A privacy protection based on actual distance strategy distributed control method for vehicle platoon

Abstract

This invention discloses a privacy-preserving distributed platoon control method based on a practical spacing strategy, specifically including the following steps: constructing a connected platoon model containing leader and follower vehicles; designing a practical spacing strategy to adaptively adjust vehicle spacing by combining various real-time road conditions; encrypting and compressing vehicle transmission data using an encoding-decoding scheme, and injecting Laplace noise into the original data using a differential privacy mechanism to protect information such as vehicle initial state, position, speed, and acceleration; designing a distributed cooperative controller to optimize privacy protection and system stability; verifying mean-square consistency and privacy protection effects through simulation, and highlighting the superiority of the practical spacing strategy through comparison. This invention significantly improves the privacy and security of data transmission while ensuring the accuracy of platoon cooperative control.

Description

Technical Field

[0001] This invention belongs to the fields of intelligent transportation systems, vehicle cooperative control and privacy protection, and specifically relates to a distributed fleet control method based on actual spacing strategy under privacy protection. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) technology, the collaborative control of connected vehicle fleets plays a crucial role in improving traffic efficiency and safety. However, privacy breaches in data transmission between vehicles have become a key challenge hindering the technology's practical application. Traditional methods rely on manually designed encryption strategies, which are vulnerable to complex attacks and do not fully consider the impact of fleet dynamics on privacy protection. While existing differential privacy mechanisms can protect data through noise injection, they are prone to system instability in collaborative fleet control.

[0003] Furthermore, vehicle platooning spacing strategies directly impact driving safety, passenger comfort, and overall traffic throughput. Existing spacing strategies (such as constant spacing, constant interval, and variable interval strategies) often fail to adequately account for sudden fluctuations in traffic density, sudden deterioration of road conditions, or changes in road friction characteristics. These shortcomings can lead to increased collision risk, reduced platoon stability, or decreased traffic throughput.

[0004] Therefore, there is an urgent need for a vehicle platooning cooperative control method that takes into account privacy, stability, and road condition adaptability. Summary of the Invention

[0005] To address the aforementioned issues, this invention discloses a privacy-preserving distributed fleet control method based on a real-spaced spacing strategy. This method integrates vehicle longitudinal dynamics, a general V2V (vehicle-to-vehicle) communication topology, an adaptive spacing strategy, and privacy-preserving vehicle data exchange. The real-spaced spacing strategy is designed to adaptively adjust vehicle distances by incorporating real-time traffic flow changes, road friction conditions, and safety distance requirements. Through a differential privacy-based mechanism combining noise injection and encoder-decoder, along with a collaborative design strategy, the privacy-preserving scheme, estimator, and distributed controller are combined to ensure estimation accuracy, data privacy, and reliable fleet tracking.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] A privacy-preserving, distributed fleet control method based on a real-time spacing strategy includes the following steps:

[0008] S1. Construct a connected vehicle fleet system model, including one leader vehicle and multiple follower vehicles, and define the vehicle dynamics model and actual spacing strategy;

[0009] S2. Based on the vehicle model and actual spacing strategy in step S1, construct an augmented state system and a state observer to estimate the vehicle state.

[0010] S3. After the state observer in step S2 estimates the vehicle state, it adds Laplace noise to the original data through the noise injector to generate disturbance data.

[0011] S4. Encode and compress the disturbance data generated in step S3 to obtain a transmissible codeword and a dynamic key that can adjust the encoding error and obtain vehicle data information, and construct an encoding-decoding module.

[0012] S5. Combining the above steps, define various errors such as consistency tracking error, coding error, and estimation error, and determine the control objective and the form of the distributed cooperative controller;

[0013] S6. Conduct collaborative design of distributed collaborative controllers, determine various gain matrices of the controllers, and ensure mean square consistency and privacy protection of vehicle formation.

[0014] Further, step S1 defines It is the inertial delay of the vehicle engine. It is a vehicle If the control input is given, then the connected vehicle fleet model is a leader vehicle numbered 0 and... Vehicle number The following vehicles take the following forms:

[0015] ;

[0016] in, It's the vehicle status. It is the sensor measurement output. ,in , , , , , Representing vehicles Position, velocity, and acceleration The sampling period. , Indicates vehicle The ideal distance between the leader's vehicle and the vehicle itself, in addition, , , It is a given measurement matrix.

[0017] Practical Spacing Policy:

[0018] ;

[0019] in,

[0020] ;

[0021] It is a headway function based on traffic flow. The correlation coefficient is low for traffic. For the density of blockage, For free-flowing traffic speed;

[0022] ;

[0023] This represents the time-distance function of the vehicle's front end based on road friction. This represents the maximum coefficient of friction on a dry road surface. This is an empirical attenuation coefficient, representing the effect of surface humidity on friction reduction. This represents the estimated road surface slippage derived from differences in wheel speeds; furthermore,

[0024] ;

[0025] This indicates an additional safety clearance item. Among them,

[0026] ;

[0027] Here , It is the minimum safe distance.

[0028] Furthermore, in step S2, given the vehicle model and actual spacing strategy from step S1, the vehicle's state is difficult to obtain directly. Therefore, an augmented state system and observer are constructed as follows:

[0029] In real-world scenarios, It must meet the following form:

[0030] ;

[0031] in, , It is bounded, and .

[0032] Construct the following augmented states:

[0033] ;

[0034] The augmentation system is constructed in the following form:

[0035] ;

[0036] in,

[0037] ;

[0038] ;

[0039] The following observation model is obtained by estimating the vehicle state using a state observer:

[0040] ;

[0041] in, , , and They are respectively , and The estimate, It is the observer gain matrix.

[0042] Furthermore, in step S3, after the state observer in step S2 estimates the vehicle state, Laplace noise is added to the original data through a noise injector. ,in Satisfies the following probability density function:

[0043] ;

[0044] in, Expected value ,variance , express The central location of the value distribution. express The degree of dispersion of the value distribution, and The specific value needs to be designed in step S6 in conjunction with the controller.

[0045] The disturbance state after noise injection is estimated as follows:

[0046] ;

[0047] in, .

[0048] Further, in step S4, the perturbation data generated in step S3 is encoded and compressed, combined with a dynamic key, and quantized to obtain a transmittable codeword. Based on the perturbation state estimation after noise injection in step S3, the constructed encoder and decoder are as follows:

[0049] vehicle Encoder:

[0050] ;

[0051] in, It is the code word that needs to be transmitted. , It is a dynamic key. For a vector Uniform quantization function Defined as:

[0052] ;

[0053] in, It is a given quantization parameter. .

[0054] vehicle Decoder:

[0055] ;

[0056] in, It is the state estimate after decoding. It is a vehicle The set of neighbors. For have Only with the correct dynamic key can information such as the vehicle's initial position, driving position, and speed be decoded;

[0057] Furthermore, in step S5, the consistency tracking error is defined as... The encoding error is The estimation error is , Estimation of augmented states The error is the quantization error:

[0058] ;

[0059] have:

[0060] ;

[0061] and

[0062] ;

[0063] The form and control objective of the distributed cooperative controller for the vehicle platooning system are as follows:

[0064] vehicle Distributed collaborative controller:

[0065] ;

[0066] in, It is the gain of the controller. It is the feedback gain of the controller.

[0067] If the following conditions are met:

[0068] ;

[0069] The convoy system is then considered to have achieved bounded convoy tracking, meaning that following vehicles can maintain the desired distance, speed, and acceleration from the lead vehicle. It is a positive constant.

[0070] Furthermore, the collaborative design process of the relevant distributed collaborative controller in step S6 is as follows:

[0071] (1) Design the controller gain and feedback gain matrix Make:

[0072] ;

[0073] in, .

[0074] (2) Design the observer gain matrix Make:

[0075] ;

[0076] (3) Select a dynamic key based on the following conditions. :

[0077] ;

[0078] in, .

[0079] (4) Select an independent Laplace matrix satisfy:

[0080] ;

[0081] in, , .

[0082] (5) The design of the distributed controller gain, feedback matrix, and observer gain is obtained from the solutions of the following two discrete Riccati equations:

[0083] ;

[0084] in , Therefore, for (4) time Right now Need to meet , , , and .

[0085] ;

[0086] in , From this, we can obtain The gain matrix of the observer is designed as follows: .

[0087] The beneficial effects of this invention are as follows:

[0088] This invention proposes a distributed cooperative control scheme for connected vehicle fleets based on a real-time spacing strategy while protecting vehicle data privacy. This scheme integrates vehicle longitudinal dynamics, a general V2V (vehicle-to-vehicle) communication topology, an adaptive spacing strategy, and privacy-preserving vehicle data exchange. First, a real-time spacing strategy is designed to adaptively adjust the distance between vehicles by incorporating real-time traffic flow changes, road friction conditions, and safety distance requirements. The resulting uncertainties are effectively addressed using an augmented system-based estimation method.

[0089] Meanwhile, to protect vehicle privacy, a differential privacy-based mechanism combining noise injection and encoder-decoder was developed. Based on this, a collaborative design strategy was proposed to combine the privacy protection scheme, estimator, and distributed controller to ensure the accuracy of estimation, data privacy, and reliable fleet tracking, thereby enhancing the safety, stability, and privacy of vehicle platooning in real-world driving scenarios.

[0090] In summary, this invention provides a privacy-preserving, stable solution for vehicle platooning control that is applicable to various road conditions through the organic combination of algorithmic innovation and data information interaction mechanisms, and has good engineering adaptability and prospects for widespread application. Attached Figure Description

[0091] Figure 1 This is a flowchart of the vehicle privacy protection scheme based on encoding-decoding and noise injection according to the present invention;

[0092] Figure 2 This is a noise distribution diagram injected into the vehicle position, speed, and acceleration respectively in this invention;

[0093] Figure 3 It is the trajectory of the expected value of the tracking error in this invention;

[0094] Figure 4This invention relates to the positional error and speed trajectory of vehicles in a connected vehicle fleet.

[0095] Figure 5 and Figure 6 This is a diagram showing the relative positions of each following vehicle and the lead vehicle when using different spacing strategies in this invention, as well as when using actual spacing strategies under different road conditions.

[0096] Figure 7 The data of vehicle 1 in the connected vehicle fleet of the present invention after being decrypted using different keys are: (a) the position of vehicle 1 after decryption; (b) the speed of vehicle 1 after decryption.

[0097] Figure 8 The data of vehicle 5 in the connected vehicle fleet in this invention after being decrypted using different keys are: (a) the decrypted position of vehicle 5; (b) the decrypted speed of vehicle 5.

[0098] Figure 9 This is a trajectory diagram of the data released by vehicles in the connected vehicle fleet under different initial states in this invention.

[0099] Figure 10 This is a flowchart illustrating the implementation process of the present invention. Detailed Implementation

[0100] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0101] Example

[0102] like Figure 1 As shown, the distributed fleet control method based on actual spacing strategy under privacy protection according to the present invention includes the following steps:

[0103] S1. Construct a connected vehicle fleet system model, including one lead vehicle and multiple follower vehicles, and define the vehicle dynamics model and actual spacing strategy.

[0104] Consider a convoy consisting of 5 following vehicles and 1 leader vehicle traveling on a flat road. The specific form of its connected convoy model is as follows:

[0105] ;

[0106] Assuming the vehicle's system matrix Input matrix , The sampling period is set to 0.01, and the vehicle engine inertial delay is considered. The value is set to 0.2, the V2V communication topology is selected as LPF, and the ideal distance between vehicles is set to 10. The simulation time is set to 120. The key required for encoding and decoding is designed as follows: The acceleration of the leader's vehicle Choose a piecewise function, the specific function is shown below: , ; , ; , ; , ; , .

[0107] Practical Spacing Policy:

[0108] ;

[0109] in,

[0110] ;

[0111] It is a headway function based on traffic flow. The correlation coefficient is low for traffic. For the density of blockage, For free-flowing traffic speed;

[0112] ;

[0113] This represents the time-distance function of the vehicle's front end based on road friction. This represents the maximum coefficient of friction on a dry road surface. This is an empirical attenuation coefficient, representing the effect of surface humidity on friction reduction. This represents the estimated road surface slippage derived from differences in wheel speeds; furthermore,

[0114] ;

[0115] This indicates an additional safety clearance item. Among them,

[0116] ;

[0117] Here , It is the minimum safe distance.

[0118] In this example, we assume that the initial traffic has a low correlation coefficient. Set to 10, clogging density = 0.1, maximum coefficient of friction Take 0.2, empirical attenuation coefficient We set the value to 0.8. The initial spacing policy is a constant spacing policy (CS): =10 .

[0119] S2. Given the vehicle model and actual spacing strategy in step S1, the vehicle state is difficult to obtain directly. Therefore, an augmented state system and a state observer are constructed to estimate the vehicle state.

[0120] Construct the following augmented states:

[0121] ;

[0122] The augmentation system is constructed in the following form:

[0123] ;

[0124] in,

[0125] ;

[0126] The following observation model is obtained by estimating the vehicle state using a state observer:

[0127] ;

[0128] in, , , and They are respectively , and The estimate, It is the observer gain matrix.

[0129] S3. After the state observer in step S2 estimates the vehicle state, it generates disturbance data by injecting Laplace noise into the original data through the noise injector.

[0130] Laplace noise Satisfies the following probability density function:

[0131] ;

[0132] Laplace noise The expected value and variance are set as follows: , , , . Figure 2 The noise distribution and noise variance decay curves injected into position, velocity, and acceleration based on the probability density function are shown.

[0133] S4. Encode and compress the disturbance data generated in step S3 to obtain a transmissible codeword and a dynamic key that can adjust the encoding error and obtain vehicle data information, and construct an encoding-decoding module.

[0134] Suppose that a potential eavesdropper has obtained the decryption scheme of the fleet system in advance and has intercepted all data transmitted in the VANET (Vehicular ad-hoc network). However, the eavesdropper cannot obtain the accurate decryption key, so he can only rely on an inaccurate decryption key to decode the intercepted data. The decryption performance is evaluated by comparing the decryption of the data using different decryption keys and comparing the actual results. The different key selections are as follows: (1) (2) (3) (4) (5) ;

[0135] S5. Combine the above steps to define consistency tracking error, estimation error, etc., and determine the control objective and the form of the distributed cooperative controller;

[0136] S6. Conduct collaborative design of distributed collaborative controllers, determine various gain matrices of the controllers, and ensure mean square consistency and privacy protection of vehicle formation.

[0137] choose Then the feedback matrix can be calculated. Other relevant parameters are as follows: , , , , .like Figure 3 As shown, when the simulation time gradually increases, Gradually approaching zero, which means , Figure 4 The performance in position error and velocity tracking is demonstrated. It can be observed that the proposed formation control strategy can achieve the desired mean-square formation performance.

[0138] Figure 5 , 6 The relative positions of following vehicles to the lead vehicle were compared under different spacing strategies and traffic scenarios. Assuming the default road conditions are uncongested and the road surface is dry, the convoy first traveled under different spacing strategies:

[0139] (a) Constant Spacing Policy (CS): =10 .

[0140] (b) Constant Time Headway Policy (CTH): ,in and .

[0141] (c) Variable Time Headway Policy (VTH): ,in , and .

[0142] From strategies (b) and (c), it can be concluded that the spacing changes with the speed of the leading vehicle, such as Figure 5 As shown.

[0143] The convoy now employs the practical spacing strategy designed in this invention, assuming:

[0144] (e) A sudden traffic jam occurs at t = 40s, and the congestion density is... It rises from 0.1 to 0.2, and then decreases to 0.1 at t = 80s.

[0145] (f) On dry road surfaces ( ) and completely slippery road surface ( (Drive)

[0146] (g) Guide the vehicle to perform emergency braking under both the CS strategy and the actual distance strategy. The minimum safe distance in the actual distance strategy is set to... = 6m.

[0147] like Figure 5 As shown, other spacing strategies have significant limitations and are less practical than the actual spacing strategies. For example... Figure 6 The diagram illustrates a scenario where, under the CS strategy, a collision occurs between the leading vehicle and following vehicles when the leading vehicle brakes suddenly, resulting in poor safety. However, regardless of deteriorating road conditions or emergency braking, the actual spacing strategy designed in this invention adaptively adjusts vehicle spacing based on changes in road congestion density, slipperiness, and sudden emergencies, ensuring that the vehicle spacing does not fall below a safe threshold, thereby preventing potential collisions and rear-end collisions.

[0148] Figure 7 and Figure 8 The decryption results for following vehicle 1 and following vehicle 5 under different keys are shown. Clearly, only when the correct key is used ( Only when the decrypted data is aligned with the actual vehicle status information can it be accurately determined. Even if the difference between the incorrect and correct keys is small, the error between the decrypted data and the actual vehicle information will amplify over time if an eavesdropper uses an incorrect key. This ensures the privacy of the transmitted data, making it difficult for unauthorized entities to easily access the actual vehicle information.

[0149] Next, we analyze the performance of the differential privacy protection scheme in protecting the initial state of the vehicle. Assume an eavesdropper intercepts data transmitted via VANET and attempts to estimate the initial state of vehicle 1, i.e. In addition, the extra controller gain is designed for Consider the initial state of a pair of vehicles, one of which is private. and ,mean .according to Inject separately and .if and right satisfy The data released into VANET , In exactly the same, such as Figure 9 As shown, this indicates that there exists a situation where the data released into the VANET is the same regardless of the vehicle's initial state. This prevents eavesdroppers from making an accurate estimate of the initial data value, thus protecting the privacy of the vehicle's initial state to some extent.

[0150] The implementation process framework of this invention is as follows: Figure 10 As shown.

[0151] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A distributed fleet control method based on a real-time spacing strategy under privacy protection, characterized in that, Includes the following steps: S1. Construct a connected vehicle fleet system model, including one leader vehicle and multiple follower vehicles, and define the vehicle dynamics model and actual spacing strategy; S2. Based on the vehicle model and actual spacing strategy in step S1, construct an augmented state system and a state observer to estimate the vehicle state. S3. After the state observer in step S2 estimates the vehicle state, it adds Laplace noise to the original data through the noise injector to generate disturbance data. S4. Encode and compress the disturbance data generated in step S3 to obtain a transmissible codeword and a dynamic key that can adjust the encoding error and obtain vehicle data information, and construct an encoding-decoding module. S5. Combine the above steps to define various errors and determine the control objective and the form of the distributed cooperative controller. The various errors include consistency tracking error, coding error, estimation error, augmented state estimation error, and quantization error. S6. Conduct collaborative design of distributed collaborative controllers, determine the various gain matrices of the controllers, and ensure mean square consistency and privacy protection of vehicle formation.

2. The method according to claim 1, characterized in that, In step S1, it is assumed that It is the inertial delay of the vehicle engine. It is a vehicle The control input, the connected fleet model is a leader vehicle numbered 0 and Vehicle number The following vehicles take the following forms: ； in, Indicates the vehicle's status. It is the sensor's measurement output; ,in , , , , , Representing vehicles Position, velocity, and acceleration The sampling period; , Indicates vehicle The ideal distance between the leader's vehicle and the vehicle itself, in addition, , , It is a given measurement matrix; Actual spacing strategies based on real-world traffic scenarios: ； in, ； It is a headway function based on traffic flow. The correlation coefficient is low for traffic. For the density of blockage, For free-flowing traffic speed; ； This represents the time-distance function of the vehicle's front end based on road friction. This represents the maximum coefficient of friction on a dry road surface. This is an empirical attenuation coefficient, representing the effect of surface humidity on friction reduction. This represents the estimated road surface slippage derived from differences in wheel speeds; furthermore, ； This indicates the additional safety clearance item; where, ； Here , It is the minimum safe distance.

3. The method according to claim 1, characterized in that, In step S2, given the vehicle model and actual spacing strategy from step S1, the vehicle's state is difficult to obtain directly. Therefore, an augmented state system and observer are constructed as follows: In real-world scenarios, It must meet the following form: ； in, , It is bounded, and ; Construct the following augmented states: ； The augmentation system is constructed in the following form: ； in, ； ； The following observation model is obtained by estimating the vehicle state using a state observer: ； in, , , and They are respectively , and The estimate, It is the observer gain matrix.

4. The method according to claim 1, characterized in that, In step S3, after the state observer in step S2 estimates the vehicle state, Laplace noise is added to the original data through the noise injector. ,in Satisfies the following probability density function: ； in, Expected value ,variance , express The central location of the value distribution. express The degree of dispersion of the value distribution, and The design needs to be combined with the controller in step S6; The disturbance state after noise injection is estimated as follows: ； in, .

5. The method according to claim 1, characterized in that, In step S4, the perturbation data generated in step S3 is encoded and compressed, combined with a dynamic key, and quantized to obtain a transmissible codeword. Based on the perturbation state estimation after noise injection in step S3, the constructed encoder and decoder are as follows: vehicle Encoder: ； in, It is the code word that needs to be transmitted. It is used to obtain Auxiliary variables, , It is a dynamic key; for a vector Uniform quantization function Defined as: ； in, These are the given quantization parameters. ; vehicle Decoder: ； in, It is the state estimate after decoding. It is a vehicle The set of neighbors; for have Only with the correct dynamic key can the vehicle's initial information be decoded.

6. The method according to claim 1, characterized in that, In step S5, the consistency tracking error is defined as... The encoding error is The estimation error is , Estimation of augmented states The error is the quantization error: ； have: ； and ； The form and control objective of the distributed cooperative controller for the vehicle platooning system are as follows: vehicle Distributed collaborative controller: ； in, It is the gain of the controller. It is the feedback gain of the controller; If the following conditions are met: ； The convoy system is then considered to have achieved bounded convoy tracking, meaning that following vehicles can maintain the desired distance, speed, and acceleration from the leader vehicle. It is a positive constant.

7. The method according to claim 1, characterized in that, The collaborative design process of the relevant distributed collaborative controller in step S6 is as follows: (1) Design the controller gain and feedback gain matrix Make: ； in, ; (2) Design the observer gain matrix Make: ； (3) Select a dynamic key based on the following conditions. : ； in ; (4) Select an independent Laplace matrix satisfy: ； in, , ; (5) The design of the distributed controller gain, feedback matrix, and observer gain is obtained from the solutions of the following two discrete Riccati equations: in , Therefore, for (4) time Right now Need to meet ,in , , and ; ； in , Thus we obtain The gain matrix of the observer is designed as follows: .

Images