Vehicle formation memory event triggering adaptive fuzzy control method under privacy protection

Abstract

The invention relates to a vehicle formation memory event triggering adaptive fuzzy control method under privacy protection. Compared with the prior art, the defect that vehicle formation control is difficult to meet privacy protection, communication resource saving and distributed cooperative control at the same time is overcome. The method comprises the following steps: establishing a vehicle dynamics model; establishing an encryption domain equivalent system model; designing an encryption domain distributed virtual control law; designing a normalized collaborative learning adaptive law; constructing a memory updating rule; and controlling the vehicle formation. According to the invention, through a full-state time mask privacy protection mechanism and a normalized collaborative learning adaptive fuzzy approximation and memory event triggering mechanism, bounded stable tracking of the track of the pilot vehicle and maintenance of the expected formation configuration are realized on the premise of ensuring that all following vehicles protect initial sensitive state information.

Description

Technical Field

[0001] This invention relates to the field of multi-agent system control technology, specifically a privacy-preserving vehicle platooning memory event-triggered adaptive fuzzy control method. Background Technology

[0002] Currently, research on vehicle platooning system control has many branches, including consensus control, platooning control, and swarm control, among which consensus control is the foundation and core of other research branches. Consensus control in vehicle platooning systems refers to designing appropriate control protocols for each agent, enabling them to interact only with their neighbors, ultimately leading all agents to the same target state. After nearly two decades of development, consensus research in vehicle platooning systems has gradually matured, giving rise to important directions such as adaptive control and event-triggered control.

[0003] In actual vehicle platooning operations, vehicle dynamics models often contain parameter uncertainties and unknown nonlinearities, including measurement deviations in parameters such as vehicle mass error, engine time delay, and air resistance. Traditional control methods typically rely on precise vehicle model parameters, making it difficult to guarantee ideal control performance when model uncertainties exist. Fuzzy logic systems and neural networks can approximate lumped unknown nonlinearities with lower-dimensional parameter vectors, effectively reducing dependence on precise model parameters. However, most existing adaptive approximation methods independently design adaptive laws for each vehicle, failing to fully utilize the learning information from neighboring vehicles, resulting in slow convergence speeds and low communication resource utilization.

[0004] Meanwhile, vehicle platooning systems are typical networked multi-agent systems, with vehicle position, speed, and acceleration status information continuously exchanged between vehicles. However, frequent and continuous communication not only leads to excessive network bandwidth consumption but also exposes sensitive information such as vehicle initial states to the risk of eavesdropping, inference, or even reconstruction. While traditional event-triggered control methods can reduce unnecessary data transmission, most employ fixed trigger thresholds, limiting their adaptability to dynamic system changes. Existing privacy protection methods primarily target general multi-agent systems, and research on third-order nonlinear uncertain vehicle platooning systems that balance privacy and control accuracy remains relatively insufficient.

[0005] Furthermore, achieving both communication resource conservation and high-performance collaborative control under privacy constraints is a significant challenge in the field of vehicle platooning. Memory event triggering mechanisms, by dynamically constructing triggering criteria using historical sampling data, can more fully utilize past system information to improve transient performance and reduce redundant triggering. However, further research is needed on combining these mechanisms with privacy protection and collaborative learning for application in third-order nonlinear uncertain vehicle platooning.

[0006] Therefore, there is an urgent need to implement an adaptive fuzzy control method for vehicle platooning triggered by memory events under privacy protection, in order to solve the above-mentioned technical problems. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing vehicle platooning control technologies in simultaneously satisfying privacy protection, communication resource conservation, and distributed collaborative control, and to provide a privacy-protected vehicle platooning memory event-triggered adaptive fuzzy control method to solve the above problems.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows:

[0009] A privacy-preserving vehicle platooning memory event-triggered adaptive fuzzy control method includes the following steps:

[0010] Establish a vehicle dynamics model;

[0011] Establish an equivalent system model for the encrypted domain;

[0012] Design a distributed virtual control law for the encrypted domain;

[0013] Design a normalized collaborative learning adaptive law: Introduce a fuzzy logic system to approximate the unknown nonlinear terms in the encrypted domain, and design a normalized collaborative learning adaptive law;

[0014] Constructing memory update rules: Design normalized continuous control signals, construct memory event triggering mechanisms and threshold memory update rules;

[0015] Vehicle platoon control: Generate the final control input for each following vehicle to achieve distributed collaborative control under privacy protection constraints.

[0016] The establishment of the vehicle dynamics model involves: constructing a vehicle platooning communication topology and establishing a third-order nonlinear uncertain vehicle dynamics model; including the following steps:

[0017] Constructing the vehicle platoon communication topology: A directed weighted graph is used to describe the communication topology of a vehicle platoon consisting of one lead vehicle and its platoon. The platoon communication structure consisting of following vehicles constructs an augmented matrix using an adjacency matrix, a Laplacian matrix, and a pinpoint matrix. ,in To augment the topology matrix, the extended directed graph must contain a directed spanning tree rooted at node 0 of the lead vehicle, ensuring that the lead vehicle's reference information is reachable by all following vehicles.

[0018] Establish a third-order nonlinear uncertain longitudinal dynamic model: establish the first A following vehicle Vertical position of time Longitudinal velocity Longitudinal acceleration Third-order dynamic model; ensembles the parameter uncertainties and unknown nonlinearities into an unknown function. ,in For the first A lumped unknown nonlinear function for a following vehicle, used to control the input. Drive acceleration subsystem, For the first The actual control input of the following vehicle;

[0019] Define the desired formation configuration: based on the desired net spacing. With vehicle body length Define the expected cumulative distance between each following vehicle and the lead vehicle. Introducing offset position output ,in For the first Output the offset position of each following vehicle; transform the formation maintenance problem into the offset position output of each following vehicle relative to the position of the lead vehicle. The tracking issue.

[0020] The establishment of the equivalent system model in the encrypted domain is as follows: Based on a third-order nonlinear uncertain vehicle dynamics model, a full-state time masking privacy protection mechanism is designed to perform finite-time encryption on the position, velocity, and acceleration states, thereby establishing the equivalent system model in the encrypted domain; including the following steps:

[0021] For the first Each state dimension of the following vehicle Define the time mask function:

[0022] ,

[0023] in, For the first The following vehicle The time mask function of the dimensional state, These correspond to the three state dimensions of position, velocity, and acceleration, respectively. For time, The duration of privacy protection defined for the user. It is a positive integer. The decay rate parameter can be designed; the function and its derivative are continuous at any time.

[0024] Define full-state encryption transformation:

[0025] ,

[0026] in, For the first The encryption status of the offset position of the following vehicle. These are the encryption states for velocity and acceleration, respectively. For the first The longitudinal position of the following vehicle. For the first The original state variables are preserved; the navigator's reference signal is also encrypted, and an encrypted reference signal is defined. ,in Encrypt the reference signal for the lead vehicle. For the time mask function corresponding to the lead car, when The time mask function degenerates into the identity transformation, and the encrypted state is automatically restored to the plaintext state.

[0027] Define the inverse mask ,in For the first The following vehicle Using the inverse of the dimensional mask function, establish an equivalent system model for the cryptographic domain:

[0028] ,

[0029] in, Encrypted status Time derivative, , Given an auxiliary function, For terms containing unknown nonlinear terms Unknown auxiliary functions, To maintain control input for event triggering, Given a known bounded control gain, , These are the known upper bounds for vehicle mass and engine time lag, respectively. For the first The vehicle quality of the following vehicle For the first Engine lag of the following vehicle.

[0030] The design of the encrypted domain distributed virtual control law is as follows: based on the encrypted domain equivalent system model, the encrypted domain synchronization error and coordinate transformation are constructed, and the encrypted domain distributed virtual control law is designed based on the backstepping method; including the following steps:

[0031] Define the encryption domain synchronization error and coordinate transformation:

[0032] ,

[0033] in, For the first Synchronization error of the first-layer encrypted domain of the vehicle-following vehicle fusion communication topology weighting. , For the first Layer coordinate transformation error For vehicles in the adjacency matrix To the vehicle Communication weight, For the navigator and the first in the fixed matrix The connection weight of the following vehicle , The first , The location of the following vehicle is encrypted. Encrypt the reference signal for the lead vehicle. For the first The following vehicle 3D encryption state, For the virtual control quantity to be designed;

[0034] Step-by-step design: Selecting Lyapunov functions ,in For the first step Lyapunov function, design the first-level virtual control law:

[0035] ,

[0036] in, For the first The first layer of virtual control variables for the following vehicle. , It is the inverse function of the mask. For the first The in-degree of the following vehicle in the communication topology, For design parameters, The time derivative of the encrypted state of neighboring vehicle locations. The time derivative of the encrypted reference signal for the navigator vehicle. mask function The time derivative;

[0037] The second step is the backstep design: using a second-order sliding mode integral filter for estimation. Choose the Lyapunov function ,in Design a second-level virtual control law for the second-step Lyapunov function:

[0038] ,

[0039] in, For the first The second-level virtual control quantity for the following vehicle. It is the inverse function of the 3rd dimension mask. For design parameters, For filter pairs The estimated output, mask function Time derivative, For the first The speed of the following vehicle is encrypted. This represents the error in the second-level coordinate transformation.

[0040] The design of the normalized collaborative learning adaptive law is as follows: based on the distributed virtual control law of the cryptographic domain, a fuzzy logic system is introduced to approximate the unknown auxiliary function, and a normalized collaborative learning adaptive law is designed; including the following steps:

[0041] Using fuzzy logic systems to handle unknown auxiliary functions in the encrypted domain Approximation:

[0042] ,

[0043] in, For the fuzzy system input vector, For the ideal weight vector, For fuzzy rule numbers, For fuzzy basis function vectors, For fuzzy approximation error, satisfy , To approximate the upper bound of the error;

[0044] Choose a Lyapunov function that includes weight estimation error. ,That This is the error in the third-level coordinate transformation. For adaptive gain, For weight estimation error, For ideal weights Estimated value; Design nominal continuous control quantity:

[0045] ,

[0046] in, This is the third layer of nominal continuous control quantity. For the first The inverse function of the acceleration mask of the following vehicle, i.e. , For the first The inverse function of the speed mask of the following vehicle, i.e. , for The reverse, For design parameters, For the second-order sliding mode integral filter, the second-level virtual control quantity Output of the time derivative estimate;

[0047] Design a normalized collaborative learning fuzzy weight adaptive law:

[0048] ,

[0049] in, Weight estimates Time derivative, For the first Fuzzy weight estimates of neighboring vehicles following each other. for The Euclidean norm, As the normalization factor, For the first The third-level coordinate transformation error of the following vehicle. It is an exponential decay factor. The weight decay coefficient is used; the normalization factor and the exponential decay factor together limit the magnitude of the synergistic term, without the need to presuppose that the neighbor weights are bounded.

[0050] The memory update rule is based on the nominal continuous control quantity. The design of a normalized continuous control signal and the construction of a memory event triggering mechanism and threshold memory update rules include the following steps:

[0051] Define a normalized continuous control signal:

[0052] ,

[0053] in, For the first Normalized continuous control signals for following vehicles. This is the relative trigger threshold coefficient. Let be the shape parameter of the hyperbolic tangent function. This is the amplitude limiting parameter;

[0054] Define sample hold error ,in For the first The sampling and holding error of the following vehicle; design a memory event triggering mechanism:

[0055]

[0056] in, For the first The following vehicle Next trigger moment For the next trigger time, The threshold is a time-varying absolute threshold, maintaining a piecewise constant value between adjacent threshold update times; the design parameters must meet the following requirements. ;

[0057] Design threshold memory update rules: Incorporate historical control data. With memory sensitivity :

[0058] ,

[0059] in, For the first The amount of historical control data for each following vehicle reflects the historical trend of control signal changes. This is a measure of memory sensitivity. This represents the number of historical data packets. For the weighting coefficients, satisfying , , For the first The normalized continuous control signal sample value corresponding to the next trigger time. For the first The historical sampled value corresponding to the next trigger time; when the condition is met At that time, update the threshold according to the rolling average rule:

[0060]

[0061] in, For the first The time-varying threshold after the next update For the first The threshold update time; when The timing mechanism degenerates into a fixed threshold event triggering; when The time memory mechanism comes into play; the richer the historical information, the stronger the threshold adaptability.

[0062] The vehicle formation control is based on a memory event triggering mechanism and the first... Normalized continuous control signal for following vehicle Generate the final control input for each following vehicle to achieve distributed collaborative control under privacy protection constraints; including the following steps:

[0063] Generate the final control input for each following vehicle: based on the first-level virtual control quantity. Second-level virtual control quantity and nominal continuous control quantity According to the normalized continuous control signal After the memory event triggering mechanism determines the event trigger, it updates the event trigger to maintain the control input. In the untriggered zone Inside, Keep the sampled value from the previous trigger moment unchanged; the final implementation is as follows:

[0064] ;

[0065] Keep control input triggered by the event Mapped to , For the first The actual control input applied to the acceleration subsystem by the following vehicle; the control law only depends on the encrypted state of the vehicle itself, the encrypted state of the neighboring vehicles, and the encrypted reference signal of the lead vehicle, without requiring global information, thus achieving fully distributed control.

[0066] A computer-readable storage medium storing a computer program that, when executed by a processor, enables a privacy-preserving vehicle platooning memory event-triggered adaptive fuzzy control method.

[0067] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, which, when executed by the processor, enables a privacy-preserving vehicle platooning memory event-triggered adaptive fuzzy control method.

[0068] Beneficial effects

[0069] The present invention provides a privacy-preserving vehicle formation memory event-triggered adaptive fuzzy control method. Compared with the prior art, this method uses a full-state time mask privacy protection mechanism, normalized collaborative learning adaptive fuzzy approximation, and memory event triggering mechanism to ensure that all following vehicles achieve bounded stable tracking of the lead vehicle's trajectory and maintain the desired formation configuration while protecting the initial sensitive state information.

[0070] The present invention also has the following advantages:

[0071] (1) To address the privacy leakage risk arising from the continuous exchange of vehicle position, speed, and acceleration status information in vehicle platooning systems during inter-vehicle communication, most existing distributed control methods directly transmit plaintext status information, leading to the risk of eavesdropping, inference, or even reconstruction of the initial sensitive state of vehicles. Furthermore, existing privacy protection methods are primarily geared towards general multi-agent systems and are difficult to directly apply to third-order nonlinear uncertain vehicle platooning scenarios. This invention designs a full-state time masking mechanism to construct recoverable masking functions for position, speed, and acceleration states respectively. Within the preset time window Internally in the form of encrypted fields Participating in workshop information exchange and completing the full backstepping controller design within the encrypted domain fundamentally protects the vehicle's initial sensitive information. When The post-masking function automatically degenerates into an identity transformation, no longer introducing additional dynamics, and strictly limits the impact of the encryption mechanism on control precision to a finite time period, breaking through the limitation of traditional privacy protection methods and control performance being difficult to balance.

[0072] (2) Existing vehicle platooning adaptive control methods typically design adaptive laws independently for each vehicle, failing to fully utilize the learning information from neighboring vehicles. This results in slow convergence speed of fuzzy parameters, low approximation accuracy, and a reliance on the assumption of bounded neighbor weights, making them overly conservative in complex network environments. This invention constructs a normalized collaborative learning adaptive fuzzy approximation framework, where the collaborative term in the adaptive law...

[0073]

[0074] The learning of fuzzy parameters for the current vehicle is accelerated by utilizing the weight estimation information of neighboring vehicles, while a normalization factor is used. With exponential decay factor Adaptive constraint of the magnitude of the collaborative term eliminates the need to pre-assume that the neighbor weights are bounded, effectively improving the convergence speed and efficiency of neighbor information utilization in distributed adaptive estimation, and reducing the dependence on precise model parameters.

[0075] (3) Traditional event-triggered control methods mostly use fixed trigger thresholds. The threshold design has limited adaptability to dynamic changes in the system. During periods of drastic changes in system state, the tracking accuracy may be sacrificed due to an excessively large threshold, while during steady-state periods, a large number of redundant triggers may occur due to an excessively small threshold. This invention introduces a memory event triggering mechanism, utilizing historical control data. With memory sensitivity Dynamically adjust trigger threshold The proposed memory event triggering mechanism maintains a high triggering frequency during periods of rapid system state change to ensure tracking accuracy, while adaptively reducing the triggering frequency during steady-state periods to conserve communication resources. Simulation results show that, compared with a fixed threshold event triggering strategy, this invention reduces the number of control updates and communication resource consumption while maintaining tracking performance, and avoids the Zeno phenomenon, thereby improving the engineering feasibility of the control method.

[0076] (4) This invention constructs a complete three-layer backstepping control architecture that matches a third-order nonlinear uncertain vehicle dynamics model. Virtual control laws are designed layer by layer within the encrypted domain. A second-order sliding mode integral filter is used to estimate the derivatives of the virtual control quantities to avoid complexity explosion, ultimately achieving seamless integration from high-level formation configuration planning to low-level accelerometer control. Based on Lyapunov stability theory, it is rigorously proven that the closed-loop system simultaneously satisfies: privacy protection, and the vehicle's initial sensitive state within... The internal structure is effectively protected; the signals are bounded, and all signals in the closed-loop system are semi-globally eventually consistent and bounded; the formation achieves practical consistency, and the position error satisfies... There is no Zeno phenomenon, and there is a positive minimum interval between adjacent trigger times.

[0077] Theoretical analysis and simulation verification show that the method proposed in this invention can achieve stable formation tracking in a formation system consisting of one lead vehicle and four follower vehicles under conditions of parameter uncertainty and unknown nonlinearity, while protecting the initial privacy information of the vehicles. The formation spacing error quickly converges to the bounded neighborhood after the initial transition phase. Compared with the fixed threshold strategy, the memory event triggering mechanism effectively saves communication resources, verifying the effectiveness and practicality of the control method. Attached Figure Description

[0078] Figure 1 This is a schematic diagram illustrating the sequence of the method of the present invention;

[0079] Figure 2 This is a flowchart of the vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to the present invention;

[0080] Figure 3 This is a schematic diagram of the privacy-protected vehicle platooning memory event-triggered adaptive fuzzy control system architecture of the present invention;

[0081] Figure 4 This is a schematic diagram of the communication topology of the vehicle formation in an embodiment of the present invention;

[0082] Figure 5 This is a simulation result diagram of the position trajectory of the lead vehicle and four following vehicles in an embodiment of the present invention;

[0083] Figure 6 Vehicle formation spacing error in embodiments of the present invention Evolutionary process diagram;

[0084] Figure 7 This is a comparison diagram of the encrypted and plaintext states of vehicle 1 under the proposed privacy protection mechanism in an embodiment of the present invention;

[0085] Figure 8 This is a comparison diagram of the encrypted and plaintext states of vehicle 3 under the proposed privacy protection mechanism in an embodiment of the present invention;

[0086] Figure 9 The speed masking error of the four following vehicles in this embodiment of the invention. Evolutionary diagram;

[0087] Figure 10 The actual control inputs for the four following vehicles in this embodiment of the invention. Simulation result diagram;

[0088] Figure 11 This is a distribution diagram of the triggering intervals of vehicles 1 and 3 under the memory event triggering mechanism in this embodiment of the invention;

[0089] Figure 12 This is a distribution diagram of the triggering intervals for vehicles 1 and 3 under the fixed threshold event triggering mechanism in this embodiment of the invention. Detailed Implementation

[0090] To provide a better understanding of the structural features and effects achieved by the present invention, a detailed description is provided below, accompanied by preferred embodiments and accompanying drawings:

[0091] Example 1

[0092] For vehicle platooning systems with privacy protection requirements, this invention proposes a privacy-preserving vehicle platooning memory-triggered adaptive fuzzy control method. This method implements finite-time encryption of position, velocity, and acceleration through a full-state time masking mechanism. Within the encrypted domain, a distributed virtual control law is constructed based on a backstepping method. A normalized collaborative learning fuzzy adaptive law is used to approximate the unknown nonlinear terms in the encrypted domain online. Furthermore, an event-triggered mechanism incorporating historical memory reduces the communication burden, ultimately generating a distributed collaborative control input that satisfies privacy constraints. Theoretical analysis and simulation verification show that the proposed method can achieve final convergence of platooning errors without revealing the privacy state of each following vehicle, and without exhibiting Zeno's phenomenon.

[0093] like Figure 1 and Figure 2 As shown, consider a vehicle platooning system consisting of one lead vehicle and four follower vehicles. The overall system architecture comprises five main parts: communication topology, vehicle dynamics model, encrypted domain equivalent system, fuzzy adaptive controller, and memory event triggering mechanism. The control flow includes six main steps: Step 11) Establishing the vehicle dynamics model; Step 12) Establishing the encrypted domain equivalent system model; Step 13) Designing the encrypted domain distributed virtual control law; Step 14) Designing the normalized collaborative learning adaptive law; Step 15) Constructing the memory update rule; and Step 16) Generating the final control input for each follower vehicle, achieving distributed collaborative control under privacy protection constraints. There is a clear progressive relationship between the steps, with the output of the previous step serving as the input for the subsequent step, ultimately achieving a complete mapping from the privacy-protected platooning configuration to the actual control input.

[0094] S1. Establish vehicle dynamics model

[0095] The establishment of the vehicle dynamics model involves: constructing a vehicle platooning communication topology and establishing a third-order nonlinear uncertain vehicle dynamics model; including the following steps:

[0096] S11. Constructing the vehicle platoon communication topology: A directed weighted graph is used to describe the communication topology of a platoon consisting of one lead vehicle and its platoon. The platoon communication structure consisting of following vehicles constructs an augmented matrix using an adjacency matrix, a Laplacian matrix, and a pinpoint matrix. The requirement is to expand the directed graph to include a directed spanning tree rooted at node 0 of the lead vehicle, ensuring that the lead vehicle's reference information is reachable by all following vehicles.

[0097] S12. Establish a third-order nonlinear uncertain longitudinal dynamic model: Establish the... A following vehicle Vertical position of time Longitudinal velocity Longitudinal acceleration Third-order dynamic model; ensembles the parameter uncertainties and unknown nonlinearities into an unknown function. ,in For the first A lumped unknown nonlinear function for a following vehicle, used to control the input. Drive acceleration subsystem, For the first The actual control input of the following vehicle.

[0098] S13. Define the desired formation configuration: based on the desired net spacing. With vehicle body length Define the expected cumulative distance between each following vehicle and the lead vehicle. Introducing offset position output ,in For the first Output the offset position of each following vehicle; transform the formation maintenance problem into the offset position output of each following vehicle relative to the position of the lead vehicle. The tracking issue.

[0099] like Figure 4 The diagram illustrates the vehicle platooning communication topology used in this embodiment. This topology consists of one lead vehicle and four follower vehicles. The lead vehicle node 0 directly sends reference information to follower vehicles 1 and 3. The information transmission relationship between the follower vehicles is as follows: follower vehicle 1 receives information from follower vehicle 4, follower vehicle 2 receives information from follower vehicle 1, follower vehicle 3 receives information from follower vehicle 2, and follower vehicle 4 simultaneously receives information from both follower vehicles 2 and 3. The arrows in the diagram indicate the direction of information transmission. The resulting extended directed graph contains a directed spanning tree rooted at the lead vehicle node 0, thus ensuring that the lead vehicle's reference information is reachable by all follower vehicles.

[0100] S2. Establish an equivalent system model for the encrypted domain.

[0101] The establishment of the equivalent system model in the encrypted domain is as follows: Based on a third-order nonlinear uncertain vehicle dynamics model, a full-state time masking privacy protection mechanism is designed to perform finite-time encryption on the position, velocity, and acceleration states, thereby establishing the equivalent system model in the encrypted domain; including the following steps:

[0102] S21, for the first Each state dimension of the following vehicle Define the time mask function:

[0103]

[0104] in, For the first The following vehicle The time mask function of the dimensional state, These correspond to the three state dimensions of position, velocity, and acceleration, respectively. For time, The duration of privacy protection defined for the user. It is a positive integer. The decay rate parameter can be designed; the function and its derivative are continuous at any time.

[0105] S22. Define the full-state encryption transformation:

[0106]

[0107] in, For the first The encryption status of the offset position of the following vehicle. , These are the encryption states for velocity and acceleration, respectively. For the first The longitudinal position of the following vehicle. For the first The original state variables are preserved; the navigator's reference signal is also encrypted, and an encrypted reference signal is defined. ,in Encrypt the reference signal for the lead vehicle. For the time mask function corresponding to the lead car, when The time mask function degenerates into an identity transformation, and the encrypted state is automatically restored to the plaintext state.

[0108] S23. Define the inverse mask. ,in For the first The following vehicle Using the inverse of the dimensional mask function, establish an equivalent system model for the cryptographic domain:

[0109]

[0110] in, Encrypted status Time derivative, , Given an auxiliary function, For terms containing unknown nonlinear terms Unknown auxiliary functions, To maintain control input for event triggering, Given a known bounded control gain, , These are the known upper bounds for vehicle mass and engine time lag, respectively. For the first The vehicle quality of the following vehicle For the first Engine lag of the following vehicle.

[0111] S3. Design a distributed virtual control law for the encrypted domain.

[0112] The design of the encrypted domain distributed virtual control law is as follows: based on the encrypted domain equivalent system model, the encrypted domain synchronization error and coordinate transformation are constructed, and the encrypted domain distributed virtual control law is designed based on the backstepping method; including the following steps:

[0113] S31. Define the encryption domain synchronization error and coordinate transformation:

[0114]

[0115] in, For the first Synchronization error of the first-layer encrypted domain of the vehicle-following vehicle fusion communication topology weighting. , For the first Layer coordinate transformation error For vehicles in the adjacency matrix To the vehicle Communication weight, For the navigator and the first in the fixed matrix The connection weight of the following vehicle , The first , The location of the following vehicle is encrypted. Encrypt the reference signal for the lead vehicle. For the first The following vehicle 3D encryption state, For the virtual control quantity to be designed.

[0116] S32. First step of the backstep design: Selecting the Lyapunov function ,in For the first step Lyapunov function, design the first-level virtual control law:

[0117]

[0118] in, For the first The first layer of virtual control variables for the following vehicle. , It is the inverse function of the mask. For the first The in-degree of the following vehicle in the communication topology, For design parameters, The time derivative of the encrypted state of neighboring vehicle locations. The time derivative of the encrypted reference signal for the navigator vehicle. mask function The time derivative.

[0119] S33, Second step backstep design: Using a second-order sliding mode integral filter for estimation. Choose the Lyapunov function ,in Design a second-level virtual control law for the second-step Lyapunov function:

[0120]

[0121] in, For the first The second-level virtual control quantity for the following vehicle. It is the inverse function of the 3rd dimension mask. For design parameters, For filter pairs The estimated output, mask function Time derivative, For the first The speed of the following vehicle is encrypted. This represents the error in the second-level coordinate transformation.

[0122] S4. Design a normalized collaborative learning adaptive law

[0123] The design of the normalized collaborative learning adaptive law is as follows: based on the distributed virtual control law of the cryptographic domain, a fuzzy logic system is introduced to approximate the unknown auxiliary function, and a normalized collaborative learning adaptive law is designed; including the following steps:

[0124] S41. Using fuzzy logic systems to handle unknown auxiliary functions in the encrypted domain Approximation:

[0125]

[0126] in, For the fuzzy system input vector, For the ideal weight vector, For fuzzy rule numbers, For fuzzy basis function vectors, For fuzzy approximation error, satisfy , To approximate the upper bound of the error.

[0127] S42. Select a Lyapunov function that includes the weight estimation error. ,in This is the error in the third-level coordinate transformation. For adaptive gain, For weight estimation error, For ideal weights Estimated value; Design nominal continuous control quantity:

[0128]

[0129] in, This is the third layer of nominal continuous control quantity. For the first The inverse function of the acceleration mask of the following vehicle, i.e. , For the first The inverse function of the speed mask of the following vehicle, i.e. , for The reverse, For design parameters, For the second-order sliding mode integral filter, the second-level virtual control quantity Output of the estimated time derivative.

[0130] S43. Design a normalized collaborative learning fuzzy weight adaptive law:

[0131]

[0132] in, Weight estimates Time derivative, For the first Fuzzy weight estimates of neighboring vehicles following each other. for The Euclidean norm, As the normalization factor, For the first The third-level coordinate transformation error of the following vehicle. It is an exponential decay factor. The weight decay coefficient is used; the normalization factor and the exponential decay factor together limit the magnitude of the synergistic term, without the need to presuppose that the neighbor weights are bounded.

[0133] S5. Construct memory update rules

[0134] The memory update rule is based on the nominal continuous control quantity. The design of a normalized continuous control signal and the construction of a memory event triggering mechanism and threshold memory update rules include the following steps:

[0135] S51. Define the normalized continuous control signal:

[0136]

[0137] in, For the first Normalized continuous control signals for following vehicles. This is the relative trigger threshold coefficient. Let be the shape parameter of the hyperbolic tangent function. This is the amplitude limiting parameter.

[0138] S52, Define Sample and Hold Error ,in For the first The sampling and holding error of the following vehicle; design a memory event triggering mechanism:

[0139]

[0140] in, For the first The following vehicle Next trigger moment For the next trigger time, It is a time-varying absolute threshold that maintains a piecewise constant value between adjacent threshold update times; To control the gain The known upper bound satisfies The design parameters must meet the following requirements. .

[0141] S53. Design threshold memory update rules: Introduce historical control data. With memory sensitivity :

[0142]

[0143] in, For the first The amount of historical control data for each following vehicle reflects the historical trend of control signal changes. This is a measure of memory sensitivity. This represents the number of historical data packets. For the weighting coefficients, satisfying , , For the first The normalized continuous control signal sample value corresponding to the next trigger time. For the first The historical sampled value corresponding to the next trigger time; when the condition is met At that time, update the threshold according to the rolling average rule:

[0144]

[0145] in, For the first The time-varying threshold after the next update For the first The threshold update time; when The timing mechanism degenerates into a fixed threshold event triggering; when The time memory mechanism comes into play; the richer the historical information, the stronger the threshold adaptability.

[0146] S6, Vehicle Formation Control

[0147] The vehicle formation control is based on a memory event triggering mechanism and the first... Normalized continuous control signal for following vehicle Generate the final control input for each following vehicle to achieve distributed collaborative control under privacy protection constraints; including the following steps:

[0148] S61. Generate the final control input for each following vehicle: based on the first-level virtual control quantity. Second-level virtual control quantity and nominal continuous control quantity According to the normalized continuous control signal After the memory event triggering mechanism determines the event trigger, it updates the event trigger to maintain the control input. In the untriggered zone Inside, Keep the sampled value from the previous trigger moment unchanged; the final implementation is as follows:

[0149]

[0150] S62. Keep control input triggered by the event. Mapped to , For the first The actual control input applied to the acceleration subsystem by the following vehicle; the control law only depends on the encrypted state of the vehicle itself, the encrypted state of the neighboring vehicles, and the encrypted reference signal of the lead vehicle, without requiring global information, thus achieving fully distributed control.

[0151] Simulation parameter settings

[0152] In this embodiment, a vehicle platooning system consisting of one lead vehicle and four follower vehicles is considered, and the simulation parameters are set as follows.

[0153] Vehicle parameters: mass kg, vehicle length m, expected vehicle spacing m, transmission efficiency Wheel radius m, aerodynamic drag coefficient Gravitational acceleration m / s², rolling resistance coefficient Actuator time constant s, the known upper bounds for vehicle mass and time delay are respectively taken as , .

[0154] Privacy protection parameters: Protection duration s, positive integer parameter Attenuation parameter ( ).

[0155] Controller parameters: Position layer gain Velocity layer gain Acceleration layer gain Hyperbolic tangent shape parameters Amplitude limiting parameters .

[0156] Fuzzy system parameters: number of fuzzy rules Adaptive gain Weight decay coefficient The fuzzy basis functions employ Gaussian membership functions, with their centers uniformly distributed at... The width is 1.5.

[0157] Event triggering parameters: relative threshold coefficient Initial absolute threshold Number of historical data packets Weighting coefficients .

[0158] The speed trajectory of the navigator is set as follows: Maintaining a constant speed of 20 m / s at time s, Smoothly decelerate to 16 m / s at time s, Accelerating to 21 m / s at s, and in The speed is kept constant at 21 m / s.

[0159] The initial values ​​of the system state variables are: vehicle position. Based on the expected vehicle spacing m setting, i.e. m, where m; initial velocity m / s; initial acceleration .

[0160] Simulation Results and Analysis

[0161] A simulation model was built in the MATLAB / Simulink environment to conduct simulation verification of a vehicle platooning system consisting of one lead vehicle and four follower vehicles. The simulation results are as follows: Figures 5 to 12 As shown.

[0162] like Figure 4 The diagram illustrates the communication topology considered in this embodiment, where the lead vehicle node 0 is connected to the four following vehicle nodes 1 to 4 via directed weighted edges, and the augmented matrix... The condition of having a directed spanning tree rooted at node 0 is satisfied, ensuring that the reference information of the lead vehicle is reachable by all following vehicles.

[0163] like Figure 5 As shown, the simulation results of the position trajectories of the lead vehicle and the four following vehicles are presented. Figure 5 It includes the longitudinal position curves of the lead car and four following cars. From Figure 5 It can be seen that during the privacy protection period ( The position status of each following vehicle is encrypted with a time mask before being used in workshop communication. The mask function degenerates into an identity transformation, and the encrypted state automatically reverts to the plaintext state; all four following vehicles can achieve the desired cumulative spacing. By tracking the position trajectory of the lead vehicle, the formation configuration remained stable throughout the entire motion, verifying the ability of the proposed control method to maintain the desired formation configuration.

[0164] like Figure 6 As shown, this illustrates the vehicle formation spacing error. The evolutionary process. Figure 6 It contains four colored curves, each corresponding to the spacing error of one of the four following vehicles. From Figure 6 It can be seen that the spacing error of each following vehicle, after exhibiting bounded fluctuations in the initial stage, quickly converged and eventually stabilized within a small range; at the end of the privacy protection period ( After that, the spacing error further approaches zero, verifying that the proposed method can achieve accurate formation spacing maintenance under privacy protection constraints.

[0165] like Figure 7 As shown, a comparison diagram of the encrypted and plaintext states of vehicle 1 under the proposed privacy protection mechanism is presented. Figure 7 The location encryption status of vehicle 1 Speed ​​encryption status and acceleration encryption state Compare each with the corresponding plaintext state. From Figure 7 It can be seen that, Stage, mask function This creates a significant difference between the encrypted and plaintext states, effectively preventing neighboring vehicles from inferring the initial state of vehicle 1 by intercepting communication information; when At this point, the masking function degenerates into an identity transformation, and the encrypted state completely overlaps with the plaintext state, thus protecting privacy. The subsequent control accuracy is not affected in any way.

[0166] like Figure 8 As shown, a comparison diagram of the encrypted and plaintext states of vehicle 3 under the proposed privacy protection mechanism is presented, along with the subgraph layout and... Figure 7 Same. From Figure 8 It can be seen that the encryption state evolution characteristics of vehicle 3 are consistent with those of vehicle 1, and the encryption state is in... The plaintext information was effectively masked during this period. The system automatically recovered afterward, verifying the consistent effectiveness of the time mask privacy protection mechanism across different following vehicles.

[0167] like Figure 9 The image shows the speed masking errors of the four following vehicles. Evolutionary diagram. Figure 9 It includes the speed masking error curves of four following vehicles. From Figure 9 It can be seen that the velocity mask error is in The phase decays monotonically over time, reflecting the mask function. A dynamic process that gradually approaches 1; when At that time, the mask error converged precisely to zero, and the encrypted speed state of each following vehicle was completely restored to the plaintext speed state, verifying the recoverability of the encryption mechanism.

[0168] like Figure 10 The diagram shows the actual control inputs of the four following vehicles. Simulation results diagram. Figure 10 It includes acceleration control signal curves for four following vehicles. From Figure 10It can be seen that the control inputs of each following vehicle exhibit bounded fluctuations in the initial stage, reflecting the effect of the normalized cooperative fuzzy control law on the unknown auxiliary function in the encrypted domain. The online approximation process; with fuzzy weights The control input gradually converges and becomes stable; the control input remains bounded throughout the process, with no obvious chattering phenomenon, verifying the effectiveness of the hyperbolic tangent boundary layer function in suppressing sliding mode chattering.

[0169] like Figure 11 As shown, the trigger interval distribution diagram of vehicle 1 and vehicle 3 under the memory event triggering mechanism is displayed. Figure 11 The table below presents the event trigger time sequences and trigger interval distribution histograms for vehicles 1 and 3 throughout the entire simulation period. Figure 11 It can be seen that the memory mechanism makes the time-varying absolute threshold... Able to control the amount of historical data With memory sensitivity Adaptive adjustment: During the initial formation establishment phase and when the speed of the lead vehicle changes abruptly, the trigger interval is small to ensure tracking accuracy; after the system enters steady state, the trigger interval increases significantly, effectively reducing redundant communication; there is always a positive minimum interval between two adjacent trigger moments, rigorously proving that Zeno's phenomenon does not exist.

[0170] like Figure 12 As shown, a fixed threshold event triggering mechanism is illustrated. The trigger interval distribution diagram for vehicles 1 and 3 is shown below. Figure 11 Forming a contrast. From Figure 12 It can be seen that the trigger interval distribution of the fixed threshold strategy is relatively uniform, and it cannot dynamically and adaptively adjust the trigger frequency according to the system; compared with Figure 11 In comparison, while ensuring similar control performance, the memory event triggering mechanism ( Compared with the fixed threshold strategy, it significantly reduces the total number of triggers, effectively reduces the network communication burden, and verifies the superiority of the memory mechanism in introducing historical sampling information to dynamically update the threshold.

[0171] Comprehensive comparative analysis: Simulation results show that the proposed method has the following performance: (1) Effective privacy protection: all following vehicles are in During this period, effective encryption is implemented for the initial sensitive state. (1) Automatic decryption afterward, without affecting long-term formation control performance; (2) Formation error convergence: the distance error between each following vehicle eventually converges to a bounded range, achieving accurate maintenance of the desired formation configuration; (3) Bounded control signal: the control input remains bounded throughout the process without jitter, and the second-order sliding mode integral filter effectively avoids the complexity explosion problem; (4) Advantage of memory mechanism: compared with the fixed threshold strategy, the memory event triggering mechanism effectively saves communication resources while ensuring similar control accuracy; (5) No Zeno phenomenon: there is a positive minimum interval between adjacent triggering times, and the control system can be physically implemented.

[0172] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A privacy-preserving vehicle platooning memory event-triggered adaptive fuzzy control method, characterized in that, Includes the following steps: 11) Establish a vehicle dynamics model; 12) Establish an equivalent system model for the encrypted domain; 13) Design a distributed virtual control law for the encrypted domain; 14) Design a normalized collaborative learning adaptive law: Introduce a fuzzy logic system to approximate the unknown nonlinear terms in the encrypted domain, and design a normalized collaborative learning adaptive law; 15) Construct memory update rules: Design normalized continuous control signals, construct memory event triggering mechanisms and threshold memory update rules; 16) Vehicle platoon control: Generate the final control input for each following vehicle to achieve distributed collaborative control under privacy protection constraints.

2. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 1, characterized in that, The establishment of the vehicle dynamics model involves: constructing a vehicle platooning communication topology and establishing a third-order nonlinear uncertain vehicle dynamics model; including the following steps: 21) Constructing the vehicle platoon communication topology: A directed weighted graph is used to describe the communication topology of a vehicle platoon consisting of one lead vehicle and its platoon. The platoon communication structure consisting of following vehicles constructs an augmented matrix using an adjacency matrix, a Laplacian matrix, and a pinpoint matrix. ,in To augment the topology matrix, the extended directed graph must contain a directed spanning tree rooted at node 0 of the lead vehicle, ensuring that the lead vehicle's reference information is reachable by all following vehicles. 22) Establish a third-order nonlinear uncertain longitudinal dynamic model: establish the first... A following vehicle Vertical position of time Longitudinal velocity Longitudinal acceleration Third-order dynamic model; ensembles the parameter uncertainties and unknown nonlinearities into an unknown function. ,in For the first A lumped unknown nonlinear function for a following vehicle, used to control the input. Drive acceleration subsystem, For the first The actual control input of the following vehicle; 23) Define the desired formation configuration: based on the desired net spacing. With vehicle body length Define the expected cumulative distance between each following vehicle and the lead vehicle. Introducing offset position output ,in For the first Output the offset position of each following vehicle; transform the formation maintenance problem into the offset position output of each following vehicle relative to the position of the lead vehicle. The tracking issue.

3. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 2, characterized in that, The establishment of the equivalent system model in the encrypted domain is as follows: Based on a third-order nonlinear uncertain vehicle dynamics model, a full-state time masking privacy protection mechanism is designed to perform finite-time encryption on the position, velocity, and acceleration states, thereby establishing the equivalent system model in the encrypted domain; including the following steps: 31) is the first Each state dimension of the following vehicle Define the time mask function: ； in, For the first The following vehicle The time mask function of the dimensional state, These correspond to the three state dimensions of position, velocity, and acceleration, respectively. For time, The duration of privacy protection defined for the user. It is a positive integer. The decay rate parameter can be designed; the function and its derivative are continuous at any time. 32) Define the full-state encryption transformation: ； in, For the first The encryption status of the offset position of the following vehicle. These are the encryption states for velocity and acceleration, respectively. For the first The longitudinal position of the following vehicle. For the first The original state variables are defined; encryption processing is applied to the reference signal of the navigator, and the encrypted reference signal is defined. ,in Encrypt the reference signal for the lead vehicle. For the time mask function corresponding to the lead car, when The time mask function degenerates into the identity transformation, and the encrypted state is automatically restored to the plaintext state. 33) Define the inverse mask ,in For the first The following vehicle Using the inverse of the dimensional mask function, establish an equivalent system model for the cryptographic domain: ； in, Encrypted status Time derivative, , Given an auxiliary function, For terms containing unknown nonlinear terms Unknown auxiliary functions, To maintain control input for event triggering, Given a known bounded control gain, , These are the known upper bounds for vehicle mass and engine time lag, respectively. For the first The vehicle quality of the following vehicle For the first Engine lag of the following vehicle.

4. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 3, characterized in that, The design of the encrypted domain distributed virtual control law is as follows: based on the encrypted domain equivalent system model, the encrypted domain synchronization error and coordinate transformation are constructed, and the encrypted domain distributed virtual control law is designed based on the backstepping method; including the following steps: 41) Define the encryption domain synchronization error and coordinate transformation: ， in, For the first Synchronization error of the first-layer encrypted domain of the vehicle-following vehicle fusion communication topology weighting. , For the first Layer coordinate transformation error For vehicles in the adjacency matrix To the vehicle Communication weight, For the navigator and the first in the fixed matrix The connection weight of the following vehicle , The first , The location of the following vehicle is encrypted. Encrypt the reference signal for the lead vehicle. For the first The following vehicle 3D encryption state, For the virtual control quantity to be designed; 42) First step of the backstepping design: Selecting the Lyapunov function ,in For the first step Lyapunov function, design the first-level virtual control law: ， in, For the first The first layer of virtual control variables for the following vehicle. , It is the inverse function of the mask. For the first The in-degree of the following vehicle in the communication topology, For design parameters, The time derivative of the encrypted state of neighboring vehicle locations. The time derivative of the encrypted reference signal for the navigator vehicle. mask function The time derivative; 43) Second step backstep design: Use a second-order sliding mode integral filter for estimation. Choose the Lyapunov function ,in Design a second-level virtual control law for the second-step Lyapunov function: ， in, For the first The second-level virtual control quantity for the following vehicle. It is the inverse function of the 3rd dimension mask. For design parameters, For filter pairs The estimated output, mask function Time derivative, For the first The speed of the following vehicle is encrypted. This represents the error in the second-level coordinate transformation.

5. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 4, characterized in that, The design of the normalized collaborative learning adaptive law is as follows: based on the distributed virtual control law of the cryptographic domain, a fuzzy logic system is introduced to approximate the unknown auxiliary function, and a normalized collaborative learning adaptive law is designed; including the following steps: 51) Using fuzzy logic systems to handle unknown auxiliary functions in the encryption domain Approximation: ， in, For the fuzzy system input vector, For the ideal weight vector, For fuzzy rule numbers, For fuzzy basis function vectors, For fuzzy approximation error, satisfy , To approximate the upper bound of the error; 52) Select a Lyapunov function that includes the weight estimation error. ,in This is the error in the third-level coordinate transformation. For adaptive gain, For weight estimation error, For ideal weights Estimated value; Design nominal continuous control quantity: ， in, This is the third layer of nominal continuous control quantity. For the first The inverse function of the acceleration mask of the following vehicle, i.e. , For the first The inverse function of the speed mask of the following vehicle, i.e. , for The reverse, For design parameters, For the second-order sliding mode integral filter, the second-level virtual control quantity Output of the time derivative estimate; 53) Design a normalized collaborative learning fuzzy weight adaptive law: ， in, Weight estimates Time derivative, For the first Fuzzy weight estimates of neighboring vehicles following each other. for The Euclidean norm, As the normalization factor, For the first The third-level coordinate transformation error of the following vehicle. It is an exponential decay factor. The weight decay coefficient is used; the normalization factor and the exponential decay factor together limit the magnitude of the synergistic term, without the need to presuppose that the neighbor weights are bounded.

6. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 5, characterized in that, The memory update rule is based on the nominal continuous control quantity. The design of a normalized continuous control signal and the construction of a memory event triggering mechanism and threshold memory update rules include the following steps: 61) Define the normalized continuous control signal: ， in, For the first Normalized continuous control signals for following vehicles. This is the relative trigger threshold coefficient. Let be the shape parameter of the hyperbolic tangent function. This is the amplitude limiting parameter; 62) Define the sample-and-hold error ,in For the first The sampling and holding error of the following vehicle; design a memory event triggering mechanism: ， ， in, For the first The following vehicle Next trigger moment For the next trigger time, The threshold is a time-varying absolute threshold, maintaining a piecewise constant value between adjacent threshold update times; the design parameters must meet the following requirements. ; 63) Design threshold memory update rules: Introduce historical control data volume With memory sensitivity : ， in, For the first The amount of historical control data for each following vehicle reflects the historical trend of control signal changes. This is a measure of memory sensitivity. This represents the number of historical data packets. For the weighting coefficients, satisfying , , For the first The normalized continuous control signal sample value corresponding to the next trigger time. For the first The historical sampled value corresponding to the next trigger time; when the condition is met At that time, update the threshold according to the rolling average rule: ， in, For the first The time-varying threshold after the next update For the first The threshold update time; when The timing mechanism degenerates into a fixed threshold event triggering; when The time memory mechanism comes into play; the richer the historical information, the stronger the threshold adaptability.

7. The vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection according to claim 6, characterized in that, The vehicle formation control is based on a memory event triggering mechanism and the first... Normalized continuous control signal for following vehicle Generate the final control input for each following vehicle to achieve distributed collaborative control under privacy protection constraints; including the following steps: 71) Generate the final control input for each following vehicle: based on the first-level virtual control quantity Second-level virtual control quantity and nominal continuous control quantity According to the normalized continuous control signal After the memory event triggering mechanism determines the event trigger, it updates the event trigger to maintain the control input. In the untriggered zone Inside, Keep the sampled value from the previous trigger moment unchanged; the final implementation is as follows: ； 72) Keep control input triggered by the event Mapped to , For the first The actual control input applied to the acceleration subsystem by the following vehicle; the control law only depends on the encrypted state of the vehicle itself, the encrypted state of the neighboring vehicles, and the encrypted reference signal of the lead vehicle, without requiring global information, thus achieving fully distributed control.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, enables the implementation of the vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection as described in any one of claims 1-7.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it can implement the vehicle platooning memory event-triggered adaptive fuzzy control method under privacy protection as described in any one of claims 1-7.

Images