all subgroups within the mixed formation have completed their billing settlements, the cost details for each subgroup are summarized. The total formation cost and the deviation between the actual cost for each type and the differentiated pre-authorized target amount are calculated, generating a complete mixed formation-level settlement record. This record includes the formation identifier, cost details for each subgroup type, the total formation cost, and deviation analysis data for each type. Based on this complete mixed formation-level settlement record, separate reconciliation reports for each type and a summary settlement report for formation management are output.

[0048] The actual travel path information of vehicles in each type of subgroup in the mixed formation is obtained, and the path cluster identifiers actually traversed by each vehicle are counted to generate actual path cluster travel distribution data for the mixed formation. This actual path cluster travel distribution data is then fed back to the path cluster-level probability preference vector of each type according to its type affiliation, and the historical travel frequency of each path cluster is adjusted using an incremental update method. Specifically, let the type... After this formation settlement, the total number of historical passage records has been updated to [number]. The actual path cluster The cumulative number of records has been updated to The updated path cluster-level historical passage frequency for: ; in, For the updated type For path clusters Historical frequency of passage For type Actual path cluster The cumulative number of records, For type The total number of historical access records.

[0049] Meanwhile, based on the deviation between the actual costs of each type and the differentiated pre-authorization target amount, the size reduction factor in cross-type mixed formation scenarios is adjusted. Generate model parameter update records.

[0050] Specifically, let the sum of the absolute values ​​of the deviations between the actual costs of each type and the differentiated pre-authorization target amount after the mixed formation settlement be... ,like If the deviation exceeds the preset tolerance threshold, the size reduction factor will be adjusted. Execute one step size Gradient direction adjustment: When the actual deviation indicates that the current overall safety factor is too high (i.e., the frozen amount is systematically higher than the actual cost), the scale reduction factor should be appropriately increased. To further reduce the safety factor in subsequent formations; when the actual deviation indicates that the current safety factor is generally low (i.e., the frozen amount is systematically lower than the actual cost), the scale reduction factor should be appropriately reduced. To improve safety in subsequent formations, among which This is the adjustment step size for the scale reduction factor, and is a preset positive hyperparameter.

[0051] Furthermore, the aforementioned size reduction factor The direction of adjustment and The correspondence between the sign directions is as follows: Let type The differentiated pre-authorization target amount is The actual cost is Then the sum of the absolute values ​​of the deviations of each type is ,in It is the sum of the absolute values ​​of all types of deviations. For type Differentiated pre-authorization target amount, For type The actual cost; Furthermore, define the total amount of signed bias. ,in For the total signed bias, when This indicates that the frozen amount is systematically higher than the actual expenses, affecting the scale reduction factor. Execute the increase adjustment; when This indicates that the frozen amount is systematically lower than the actual expenses, affecting the scale reduction factor. Perform a reduction adjustment; only when The above adjustments are only triggered when the preset deviation tolerance threshold is exceeded, in order to avoid frequent parameter modifications due to random fluctuations in a single formation.

[0052] This implementation divides the highway network topology into sub-regions and aggregates reachable paths into path clusters according to the sequence of sub-regions. This compresses the original path space from thousands to tens of thousands of paths to tens to hundreds of path clusters. This allows all vehicles in the formation that share the same entrance gantry to share basic data and filtering calculation results at the path cluster level, instead of enumerating and traversing all reachable paths independently for each vehicle. Therefore, the data dimension of the formation-level cost probability calculation is reduced from the order of magnitude of the original paths to the order of magnitude of path clusters. Moreover, the path cluster filtering operation only needs to be performed once on the shared entrance data. Differentiated calculations for each type only need to be superimposed on the shared filtering results with type-specific probability weights, thereby reducing the redundant calculations caused by the independent execution of the full path probability calculation for each vehicle in mixed formations.

[0053] This implementation calculates path cluster-level posterior probability distribution vectors for each type of subgroup in a mixed formation, ensuring that the estimated cost calculation for each type is based on its own path cluster selection preference rather than a uniform formation-level exit probability aggregation. Therefore, types that prefer short-distance, low-cost path clusters correspond to lower expected estimated costs and lower frozen amounts, while types that prefer long-distance, high-cost path clusters correspond to higher expected estimated costs and higher frozen amounts. This matches the frozen amount for each type with its actual cost distribution characteristics, avoiding inter-type freezing bias caused by coarse-grained allocation based on vehicle quantity proportions. Simultaneously, by introducing a safety factor to correct for formation size, and utilizing the partial hedging effect of cross-type cost bias when the formation size increases, the total pre-authorized frozen amount for the mixed formation is reduced compared to the sum of the individual vehicle frozen amounts. This allows mixed formations that were originally excluded from formation optimization due to not meeting the single-type proportion threshold to achieve the optimization effect of formation-level cost uncertainty averaging.

[0054] In addition, through independent group pre-authorization freezes, independent gantry adjustments, and independent export settlement processes for each type, each type can enjoy fleet-level optimization while maintaining independent cost accounting and timely fund release.

[0055] like Figure 8 As shown, based on the above-mentioned method for automatic toll deduction of highways based on credit pre-authorization, a system for automatic toll deduction of highways based on credit pre-authorization is also proposed, including the following modules: The path cluster compression module is used to divide the highway network topology map into sub-regions based on the graph partitioning algorithm. It aggregates the reachable paths from each entrance gantry into sub-regions by sequence through equivalence classes, calculates the expected cost and cost variance of each path cluster, and generates a compressed path cluster graph. The mixed formation identification module is used to perform type classification analysis on candidate formation clusters. When the proportion of no single type in the cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. The shared data extraction module is used to extract the corresponding path cluster set and cost statistics from the compressed path cluster map based on the entry gantry identifier in the mixed formation record, and generate basic data of the formation shared path cluster. The probability fusion module is used to calculate the path cluster-level probability preference vector for each type of subgroup in the mixed formation, and perform Bayesian fusion with the statistical prior probability distribution vector of the whole network to generate the path cluster-level posterior probability distribution vector for each type. The differentiated pre-authorization calculation module is used to calculate the differentiated pre-authorization target amount for each type of subgroup based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics. The safety factor for each type is reduced by the formation size correction. The pre-authorization certificate generation module is used to aggregate the differentiated pre-authorization target amounts of various types into a total pre-authorization amount at the group level, send group freeze requests to the payment accounts bound to each type respectively, and generate mixed group-level combined pre-authorization certificates.

[0056] like Figures 2-7 As shown, the above-mentioned method and system for automatic deduction of highway tolls based on credit pre-authorization is applied to the following application scenario: A highway network toll system of a certain urban agglomeration (hereinafter referred to as "road network toll system") covers the Yangtze River Delta region and deploys a number of gantry nodes. The road network topology map has completed the sub-region division preprocessing.

[0057] July 15, 20XX, 08:23:41, Entrance gantry At this point, the temporal density clustering module detected a candidate platoon cluster with a total of 11 vehicles, including 3 types: Type A (network freight platform trucks, 5 vehicles), Type B (logistics line trucks, 4 vehicles), and Type C (construction machinery transport vehicles, 2 vehicles).

[0058] Type A accounted for 45.5%, Type B accounted for 36.4%, and Type C accounted for 18.2%, none of which reached the effective formation confirmation threshold (set at 60%). They did not meet the traditional single-type formation confirmation conditions, but met the mixed formation confirmation conditions (number of types ≥ 2, total number of vehicles 11 > minimum mixed formation size threshold 8).

[0059] The road network toll collection system initiates a mixed formation processing procedure for the candidate formation cluster, based on the entrance gantry identifier. Extract the corresponding path cluster set (S300) from the compressed path cluster graph generated by S100, and perform subsequent differentiated pre-authorization allocation.

[0060] Based on the above application scenarios, the following implementation process example is given: Implementation example of core S100: Corresponding to S100 in the specific implementation, the road network toll collection system has already performed a graph partitioning algorithm on the highway network topology map during the offline preprocessing stage, and will... The area is divided into several sub-regions.

[0061] from The reachable paths originating from the origin are aggregated by sub-regions, resulting in four path clusters, each named after its sub-region sequence. , , , .

[0062] The cost statistics for each path cluster are as follows: With path clusters For example, it includes There are 10 routes, and the average historical toll for each route is as follows: Yuan, Yuan, Yuan, of which , , Path clusters The average historical toll cost of routes 1, 2, and 3 within the route cluster is then used to determine the route cluster. Expected cost for: ; Path clusters Cost variance for: ; ; Table 1. Compression Path Cluster Diagram Corresponding path cluster basic data in, For path cluster identification, For path clusters The number of paths contained within. For path clusters Expected cost For path clusters The cost variance.

[0063] Implementation example of core S200: Corresponding to S200 in the specific implementation, the road network toll collection system performs type attribution analysis on candidate formation clusters, confirms mixed formations, and generates mixed formation records. .

[0064] Table 2 Mixed Formation Records in, This represents the total number of vehicles in the mixed platoon, used for calculating the platoon size correction safety factor in the subsequent S500. The road network toll system is based on the entrance gantry identifiers in Table 2. The corresponding path cluster set (i.e., the data in Table 1) is retrieved and extracted from the compressed path cluster map generated by S100, and associated with the formation identifier. Generate basic data for the formation shared path cluster, which can be shared by S400 to S700.

[0065] Implementation example of core S400: Corresponding to S400 in the specific implementation, the road network toll collection system extracts types A, B, and C from the historical passage database respectively. The historical records of departures are used to calculate the probability preference vectors of path clusters of each type, and then Bayesian fusion is performed with the statistical priors of the entire network.

[0066] Prior knowledge of network-wide statistics: Total number of historical records triggered across the entire network The number of records hit in each path cluster is , , , ,in For the actual path clusters passed through the entire network The number of records determines the prior probability of each path cluster across the entire network. for: ; ; ; ; Type A Total number of historical records If the sufficiency threshold is exceeded, the confidence weight will be adjusted accordingly. Type B: Total number of historical records If the sufficiency threshold is exceeded, the confidence weight will be adjusted accordingly. Type C: Total number of historical records Below the sufficiency threshold, confidence weight .in, For type Total number of historical access records For type Historical data confidence weights.

[0067] Path clusters of type A Taking the posterior probability as an example, type A actually passes through the path cluster Number of historical records Historical preference frequency , fusion calculation of posterior probability : ; molecular Since the preference vectors of each type have been normalized, the denominator Therefore .in, The index variable is used to sum and traverse the path cluster.

[0068] Table 3. Distribution of Historical Preference Frequency and Posterior Probability for Each Path Cluster in, For type Vehicles select route clusters in historical traffic. frequency, Select route clusters for all vehicles on the network The prior probability, For type For path clusters The posterior probability, , .

[0069] Implementation example of core S500: Corresponding to S500 in the specific implementation, the road network tolling system calculates the expected value of the estimated cost per vehicle for each type based on the posterior probability distribution of each type in Table 3 and the path cluster cost statistics in Table 1. With cost variance It also introduces a formation size-corrected safety factor to calculate differentiated pre-authorized target amounts. .

[0070] Taking type A as an example, the expected cost per vehicle : ; Type A Single-Bike Cost Variance : ; Formation size correction safety factor (based on baseline safety factor) Scale reduction factor Total number of vehicles in the formation ): ; in, For type The safety factor is the same for all types of formations, with each type sharing the same formation size correction safety factor. .

[0071] Type A Differentiated Pre-authorization Target Amount (Number of Type A Vehicles) ): ; Table 4 Calculation Results of Differentiated Pre-authorization Target Amounts for Each Type (S500 Initial Frozen Amount) in, For type Number of vehicles in the subgroup For type The expected cost per bicycle. For type The cost variance of a single bicycle For the standard deviation of costs, For type The safety factor, For type Differentiated pre-authorization target amount, .

[0072] Implementation example of the core S700: Corresponding to the S700 in the specific implementation, on July 15, 20XX at 08:51:17, the platoon member vehicles passed through the intermediate gantry. The gantry belongs to the sub-region. The sub-region sequence has been passed. .

[0073] Perform sub-region prefix matching filtering: Check if the sub-regions of each path cluster are prefixed with the given sequence. Used as a prefix. The sequence is If the prefix does not match, the probability is set to zero. The sequence is If the prefix does not match, the probability is set to zero. The sequence is Prefix matching; The sequence is Prefix matching. The effective path cluster is reduced to... This filtering operation is performed only once per group on the group's shared path cluster base data.

[0074] Renormalized posterior probabilities for each type on the effective path cluster: Taking type A as an example, the original posterior probability of the effective path cluster is... , After normalization: ; ; in, For type After the gantry update, the effective path cluster The normalized probability.

[0075] Concentration check (previous) Path clusters, concentration threshold ): There are only 2 valid path clusters left for type A, and the cumulative probability of the first 2 path clusters is... This triggers fine-grained cost calculation for high-probability path clusters.

[0076] Expected cost values ​​for each type of revised forecast (taking type A as an example), and the revised expected cost value. : ; Corrected cost variance : ; Adjust the differentiated pre-authorization target amount : ; Yuan, lower than the current frozen amount of S500. Yuan, generating type A partial unfreezing instruction, unfreezing amount Yuan.

[0077] Table 5. Differentiated Pre-authorization Target Amounts for Various Types of S700 Gantry Updates in, For type Expected cost of modifications following gantry upgrade For type Variance of correction costs after gantry upgrade For type The revised differential pre-authorization target amount after the gantry update .

[0078] Implementation example of the core S900: In the specific implementation of S900, at 10:38:52 on July 15, 20XX, all vehicles in the mixed formation completed the exit settlement, the road network toll system summarized the actual fees of each type, and the execution parameters were updated.

[0079] Table 6. Summary of Mixed Formation Settlement and Parameter Update Trigger Data The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for automatic deduction of highway tolls based on credit pre-authorization, characterized in that, Includes the following steps: Based on the graph partitioning algorithm, the topology of the highway network is divided into sub-regions. The reachable paths starting from each entrance gantry are aggregated into equivalence classes according to the sub-regions. The expected cost and cost variance of each path cluster are calculated to generate a compressed path cluster graph. Perform type classification analysis on candidate formation clusters. When the proportion of no single type in a cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. Based on the entry gantry identifier in the hybrid formation record, the corresponding path cluster set and cost statistics are extracted from the compressed path cluster map to generate basic data of the shared path cluster in the formation. Calculate the path cluster-level probability preference vector for each type of subgroup within the mixed formation, and perform Bayesian fusion with the statistical prior probability distribution vector of the entire network to generate the path cluster-level posterior probability distribution vector for each type. Based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics for each type of path cluster, the differentiated pre-authorization target amount for each type of subgroup is calculated, where the safety factor for each type is reduced by the formation size correction. The differentiated pre-authorization target amounts of each type are aggregated into a total pre-authorization amount at the group level. Group freeze requests are then sent to the payment accounts bound to each type, generating a mixed group-level combined pre-authorization certificate.

2. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 1, characterized in that, The aggregation of reachable paths originating from each entrance gantry by sub-region using equivalence classes includes: For each entrance gantry in the road network, traverse all reachable paths originating from that entrance gantry, extract the sub-region identifier sequence that each reachable path passes through in sequence, and group paths that pass through the same sub-region identifier sequence into the same path cluster; the sub-region passing sequence is an ordered list of identifiers of each sub-region that a reachable path passes through in sequence from the entrance gantry. Using the sub-region identifier sequence as the path cluster identifier, the expected cost and cost variance of all paths within each path cluster are calculated, and the set of exit gantry reached by the paths within the cluster is calculated. The above information is used as the path cluster attribute. Wherein, the expected cost of the path cluster is the arithmetic mean of the historical tolls of all paths within the path cluster, and the cost variance of the path cluster is a measure of the dispersion of the historical tolls of each path within the path cluster relative to the expected cost.

3. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 1, characterized in that, When no single type within a cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation, including: Obtain the type affiliation information of each vehicle within the cluster, group the vehicles according to their type affiliation, and extract the number of vehicles and the vehicle identifier list for each type subgroup; Determine whether a candidate formation cluster simultaneously meets the following two conditions: the cluster contains at least two different types, and the total number of vehicles in the cluster exceeds the minimum size threshold for mixed formations. When both conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. The mixed formation record includes a formation identifier, a list of members of each type of subgroup, an entrance gantry identifier, and an entry time range. The effective formation confirmation threshold is the lower limit that the ratio of the number of vehicles of a single type in a candidate formation cluster to the total number of vehicles in the cluster must reach, and the minimum size threshold of the mixed formation is the minimum total number of vehicles that the mixed formation must contain.

4. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 1, characterized in that, The method involves calculating path cluster-level probability preference vectors for each type of subgroup within the hybrid formation, and then performing Bayesian fusion with the network-wide statistical prior probability distribution vector to generate path cluster-level posterior probability distribution vectors for each type, including: For each type subgroup within the mixed formation, historical exit distribution data of vehicles of that type is obtained from the historical passage database. The frequency of vehicles of that type hitting each path cluster in the basic data of the shared path cluster in the historical passage is counted, and a path cluster-level probability preference vector of that type is generated. The frequency is the ratio of the number of historical records of that type actually passing through a certain path cluster to the total number of historical passage records of that type. Obtain the path cluster selection statistics of all vehicles in the entire network that depart from this entrance gantry, and generate the network-wide statistical prior probability distribution vector; where the network-wide prior probability is the ratio of the number of records of actual passage through a certain path cluster in the entire network to the total number of historical passage records of the entire network that depart from this entrance gantry; Bayesian fusion is performed on the path cluster-level probability preference vectors of each type and the statistical prior probability distribution vector of the entire network. The path cluster-level posterior probability of each type is the product of the historical data confidence weight of that type and the path cluster frequency of that type, plus the product of the complementary value of the confidence weight and the prior probability of the entire network, and then normalized. The historical data confidence weight is determined based on the sufficiency of the number of historical access records of this type. A larger value is taken when the number of historical access records exceeds the sufficiency threshold, and a smaller value is taken when the number of historical access records is less than the sufficiency threshold.

5. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 4, characterized in that, The differential pre-authorization target amount for each type of subgroup is calculated based on the posterior probability distribution vector of each path cluster and the path cluster cost statistics, including: For each type subgroup, calculate the expected cost of a single vehicle of that type. The expected cost is the weighted sum of the posterior probability of each path cluster and the expected cost of the corresponding path cluster. The cost variance of this type of bicycle is calculated based on the total probability variance expansion theorem. It is a weighted sum of the posterior probability of each path cluster, the cost variance of the corresponding path cluster, and the square of the expected cost value, minus the square of the estimated cost value. Obtain the number of vehicles in this type of subgroup, and calculate the differentiated pre-authorization target amount for this type as the product of the number of vehicles in this type of subgroup, the expected value of the estimated cost per vehicle for this type, plus the square root of the safety factor and the variance of the cost per vehicle for this type.

6. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 5, characterized in that, The safety factor is reduced through formation size correction, including: Obtain the total number of vehicles in the mixed formation, and calculate the safety factor for each type as the base safety factor in the single-vehicle mode divided by the sum of the product of the size reduction factor and the natural logarithm of the total number of vehicles in the formation; Wherein, the size reduction factor is a positive value and satisfies the constraint that the sum of the product of the size reduction factor and the natural logarithm of the total number of vehicles in the formation is positive under any effective formation size. The upper bound of the size reduction factor is determined by the maximum number of vehicles allowed in the formation in the system. When the total number of vehicles in the formation increases, the safety factor is appropriately reduced to reflect the partial hedging effect of cross-type cost deviations when the formation size increases.

7. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 1, characterized in that, After generating the hybrid formation-level combined pre-authorization certificate, the process also includes: When a convoy member vehicle passes through an intermediate gantry, the sub-region identifier to which the gantry belongs is obtained. Sub-region prefix matching is performed on the path cluster set in the convoy shared path cluster basic data. The probability of path clusters whose sub-region passing sequence does not match the prefix of the already passed sub-region sequence is set to zero, and the probability of the remaining valid path clusters is normalized. The filtering operation is performed only once on the convoy shared path cluster basic data, and the filtering results are shared by all types of subgroups. On the set of effective path clusters, the probability weights of the corresponding effective path clusters in the posterior probability distribution vectors of each type of path cluster are applied and normalized to obtain the path cluster-level probability distribution vectors after the gantry update for each type. For each type, a probability concentration check is performed. The cumulative probability of the top few path clusters in each type, ranked in descending order of probability, is calculated to determine whether the cumulative probability exceeds the concentration threshold. When the cumulative probability exceeds the concentration threshold, fine-grained cost calculations are performed on the intra-cluster paths for high-probability path clusters. The probability-weighted cost of the intra-cluster paths is used to approximate the cluster-level cost expectation value. For the remaining low-probability path clusters, the path cluster-level cost expectation value is used for approximation. When the cumulative probability does not exceed the concentration threshold, the path cluster-level cost expectation value is used for approximation on all valid path clusters. Based on the above processing results, the expected value of the revised estimated cost and the revised differential pre-authorization target amount for each type are calculated respectively. The revised differential pre-authorization target amount is compared with the current frozen amount to generate differential pre-authorization adjustment instructions for each type of gantry update.

8. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 7, characterized in that, Also includes: Based on the differentiated pre-authorization adjustment instructions after the gantry update for each type, additional freeze requests or partial unfreeze requests are sent to the corresponding type of payment accounts to ensure that the current frozen amount for each type is consistent with the revised differentiated pre-authorization target amount. When vehicles of each type of subgroup exit the highway, the actual exit gantry identifier and actual travel route information of each vehicle are obtained, the actual toll fee is calculated for each vehicle, and the actual toll fee is written into the corresponding toll fee details table according to the vehicle type. Once all vehicles in a certain type of subgroup have completed their export fee calculations, the total actual fees for that type of subgroup are aggregated, and a precise deduction instruction and a remaining frozen amount unfreezing instruction are sent to the payment account for that type, generating an independent deduction completion record for that type.

9. The method for automatic deduction of highway tolls based on credit pre-authorization according to claim 1, characterized in that, Also includes: Once all types of subgroups within the mixed formation have completed their billing settlements, the cost details for each type of subgroup are summarized, the total formation cost is calculated, and the deviation between the actual cost of each type and the differentiated pre-authorization target amount is calculated, generating a complete settlement record at the mixed formation level. Obtain the actual travel route information of vehicles in each type of subgroup, count the actual route cluster identifiers of each vehicle, and feed back the actual route cluster travel distribution data to the route cluster-level probability preference vector of each type according to the type affiliation. Adjust the historical travel frequency of each route cluster by incremental update. The updated route cluster-level historical travel frequency is the ratio of the cumulative number of records of a certain type of vehicle actually passing through a certain route cluster to the total number of cumulative historical records of that type. Based on the deviation between the actual cost of each type and the differentiated pre-authorization target amount, adjust the scale reduction factor; when the sum of the absolute values ​​of the deviations of each type exceeds the preset deviation tolerance threshold, calculate the signed deviation summary, which is the sum of the differences between the differentiated pre-authorization target amount and the corresponding actual cost of each type; when the signed deviation summary is positive, increase the scale reduction factor; when the signed deviation summary is negative, decrease the scale reduction factor; generate model parameter update records.

10. A highway toll automatic deduction system based on credit pre-authorization, characterized in that, The method for performing the steps of an automatic toll collection method for highways based on credit pre-authorization as described in any one of claims 1-9 includes: The path cluster compression module is used to divide the highway network topology map into sub-regions based on the graph partitioning algorithm. It aggregates the reachable paths from each entrance gantry into sub-regions by sequence through equivalence classes, calculates the expected cost and cost variance of each path cluster, and generates a compressed path cluster graph. The mixed formation identification module is used to perform type classification analysis on candidate formation clusters. When the proportion of no single type in the cluster reaches the effective formation confirmation threshold and the mixed formation conditions are met, the candidate formation cluster is confirmed as a mixed formation and a mixed formation record is generated. The shared data extraction module is used to extract the corresponding path cluster set and cost statistics from the compressed path cluster map based on the entry gantry identifier in the mixed formation record, and generate basic data of the formation shared path cluster. The probability fusion module is used to calculate the path cluster-level probability preference vector for each type of subgroup in the mixed formation, and perform Bayesian fusion with the statistical prior probability distribution vector of the whole network to generate the path cluster-level posterior probability distribution vector for each type. The differentiated pre-authorization calculation module is used to calculate the differentiated pre-authorization target amount for each type of subgroup based on the path cluster-level posterior probability distribution vector and path cluster-level cost statistics. The safety factor for each type is reduced by the formation size correction. The pre-authorization certificate generation module is used to aggregate the differentiated pre-authorization target amounts of various types into a total pre-authorization amount at the group level, send group freeze requests to the payment accounts bound to each type respectively, and generate mixed group-level combined pre-authorization certificates.

Images