Highway toll passage management and control system based on cloud toll architecture

Abstract

The invention discloses a highway toll passage management and control system based on a cloud toll architecture, which comprises a cloud intelligent billing platform, an edge decision node and a vehicle-mounted trusted terminal, and is characterized in that the cloud intelligent billing platform, the edge decision node and the vehicle-mounted trusted terminal are connected through signals; the cloud intelligent charging platform comprises an elastic rate engine, a privacy settlement middle station and an emergency visualization system, and the edge decision node comprises a heterogeneous communication unit, a security authentication unit vehicle and a lane optimizer. The LSTM model is combined with real-time data, the rate prediction error is smaller than or equal to 2.5%, compared with a traditional linear model (the error is about 8%), the ETC and C-V2X switching success rate is remarkably optimized to be larger than or equal to 99.5%, the passing efficiency is improved by 30%, and data anonymization processing consumes time lt through the differential privacy technology; and the block chain clearing efficiency reaches 5000 pieces per second in 100ms.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control systems, and particularly to a highway toll access control system based on a cloud toll collection architecture. Background Art

[0002] Traditional highway toll systems mostly adopt fixed rates or regional segmented billing models, which have the following defects:

[0003] Lack of flexibility in static rates: Unable to dynamically adjust prices according to real-time traffic flow, weather or accidents, resulting in increased congestion during peak periods or waste of resources;

[0004] Poor compatibility of multi-protocol devices: The switching of communication systems such as ETC and C-V2X is not intelligent, affecting the vehicle passing efficiency;

[0005] Low data privacy and clearing efficiency: Cross-provincial transactions require manual intervention, with a risk of data leakage;

[0006] Lag in emergency response: Congestion events rely on manual judgment, and the decision-making chain is too long. Summary of the Invention

[0007] The purpose of the present invention is to provide a highway toll access control system based on a cloud toll collection architecture to solve the problems raised in the above background art.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A highway toll access control system based on a cloud toll collection architecture, including a cloud intelligent charging platform, an edge decision-making node and an in-vehicle trusted terminal, and the cloud intelligent charging platform, the edge decision-making node and the in-vehicle trusted terminal are all connected by signals, wherein:

[0009] The cloud intelligent charging platform includes:

[0010] Elastic rate engine: Generates dynamic rates through the cooperation of a time series prediction model and a rule engine. The time series prediction model is a neural network containing 3 layers of LSTM units. The input features include section traffic flow Q, visibility W and accident level E, and it is trained based on historical traffic data. The mean absolute error (MAE) of the model ≤ 2.5%;

[0011] The training data of the LSTM model includes the traffic flow (unit: vehicles per minute), visibility (quantification level 1-5), accident level (1-3) in the historical traffic dataset, and the corresponding historical rate labels. The input features are processed by Z-Score standardization, and the model output is the benchmark rate P0. During the training process, the mean squared error (MSE) is used as the loss function, and the model parameters are optimized through the backpropagation algorithm. Finally, the model validation MAE ≤ 2.5%. The generation method of the historical rate label is: it is inversely deduced according to the actual traffic revenue of the road section under specific traffic flow, visibility, and accident level, based on the rate per unit mileage;

[0012] Accident density calculation:

[0013] Accident density = the number of accidents on the current road section within 10 minutes / road section length (km), and the data is sourced from the emergency management system;

[0014] Privacy settlement middleware: Uses differential privacy technology to anonymize cross-provincial transaction data, and combines blockchain technology to achieve automated clearing;

[0015] Emergency visualization system: The emergency visualization system aggregates Beidou positioning data (error ≤ 0.5 m), RSU status (connection success rate ≥ 99%), and meteorological data (temperature, humidity, wind speed) per second to generate the road section congestion index γ. The calculation formula is:

[0016]

[0017] where w1 = 0.6, w2 = 0.4; when γ > 0.7, a lane control plan is automatically generated and sent to the edge decision node through the MQTT protocol;

[0018] The said edge decision node includes:

[0019] Heterogeneous communication unit: Dynamically switches between C-V2X and ETC protocols through signal strength thresholds, and encrypts the communication link using the protocol;

[0020] Security authentication unit: Verifies the identity of in-vehicle terminals and generates dynamic session keys;

[0021] Lane optimizer: Dynamically allocates lane resources based on the Nash equilibrium model of non-cooperative games, and calculates the payment cost of vehicle i through the VCG mechanism. The formula is:

[0022] Payment cost i = Total system benefit -i - Total benefit

[0023] where the total system benefit is calculated by weighting the traffic efficiency T, energy consumption index E, and safety factor S:

[0024] Total benefit = 0.6T + 0.3E -1 -0.1S (weight of morning rush hour)

[0025] The morning rush hour weights β1, β2, and β3S are determined by analyzing the traffic efficiency (vehicles per hour), average energy consumption (kW·h), and accident rate (times per hour) during the morning rush hour (7:00 - 9:00) in historical data. The specific formula is as follows:

[0026]

[0027] Among them, the indicators are traffic efficiency, reciprocal of energy consumption (E -1 ) and safety factor.

[0028] In the VCG mechanism, the vehicle pays a fee equal to the reduction in the total system benefit caused by its entry into the traffic flow. This fee is superimposed on the dynamic rate to reflect the cost of the vehicle's occupation of road resources. The final toll is P + payment fee i .

[0029] The on-vehicle trusted terminal includes:

[0030] Multi-mode sensing unit: It fuses millimeter-wave radar and infrared sensing data through the Kalman filtering algorithm to generate vehicle environment perception information;

[0031] Interactive execution unit: According to cloud and edge instructions, it preferentially receives lane allocation instructions through V2X broadcasting, then switches the LED markings, and finally pushes them to the mobile terminal.

[0032] The dynamic rate adjustment method of the flexible rate engine includes the following steps:

[0033] S201. Benchmark rate prediction: Generate the initial rate P0 through the LSTM model, and normalize the input features to Z-Score;

[0034] S202. Real-time floating rule: Adjust the rate based on the percentage of traffic flow deviation from the benchmark capacity Q0:

[0035]

[0036] Among them: α = 0.15 for passenger cars or α = 0.25 for trucks, determined based on historical traffic-flow revenue elasticity experiments;

[0037] S203. Fuse mechanism: When the rate fluctuates by more than ±25% within 3 consecutive billing cycles, or the social media public opinion index λ > 5×10 4 (λ = number of negative comments × emotional intensity), freeze the rate adjustment and start the manual review process.

[0038] The linkage mechanism between the emergency visualization system and the lane optimizer includes:

[0039] When the congestion index γ > 0.7 is detected, the lane optimizer dynamically increases the number of express lanes and synchronously raises the rate floating coefficient α to 0.2 (for passenger cars) or 0.3 (for freight cars).

[0040] Emergency instructions are sent to in-vehicle terminals in real time via V2X broadcasting. If communication fails, the lane identification is switched via LED markings.

[0041] The calculation method of the public opinion index λ includes:

[0042] Analyze the sentiment polarity (-1 to 1) of social media texts through natural language processing technology;

[0043] Calculate according to the formula λ = ∑(sentiment score of a single comment × number of forwards), and update it once an hour;

[0044] Quantification of public opinion sentiment intensity:

[0045] Use the BERT model to perform sentiment analysis on social media texts, and output a score S ∈ [-1, 1]. The sentiment intensity of a single comment = |S| × 100.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] First, improve the accuracy of dynamic rates: The LSTM model combines real-time data, and the rate prediction error ≤ 2.5%, which is significantly optimized compared with the traditional linear model (error about 8%);

[0048] Second, seamless switching of multi-mode communication: The switching success rate between ETC and C-V2X ≥ 99.5%, and the passing efficiency is increased by 30%;

[0049] Third, balance data security and efficiency: The differential privacy technology makes the data anonymization process take less than 100 ms, and the blockchain clearing efficiency reaches 5000 transactions per second;

[0050] Fourth, double the emergency response speed: The whole process from data collection to instruction issuance takes less than 2 seconds, which is 150 times faster than manual decision-making (average 5 minutes);

[0051] Fifth, enhance the fairness of resource allocation: The VCG mechanism makes the vehicle payment cost match the social cost of resource occupancy, and the passing capacity of congested sections is increased by 25%. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a framework schematic diagram of the method of the present invention;

[0053] Figure 2 It is a flow schematic diagram of the lane optimizer of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0054] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0055] Please refer to Figure 1-2 , the present invention provides a technical solution: a highway toll access control system based on a cloud toll collection architecture, including a cloud intelligent billing platform:

[0056] Elastic rate engine:

[0057] Adopt a 3-layer LSTM neural network model, input road section traffic (unit: vehicles / minute), visibility (quantification level 1-5), and accident level (1-3) data, train based on the historical traffic data set (covering 5 years, 100,000 records), and predict the initial benchmark rate P0. The mean absolute error (MAE) of the model ≤ 2.5%;

[0058] LSTM model online update: The model supports incremental learning, updates the weights weekly through the incremental data set (including the latest 7-day traffic data), the update delay < 30 minutes, and the model version number is synchronized to the blockchain.

[0059] The real-time floating rule restricts the rate fluctuation range through the sigmoid function. The floating coefficients α for passenger cars and trucks are 0.15 and 0.25 respectively, which are determined based on the historical traffic-revenue elasticity experiment;

[0060] Vehicle type recognition method: The on-vehicle trusted terminal is built-in with an OBU identification code (passenger car: Class A code, truck: Class B code), and the edge node distinguishes the vehicle type by parsing the identification code;

[0061] Energy consumption index formula: E = (average speed / 60) × 0.3 + (number of hard brakes) × 0.7, unit: kW·h / 100 km, and the data source is the on-vehicle OBD interface.

[0062] The triggering conditions for the fusing mechanism include: the rate fluctuation exceeds ±25% in three consecutive billing cycles (each cycle is 5 minutes), or the social media public opinion index λ > 5 × 10 4 (λ = number of negative comments × emotional intensity), freeze the rate and start manual review.

[0063] Privacy settlement middle platform:

[0064] Apply differential privacy technology (ε = 0.1) to the cross-provincial transaction data, and add Laplace noise to achieve anonymization;

[0065] Automated clearing based on the Hyperledger Fabric blockchain framework, generating smart contracts for each transaction to ensure the immutability of data.

[0066] Emergency visualization system:

[0067] Aggregate Beidou positioning data (error ≤ 0.5 m), RSU status (connection success rate ≥ 99%), and meteorological data (temperature, humidity, wind speed) per second, and calculate the road congestion index γ:

[0068]

[0069] When γ > 0.7, automatically generate a lane control plan (such as closing the emergency lane and setting a speed limit of 60 km / h), and send it to the edge node via the MQTT protocol.

[0070] 2. Edge decision node

[0071] Heterogeneous communication unit:

[0072] Dynamically monitor the ETC signal strength (RSSI), and switch to the C-V2X mode when RSSI < -85 dBm. Use the TLS1.3 protocol to encrypt the communication link, and the transmission delay < 50 ms.

[0073] Security authentication unit:

[0074] Verify the identity of the on-vehicle terminal based on the national cryptographic SM2 algorithm, generate a dynamic session key (valid for 30 seconds), and prevent replay attacks.

[0075] Edge node offline mode: When a cloud connection interruption is detected, the edge decision node automatically switches to the local cache mode, and calls the rate table (stored in the edge node SSD) and lane allocation strategy for the last 24 hours to ensure at least 4 hours of continuous operation ability.

[0076] Lane optimizer:

[0077] Construct a non-cooperative game model, with the traffic efficiency T (vehicles / hour), energy consumption index E (kW·h), and safety factor S (0 - 1) as the optimization objectives. The weight settings for the morning rush hour are β1 = 0.6, β2 = 0.3, and β3 = 0.1;

[0078] Apply the VCG mechanism to calculate the payment fee for vehicle i:

[0079] Payment fee i = Total system benefit -i - Total benefit

[0080] Output multi-modal guidance instructions, with the priority: V2X broadcast > LED marking switching > mobile device push.

[0081] 3. On-vehicle Trusted Terminal

[0082] Multi-mode Sensing Unit:

[0083] Fuse the data of millimeter-wave radar (detection accuracy ±0.1 m) and infrared sensor (sampling frequency 10 Hz), and use the Kalman filtering algorithm to eliminate noise, generating the perception information of the vehicle surrounding environment (such as vehicle distance, obstacle position).

[0084] Interaction Execution Unit:

[0085] After receiving the instructions from the cloud and the edge, first update the lane allocation through V2X broadcast. If the communication fails, switch to the LED lane marking (response time <200 ms), and finally push it to the driver's mobile App.

[0086] Example 1: Highway Morning Rush Hour Scenario

[0087] Step 1: System Initialization and Data Collection

[0088] Cloud Intelligent Billing Platform:

[0089] The elastic rate engine loads a pre-trained 3-layer LSTM model (the number of hidden layer units is 128), inputs the traffic data of the past 5 years (traffic flow, visibility, accident level), and normalizes it into the Z-Score format;

[0090] Receive real-time data: The traffic flow of section A is Q = 1200 vehicles / minute (benchmark capacity Q0 = 1000 vehicles / minute), visibility level W = 3 (light fog), accident level E = 1 (no accident);

[0091] Predict the benchmark rate P0 = 0.5 yuan / km, and the MAE of the model is verified to be 2.3%.

[0092] Edge Decision Node:

[0093] The heterogeneous communication unit monitors the ETC signal strength (RSSI = -80 dBm) and maintains the ETC communication mode;

[0094] The security authentication unit verifies the identity of the on-vehicle terminal through the SM2 algorithm and generates a dynamic session key (validity period 30 seconds).

[0095] On-vehicle Trusted Terminal:

[0096] The multi-mode sensing unit collects millimeter-wave radar (detecting the distance to the vehicle ahead is 2.5 m ±0.1 m) and infrared sensing data (detecting lane markings) at a frequency of 10 Hz;

[0097] The Kalman filter fuses the data to generate the perception information of the vehicle surrounding environment.

[0098] Step 2: Dynamic Rate Adjustment and Lane Allocation

[0099] Real-time rate calculation:

[0100] Ratio of traffic deviation from the reference capacity

[0101] For the bus rate floating coefficient α = 0.15, substituting it into the sigmoid function:

[0102] P = 0.5×[1 + 0.15·sigmoid(0.2)] = 0.5×1.13 = 0.565 yuan / km

[0103] Rate fluctuation range is 13% (fuse mechanism not triggered).

[0104] Lane optimization and payment calculation:

[0105] The lane optimizer calculates the morning peak weight (traffic efficiency T = 0.6, energy consumption E = 0.3, safety S = 0.1) based on the non-cooperative game model;

[0106] Calculate the payment cost of vehicle i through the VCG mechanism:

[0107] Payment cost i = 8 yuan (assuming a 10% increase in the total system benefit)

[0108] Generate an instruction: preferentially allocate vehicles to the fast lane (lane number L3) through V2X broadcast.

[0109] Step 3: Emergency response and communication degradation

[0110] Sudden congestion event: The emergency visualization system detects that the congestion index γ of section B is γ = 0.6×20001500 + 0.4×0.2 = 0.68 (not exceeding the threshold of 0.7), and no control is triggered temporarily;

[0111] After 10 minutes, an accident causes γ = 0.8, and the system automatically generates an instruction: close the emergency lane, limit the speed to 60 km / h, and send it to the edge node through MQTT.

[0112] Communication link switching:

[0113] The edge node detects that the ETC signal strength drops to RSSI = -90 dBm, switches to the C-V2X mode, and the encrypted transmission delay < 50 ms;

[0114] When the in-vehicle terminal fails to receive, it degrades to LED marking switching (displaying "L3→L2"), and the response time is 180 ms.

[0115] Example 2: Cross-provincial clearing and privacy protection

[0116] Step 1: Cross-provincial transaction processing

[0117] Privacy Settlement Middle Platform:

[0118] Add Laplace noise (ε = 0.1) to the cross-provincial vehicle transaction data (such as license plate numbers, travel records) to generate an anonymized dataset;

[0119] Based on the Hyperledger Fabric blockchain, write the clearing instructions into the smart contract (processing 5,000 transactions per second), and the nodes in each province synchronize the ledger.

[0120] Step 2: Trigger the public opinion fusing mechanism

[0121] Social media public opinion monitoring:

[0122] The natural language processing module (BERT model) analyzes Weibo comments to detect negative sentiment (score -0.8) and the number of reposts (12,000 times);

[0123] Calculate the public opinion index λ = 1.2 × 10 4 × 0.8 = 9.6 × 10 4 (exceeding the threshold of 5 × 10 4 ).

[0124] Fusing execution:

[0125] The elastic rate engine freezes the rate adjustment, and a pop-up window appears on the manual review interface for prompt;

[0126] After the review is passed, dynamic price adjustment is restored, and the floating coefficient α is lowered to 0.1.

[0127] Verification experiment of the floating coefficient α: Based on the regression analysis of historical data, the elasticity coefficient of passenger car flow - revenue is 0.12 - 0.18 (taking 0.15), and that of freight cars is 0.22 - 0.28 (taking 0.25);

[0128] Public opinion threshold λ: After testing with 100,000 Weibo data, when λ > 5 × 10 4 the negative public opinion spreads at an exponential rate (R0 > 1.5), and triggering the fuse can reduce the user complaint rate by 30%.

[0129] Example 3

[0130] Linkage calculation of dynamic rate and payment fee

[0131] When vehicle i enters the section, the elastic rate engine calculates the real-time rate P = 0.5 yuan / km;

[0132] The lane optimizer calculates the payment fee for vehicle i as 1.2 yuan through the VCG mechanism (assuming its behavior causes the total system benefit to decrease by 1.2 yuan);

[0133] The final toll is 0.5 + 1.2 = 1.7 yuan, and it is pushed to the in-vehicle terminal through V2X broadcast.

[0134] Embodiment 4: Edge node offline disaster tolerance scenario

[0135] Steps: The cloud failure triggers the edge node offline mode, and the cached toll rate table (P0 = 0.5 yuan / km, α = 0.15) is read;

[0136] Based on the local camera data, the real-time traffic flow (Q = 1100 vehicles / minute) is calculated, and the toll rate is adjusted according to the offline rule: P = 0.5×[1 + 0.15×(1100 - 1000) / 1000] = 0.515 yuan / km;

[0137] The lane optimizer switches to the local game model, and the weights are fixed as A1 = 0.6, A2 = 0.3, A3 = 0.1.

[0138] Embodiment 5: Public opinion fusing verification

[0139] Steps: It is monitored that the forwarding volume of the Weibo topic "High-speed charging chaos" is 20,000 times, and the average intensity of BERT sentiment analysis is -0.7;

[0140] Calculate λ = 2×10 4 ×0.7 = 1.4×10 5 (>5×10 4 ), trigger fusing;

[0141] The toll rate adjustment is frozen for 24 hours, and it resumes after manual review and confirmation, and a notice is sent to the in-vehicle terminal.

[0142] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

Claims

1. A highway toll passage control system based on a cloud toll collection architecture, characterized in that, It includes a cloud intelligent charging platform, an edge decision-making node, and an in-vehicle trusted terminal. The cloud intelligent charging platform, the edge decision-making node, and the in-vehicle trusted terminal are all connected by signals. Among them: The cloud intelligent charging platform includes: Elastic rate engine: Adopting a 3-layer LSTM neural network, it inputs the road section traffic flow, visibility level, and accident level, and through Z-Score standardization processing, predicts the benchmark rate P0. The historical traffic data set contains historical rate label data corresponding to the road section traffic flow, visibility, and accident level; Based on the percentage of the traffic flow deviating from the benchmark capacity Q0, the rate is dynamically adjusted through the sigmoid function. The floating coefficient for passenger cars is α = 0.15, and for trucks is α = 0.25; Privacy settlement middleware: Anonymizes cross-provincial transaction data by adding Laplace noise through differential privacy and performs automated clearing based on the Hyperledger Fabric blockchain framework; Emergency visualization system: Aggregates Beidou positioning data, RSU status, and meteorological data per second, calculates the congestion index γ, and generates lane control instructions when γ > 0.7; The edge decision-making node includes: Heterogeneous communication unit: Switches modes when the ETC signal strength is low; Security authentication unit: Verifies the identity of the in-vehicle terminal based on the SM2 algorithm and generates a dynamic session key; Lane optimizer: Optimizes the weighted total benefit of the traffic efficiency T, energy consumption index E, and safety factor S through a non-cooperative game model. The weights for the morning rush hour are β1 = 0.6, β2 = 0.3, β3 = 0.1, and calculates the payment cost of vehicle i through the VCG mechanism. The formula is: Payment cost i = Total benefit - i - Total benefit Among them, the total benefit - i is the total system benefit when vehicle i does not participate, the total benefit is the total system benefit after vehicle i participates, and the total benefit is calculated by β1T + β2E -1 - β3S; The in-vehicle trusted terminal includes: Multi-mode sensing unit: Fuses millimeter-wave radar and infrared sensing data and generates environmental perception information through Kalman filtering; Interaction execution unit: Prioritizes receiving instructions through V2X broadcast, secondly selects LED marking switching, and finally pushes them to the mobile terminal.

2. The highway toll access control system based on a cloud toll collection architecture according to claim 1, characterized in that: The training method of the LSTM model of the elastic rate engine includes the following steps: Data preprocessing: Performs Z-Score standardization on the road section traffic flow, visibility, accident level, and corresponding historical rates in the historical traffic data set; Model training: Using the standardized road section traffic flow, visibility, and accident level as inputs and the historical rate as the output label, trains a 3-layer LSTM neural network, and the model mean absolute error (MAE) ≤ 2.5%; Real-time prediction: Inputs real-time road section traffic flow, visibility, and accident level data and outputs the benchmark rate P0.

3. The highway toll access control system based on a cloud toll collection architecture according to claim 1, characterized in that: The application of the VCG mechanism includes: Calculating the total benefit -i of the system when vehicle i does not participate; According to the impact of the driving behavior of vehicle i on the total benefit, calculates its payment cost as: Payment cost i = Total system benefit - i - Total benefit; The payment cost is superimposed with the dynamic rate as the final passing cost of the vehicle.

4. The highway toll access control system based on a cloud toll collection architecture according to claim 1, characterized in that: The morning rush hour weights β1, β2, and β3 are determined by analyzing the traffic efficiency, energy consumption, and accident rate during the morning rush hour in the historical traffic data and are dynamically updated to the edge decision-making node.

Images