Vehicle cooperative fault-tolerant control method based on data consistency under zero-trust architecture
By introducing a zero-trust architecture and parity space method into the vehicle queue, evaluating vehicle trust and building a residual generator, the communication security and identity authentication problems in the vehicle queue are solved, and the coordinated fault tolerance control effect and security of the vehicle queue are improved. .
Patent Information
- Application Number
- CN202411004791.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-12
AI Technical Summary
The lack of effective collaborative perception mechanisms and identity authentication mechanisms in the vehicle queues leads to communication security and identity authentication issues, affecting information sharing and driving safety among vehicles.
Using a vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture, the residual generator is constructed through trust evaluation and parity space method combined with I/O data model, estimate the actuator failure and design compensation controllers to improve the vehicle queue Coordinated fault tolerance control effect.
It realizes the safety and trust accuracy of information transmission between vehicles, improves the safety and reliability of vehicle queues, enhances the detection and diagnosis of faults, and ensures the stability and efficiency of vehicle queues.
Smart Images

Figure CN118938862A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of Internet of Vehicles and zero trust, and specifically to a vehicle collaborative fault-tolerant control method based on data consistency under a zero trust architecture. Background Art
[0002] In recent years, intelligent transportation systems (ITS) have attracted much attention, and vehicle platoon control has become one of the important applications in ITS. A vehicle platoon is composed of a group of vehicles, including a lead vehicle and several follower vehicles, which travel in coordination, usually at the same speed and a short interval. Compared with traditional driving modes, platoon driving has the advantages of increasing the throughput of road vehicles, improving traffic efficiency, enhancing road safety and reducing fuel consumption.
[0003] Based on the driving mode of vehicle platoons, V2V communication in VANETs is used to realize information exchange between platoon vehicles. V2V communication enables vehicles to share current position, speed, acceleration, heading angle and other status information, quickly respond to changes in the behavior of the vehicle in front, improve the fault tolerance of the system, and ensure the rapidity and accuracy of decision-making.
[0004] In the collaborative control of vehicle platoons, secure communication is crucial, especially considering that communication usually occurs in open wireless channels. However, when the status information between vehicles in a vehicle platoon is transmitted through the network, it faces security issues such as network-induced delays, data packet loss caused by intruders or malicious nodes in the network, information leakage, and information tampering. These problems caused by the communication network will inevitably affect the communication between vehicles and undermine the performance of the entire control system. Zero trust architecture provides a solution to the above problems. Zero trust architecture effectively responds to security threats by strictly authenticating and encrypting each network request. Although there is no unified definition of the zero trust concept and its architecture, its core concept of never trusting and constantly verifying is consistent.
[0005] Fault-tolerant control of vehicle platoons refers to mechanisms designed to handle and mitigate the impact of various faults or abnormal conditions (such as vehicle communication failures, sensor errors, software failures, hardware failures, etc.) on the performance of vehicle platoons. However, current vehicle platoons lack an effective collaborative perception mechanism and cannot fully utilize the state information and environmental information of surrounding vehicles to improve driving safety and efficiency. In addition, communications between vehicles may be affected by eavesdropping or tampering, and there is a lack of effective identity authentication mechanisms. Therefore, it is necessary to strengthen information sharing between vehicles, improve communication security and identity authentication, and design effective privacy protection mechanisms. Summary of the invention
[0006] In view of the deficiencies of the above-mentioned prior art, the technical problem to be solved by the present invention is: how to provide a vehicle collaborative fault-tolerant control method based on data consistency under a zero-trust architecture, firstly, a zero-trust architecture is introduced into the vehicle queue to ensure the security of information transmission and interaction; secondly, the trust of the vehicle is determined based on the consistency of the output data to ensure the accuracy of the trust; then, a residual generator is constructed by combining the I / O data model with the parity space method, and the fault estimation value of the actuator of the residual generator is estimated to assist the collaborative fault-tolerant control of the vehicle queue; finally, a compensation-based collaborative fault-tolerant controller is designed by estimating the trust of the vehicle and the fault estimation value of the actuator, thereby improving the effect of vehicle collaborative fault-tolerant control.
[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:
[0008] The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture includes:
[0009] S1: Under the zero-trust architecture, the trust of each vehicle in the vehicle queue towards other vehicles is evaluated based on the data consistency evaluation method;
[0010] S2: construct system state space equations for the vehicle platoon and reconstruct the system state space equations into an I / O data model;
[0011] S3: Use the parity space method to build the corresponding residual generator in combination with the I / O data model;
[0012] S4: Calculate the fault estimate of the actuator by combining the output of the residual generator through an iterative learning method;
[0013] S5: For each vehicle in the vehicle queue, a compensation-based cooperative fault-tolerant controller is constructed according to the trust evaluation results of other vehicles and the fault estimation value of the actuator;
[0014] S6: For each vehicle in the vehicle queue, a corresponding control input is generated based on its cooperative fault-tolerant controller to cooperatively adjust the operating state of the vehicle.
[0015] Preferably, in step S1, the trust of vehicle i in vehicle j is evaluated by the following steps:
[0016] Sl01: Calculate the behavior information consistency trust Tr based on the consistency of lane change frequency, acceleration stability and deceleration stability between vehicle i and vehicle j A(i,j) ;
[0017] S102: Set the trust threshold γ1, and calculate the trust Tr of the consistency of the behavior information between vehicle i and vehicle j by using the following formula: A(i,j) is preprocessed and the behavior information consistency trust Tr A(i,j) Convert to 0 or 1;
[0018]
[0019] S103: Obtaining the state data sequence X of vehicle i itself i ={x i,1 , x i,2 , ..., x i,n} and the state data sequence X received by vehicle i from vehicle j (i,j) ={x (i,j),1 , x (i,j),2 , ..., x (i,j),n};
[0020] S104: Calculate state data sequence X i and X (i,j) Corresponding element difference Δ (0,1),i =|x 0,i -x (0,1),i |, forming a difference sequence Z (0,1) ={Δ (0,1),1 , Δ (0,1),2 , ..., Δ (0,1),n};
[0021] S105: Calculate the node status information consistency trust Tr between vehicle i and vehicle j based on the difference sequence B(i,j) ;
[0022] S106: Setting the trust threshold γ 2 , combined with the following formula, the node status information consistency trust Tr between vehicle i and vehicle j B(i,j) Perform preprocessing to make the node status information consistent with the trust degree Tr B(i,j) Convert to 0 or 1;
[0023]
[0024] S107: The consistency trust Tr of the behavior information between the pre-processed vehicles i and j A(i,j) The trust degree Tr of the node status information is consistent with B(i,j) Evaluate the trust of vehicle i in vehicle j;
[0025] The formula is:
[0026] γ ij =Tr A(i,j) ·Tr B(i,j) ;
[0027] Where: γ ij It represents the trust of vehicle i to vehicle j.
[0028] Preferably, in step S101, the behavior information consistency trust Tr between vehicle i and vehicle j is calculated by the following formula: A(i,j) :
[0029] T A(i,j) =w 1 ×F C(i,j) +w 2 ×A S(i,j) +w 3 ×D s(i,j) ;
[0030] F C(i,j) =|F Ci -F Cj |;
[0031] A S(i,j) =|A Si -A Sj |;
[0032] D S(i,j) =|D Si -D Sj |;
[0033]
[0034]
[0035]
[0036] Where: w 1 、w 2 、w 3 is the weight coefficient; F C(i,j) , A S(i,j) , D S(i,j) They represent the consistency of lane change frequency, acceleration stability and deceleration stability between vehicle i and vehicle j respectively; F Ci , A Si , D Si They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle i respectively; F Cj , A Sj , D Sj They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle j respectively; a i represents the acceleration value measured at the i-th time; d i represents the deceleration value of the i-th measurement; n represents the number of measurements.
[0037] Preferably, in step S105, the node status information consistency trust Tr between vehicle i and vehicle j is calculated by the following formula: B(i,j) :
[0038]
[0039]
[0040] Where a1 is a parameter used to control the speed at which trust decreases; n is the number of element differences in the difference sequence.
[0041] Preferably, in step S2, the system state space equation of the vehicle queue is expressed as:
[0042]
[0043] Where: x i (k) is the system status; y i (k) is the system output; u i (k) is the system input; A, B, C, D, E are the system parameter matrices; f i (k) is the fault input; w(k) is the process noise; v(k) is the measurement noise.
[0044] Preferably, in step S2, the vehicle input is used as the input of the I / O data model, and the acceleration information of the vehicle is used as the output of the I / O data model;
[0045] The reconstructed I / O data model is expressed as:
[0046] Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ;
[0047]
[0048]
[0049]
[0050] X(k)=[x(k)x(k+1)…x(k+N-1)];
[0051]
[0052] Preferably, in step S3, a residual generator is constructed by the following steps:
[0053] In step S3, the residual generator is constructed by the following steps:
[0054] S301: Extract Y from the I / O data model f , Uf , Y p , U p ;
[0055] Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ;
[0056]
[0057]
[0058]
[0059] S302: Obtaining Data Build And perform SVD decomposition;
[0060] S303: Based on The SVD decomposition results of constructing an equivalent space and the correlation matrix
[0061] S304: Calculation Get the maximum eigenvalue λ s,max and the eigenvector l s,max ;
[0062] S305: Constructing robust equivalent vectors and
[0063] S306: Construct a residual generator: r(k)=v s y x (k)-β s u s (k).
[0064] Preferably, in step S4, fault estimation is achieved through the following steps:
[0065] S401: Obtain residual generator r(k)=v s y s (k)-β s u s (k);
[0066] S402: Introduce the L2 regularization term and construct the objective function based on the residual generator;
[0067] The objective function is expressed as:
[0068]
[0069] Where: r represents the residual of the system; Q represents the design matrix; represents the estimated value of the fault signal; ||·|| 2 represents the square of the two norm (L2 norm); λ is the regularization parameter, which controls the weight of the regularization term;
[0070] S403: splitting the objective function into two parts: a residual error term and a regularization term;
[0071] The residual error term is expressed as:
[0072]
[0073] The regularization term is expressed as:
[0074]
[0075] S404: solving the gradients of the residual error term and the regularization term respectively;
[0076] 1) Gradient of the residual error term J1(f):
[0077] 2) Gradient of the regularization term J2(f):
[0078] S405: merging and simplifying the gradient of the residual error term and the regularization term;
[0079] The formula is:
[0080]
[0081] To make the gradient zero, solve for the fault estimate f:
[0082]
[0083] Simplified to:
[0084] S406: Generate a closed-form solution for the fault estimate;
[0085] The formula is:
[0086]
[0087] Where: Q represents the design matrix; r represents the residual of the system; λ represents the regularization parameter; I is the unit matrix, which is used to ensure that Q T Q+λ, reversible.
[0088] Preferably, in step S406, the fault estimation value is iteratively solved using a gradient descent method;
[0089] The formula is:
[0090]
[0091] Where: η represents the learning rate; λ represents the regularization parameter; r(k) represents the residual.
[0092] Preferably, in step S5, the controller of vehicle i is represented by the following formula:
[0093]
[0094] Where: x i1 represents the first state quantity of vehicle i, i.e., position; x i2 represents the second state quantity of vehicle i, namely speed; x i3 The third state quantity x represents the acceleration of vehicle i; j1 represents the first state of car j; x j2 represents the second state of car j; x j3 Represents the third state quantity of car j; K p , K v , K a is the controller gain; d represents the distance between vehicles in the vehicle queue; γ ij B is the trust of vehicle i in vehicle j; * is the parameter matrix to be designed; Indicates the fault estimation result.
[0095] Compared with the prior art, the vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture of the present invention has the following beneficial effects:
[0096] The present invention introduces a zero-trust architecture into the vehicle platoon, so that the information transmission and interaction between vehicles in the vehicle network need to meet the requirements of the zero-trust architecture, that is, they need to undergo strict identity authentication and authorization. This means that communication between vehicles requires real-time verification of the identities of participating entities, including vehicles, users, and data, to ensure that only legitimate entities can access corresponding resources. By dividing the vehicle network into multiple isolated segments, each segment serves as an independent security area, thereby reducing the potential attack surface. Even if a segment is attacked, it is difficult for the attacker to move laterally to other segments, thereby improving the security of vehicle platoon operation.
[0097] The present invention evaluates the trust of each vehicle in a vehicle queue for other vehicles based on a data consistency evaluation method. First, by comparing the consistency of the output data of the vehicle in multiple dimensions (including lane change frequency, acceleration change, deceleration stability and state), the trust of the vehicle can be evaluated more comprehensively. When the output data of multiple vehicles show a high degree of consistency under the same or similar conditions, the trust of these vehicles can be improved; at the same time, when the output data of a certain vehicle is significantly inconsistent with that of other vehicles, it is often a signal that the vehicle has abnormal behavior, and then these abnormal vehicles can be discovered and marked in time, thereby reducing the trust in them and improving the accuracy of the evaluation. Secondly, the output data of the vehicle involves data in multiple dimensions, such as the lane change frequency reflects the driving habits and stability of the vehicle, the acceleration change reveals the acceleration and braking performance of the vehicle, and the deceleration stability is directly related to the safety performance of the vehicle. The integration of multiple dimensional data makes the trust evaluation more comprehensive and accurate. Finally, by obtaining the output data between vehicles in real time, the trust of each vehicle can be dynamically evaluated, and the changes in the vehicle status can be reflected in time, ensuring the timeliness of the trust evaluation.
[0098] The present invention reconstructs the system state space equation into an I / O data model, and then constructs a corresponding residual generator by combining the I / O data model with the parity space method. First, the complex system state space equation is converted into an intuitive input-output relationship, namely the I / O data model, which simplifies the fault analysis process and makes fault location and diagnosis more efficient; at the same time, the I / O data model directly associates the system input and output, making the monitoring of the system state and fault detection more intuitive. Secondly, on the basis of the I / O data model, the parity space method is combined. The parity space method allows the extraction of fault-related features from the mapping of system input to output. Therefore, the residual generator constructed by the parity space can more directly reflect the system abnormality, thereby improving the sensitivity and accuracy of fault detection. Finally, the parity space method has strong real-time performance and can be designed to be suitable for real-time fault detection systems, allowing rapid calculation of residuals and providing timely feedback to the system; at the same time, the parity space method generates residuals by designing a residual generator, and these residuals have good robustness to internal and external disturbances of the system, which means that even when the system is subject to noise or other types of interference, the method can still effectively detect faults.
[0099] The present invention, based on the construction of a residual generator, calculates the fault estimation value of the actuator through an iterative learning method combined with the output of the residual generator. First, the iterative learning method can gradually improve the accuracy of fault estimation through continuous learning and optimization, thereby improving the adaptability and robustness of the system. Secondly, in the face of various complex environments and operating conditions that may occur in the vehicle queue, the iterative learning method can adaptively adjust the fault estimation model, thereby improving the robustness of the system under different conditions; at the same time, the iterative learning method supports online learning and adjustment, so that the system can continuously optimize the fault estimation model according to the actual operating conditions, and improve the adaptability and intelligence level of the system. Finally, through real-time actuator fault estimation, it is helpful to dynamically adjust the control strategy to ensure that the stability of the queue can be maintained even in complex and changing environments.
[0100] On the basis of trust evaluation and actuator fault estimation, the present invention further constructs a compensation-based collaborative fault-tolerant controller for each vehicle in the vehicle queue, and then collaboratively adjusts the vehicle's operating state based on the control input generated by the collaborative fault-tolerant controller. First, the collaborative fault-tolerant controller constructed in combination with the trust evaluation and the actuator fault estimation value can more accurately identify and compensate for the impact of the faulty vehicle, implement a more accurate fault-tolerant control strategy, and thus improve the effect of vehicle collaborative fault-tolerant control. Secondly, the compensation-based collaborative fault-tolerant controller can flexibly adjust the control strategy according to different fault types and degrees, compensate and adjust the control inputs of other vehicles in a collaborative manner, make up for the functional loss of the faulty vehicle, maintain the coordination and stability of the entire vehicle queue, and adapt to different fault scenarios and requirements. Finally, the introduction of the collaborative fault-tolerant controller enables the system to maintain the normal operation of the overall function when facing individual vehicle failures, thereby improving the reliability and availability of the entire vehicle queue. BRIEF DESCRIPTION OF THE DRAWINGS
[0101] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:
[0102] Figure 1 and Figure 2 It is a logic block diagram and flow chart of the vehicle cooperative fault-tolerant control method;
[0103] Figure 3 is a flow chart of a trust evaluation method based on data consistency;
[0104] Figure 4 Flowchart for generating residual generator;
[0105] Figure 5 Flowchart of the fault estimation method. DETAILED DESCRIPTION
[0106] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but only represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.
[0107] The following is a further detailed description through specific implementation methods:
[0108] Example:
[0109] This embodiment discloses a vehicle collaborative fault-tolerant control method based on data consistency under a zero-trust architecture.
[0110] like Figure 1 and Figure 2 As shown, the vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture includes:
[0111] S1: Under the zero-trust architecture, the trust of each vehicle in the vehicle queue towards other vehicles is evaluated based on the data consistency evaluation method; under the zero-trust architecture, it means that the information transmission and interaction between vehicles must meet the requirements of the zero-trust architecture, that is, every transmission and interaction between vehicles requires identity authentication.
[0112] S2: construct system state space equations for the vehicle platoon and reconstruct the system state space equations into an I / O data model;
[0113] S3: Use the parity space method to build the corresponding residual generator in combination with the I / O data model;
[0114] S4: Calculate the fault estimate of the actuator by combining the output of the residual generator through an iterative learning method;
[0115] S5: For each vehicle in the vehicle queue, a compensation-based cooperative fault-tolerant controller is constructed according to the trust evaluation results of other vehicles and the fault estimation value of the actuator;
[0116] S6: For each vehicle in the vehicle queue, a corresponding control input is generated based on its cooperative fault-tolerant controller to cooperatively adjust the operating state of the vehicle.
[0117] The present invention introduces a zero-trust architecture into the vehicle platoon, so that the information transmission and interaction between vehicles in the vehicle network need to meet the requirements of the zero-trust architecture, that is, they need to undergo strict identity authentication and authorization. This means that communication between vehicles requires real-time verification of the identities of participating entities, including vehicles, users, and data, to ensure that only legitimate entities can access corresponding resources. By dividing the vehicle network into multiple isolated segments, each segment serves as an independent security area, thereby reducing the potential attack surface. Even if a segment is attacked, it is difficult for the attacker to move laterally to other segments, thereby improving the security of vehicle platoon operation.
[0118] The present invention evaluates the trust of each vehicle in a vehicle queue for other vehicles based on a data consistency evaluation method. First, by comparing the consistency of the output data of the vehicle in multiple dimensions (including lane change frequency, acceleration change, deceleration stability and state), the trust of the vehicle can be evaluated more comprehensively. When the output data of multiple vehicles show a high degree of consistency under the same or similar conditions, the trust of these vehicles can be improved; at the same time, when the output data of a certain vehicle is significantly inconsistent with that of other vehicles, it is often a signal that the vehicle has abnormal behavior, and then these abnormal vehicles can be discovered and marked in time, thereby reducing the trust in them and improving the accuracy of the evaluation. Secondly, the output data of the vehicle involves data in multiple dimensions, such as the lane change frequency reflects the driving habits and stability of the vehicle, the acceleration change reveals the acceleration and braking performance of the vehicle, and the deceleration stability is directly related to the safety performance of the vehicle. The integration of multiple dimensional data makes the trust evaluation more comprehensive and accurate. Finally, by obtaining the output data between vehicles in real time, the trust of each vehicle can be dynamically evaluated, and the changes in the vehicle status can be reflected in time, ensuring the timeliness of the trust evaluation.
[0119] The present invention reconstructs the system state space equation into an I / O data model, and then constructs a corresponding residual generator by combining the I / O data model with the parity space method. First, the complex system state space equation is converted into an intuitive input-output relationship, namely the I / O data model, which simplifies the fault analysis process and makes fault location and diagnosis more efficient; at the same time, the I / O data model directly associates the system input and output, making the monitoring of the system state and fault detection more intuitive. Secondly, on the basis of the I / O data model, the parity space method is combined. The parity space method allows the extraction of fault-related features from the mapping of system input to output. Therefore, the residual generator constructed by the parity space can more directly reflect the system abnormality, thereby improving the sensitivity and accuracy of fault detection. Finally, the parity space method has strong real-time performance and can be designed to be suitable for real-time fault detection systems, allowing rapid calculation of residuals and providing timely feedback to the system; at the same time, the parity space method generates residuals by designing a residual generator, and these residuals have good robustness to internal and external disturbances of the system, which means that even when the system is subject to noise or other types of interference, the method can still effectively detect faults.
[0120] The present invention, based on the construction of a residual generator, calculates the fault estimation value of the actuator through an iterative learning method combined with the output of the residual generator. First, the iterative learning method can gradually improve the accuracy of fault estimation through continuous learning and optimization, thereby improving the adaptability and robustness of the system. Secondly, in the face of various complex environments and operating conditions that may occur in the vehicle queue, the iterative learning method can adaptively adjust the fault estimation model, thereby improving the robustness of the system under different conditions; at the same time, the iterative learning method supports online learning and adjustment, so that the system can continuously optimize the fault estimation model according to the actual operating conditions, and improve the adaptability and intelligence level of the system. Finally, through real-time actuator fault estimation, it is helpful to dynamically adjust the control strategy to ensure that the stability of the queue can be maintained even in complex and changing environments.
[0121] On the basis of trust evaluation and actuator fault estimation, the present invention further constructs a compensation-based collaborative fault-tolerant controller for each vehicle in the vehicle queue, and then collaboratively adjusts the vehicle's operating state based on the control input generated by the collaborative fault-tolerant controller. First, the collaborative fault-tolerant controller constructed in combination with the trust evaluation and the actuator fault estimation value can more accurately identify and compensate for the impact of the faulty vehicle, implement a more accurate fault-tolerant control strategy, and thus improve the effect of vehicle collaborative fault-tolerant control. Secondly, the compensation-based collaborative fault-tolerant controller can flexibly adjust the control strategy according to different fault types and degrees, compensate and adjust the control inputs of other vehicles in a collaborative manner, make up for the functional loss of the faulty vehicle, maintain the coordination and stability of the entire vehicle queue, and adapt to different fault scenarios and requirements. Finally, the introduction of the collaborative fault-tolerant controller enables the system to maintain the normal operation of the overall function when facing individual vehicle failures, thereby improving the reliability and availability of the entire vehicle queue.
[0122] 1. Trust Calculation
[0123] The trust evaluation method based on output data consistency is an algorithm that uses vehicle output data to perform trust evaluation. Due to the consistency of vehicle queues, each vehicle shares its own driving intention, maintains a unified safety distance and travels at the same speed. Therefore, the present invention calculates the trust by comparing the output state information of different vehicles within a unit time.
[0124] Based on the consistency check of vehicle behavior information. First, each vehicle node needs to collect data including lane change, acceleration, deceleration and other behavior states. After the data is collected, necessary preprocessing is performed to ensure the quality and consistency of the data. Key features are extracted from these data, including lane change frequency (Frequency Consistency, F C ), Acceleration Stability, A S ) and Deceleration Smoothness (Deceleration Smoothness, D S ), and use these features to establish a vehicle behavior model, which can be a statistical model for analyzing and predicting vehicle behavior patterns.
[0125] After the model is established, consistency checks are performed, which include time series analysis and behavior pattern matching. Time series analysis is used to check whether the vehicle behavior is consistent over time, and behavior pattern matching compares the actual behavior with the behavior predicted by the model. The following is a specific trust evaluation algorithm:
[0126] a) Data collection and preprocessing
[0127] (1) Data collection: Collect vehicle behavior status data such as lane changing, acceleration, and deceleration.
[0128] (2) Data cleaning: remove outliers and noise to ensure data quality.
[0129] (3) Data standardization: Convert data to a uniform dimension and range for easy comparison.
[0130] b) Key feature extraction
[0131] (1) Lane change frequency: Calculate the number of lane changes per unit time.
[0132] (2) Acceleration change: measure the fluctuation amplitude of acceleration.
[0133] (3) Deceleration stability: evaluates the smoothness during deceleration.
[0134] c) Build behavioral models
[0135] (1) Statistical model: Use statistical methods (such as regression analysis) to establish vehicle behavior models.
[0136] (2) Feature weight: Determine the weight of each feature’s impact on behavior prediction.
[0137] d)Consistency check
[0138] (1) Time series analysis: Analyze the consistency of vehicle behavior in time series.
[0139] (2) Behavioral pattern matching: comparing actual behavioral data with the behavior predicted by the model.
[0140] In this embodiment, combined with Figure 3 As shown, the trust of vehicle i for vehicle j is evaluated through the following steps:
[0141] S1O1: Calculate the behavior information consistency trust Tr by the consistency of lane change frequency, acceleration stability and deceleration stability between vehicle i and vehicle j (i, j represents vehicle number i, j∈{0, 1, 2, ..., N}) A(i,j) ;
[0142] Among them, the consistency trust degree Tr between vehicle i and vehicle j is calculated by the following formula: A(i,j) ;
[0143] T A(i,j) =w 1 ×F C(i,j) +w 2 ×A S(i,j) +w 3 ×D S(i,j) ;
[0144] F C(i,j) =|FCi -F Cj |;
[0145] A S(i,j) =|A Si -A Sj |;
[0146] D S(i,j) =|D Si -D Sj |;
[0147]
[0148]
[0149]
[0150] Where: w 1 、w 2 、w 3 is the weight coefficient; F C(i,j) , A S(i,j)、 D S(i,j) They represent the consistency of lane change frequency, acceleration stability and deceleration stability between vehicle i and vehicle j respectively; F Ci , A Si , D Si They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle i respectively; F Cj , A Sj , D Sj They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle j respectively; a i represents the acceleration value measured at the i-th time; d i represents the deceleration value of the i-th measurement; n represents the number of measurements.
[0151] S102: Setting the trust threshold γ 1 , combined with the following formula, the trust degree Tr of the behavior information consistency between vehicle i and vehicle j A(i,j) Preprocessing is performed to convert the behavior information consistency trust Tr A(i,j) ) is converted to 0 or 1;
[0152]
[0153] S103: Within a certain time period T (dividing T into n equal parts), obtain the state data sequence X of vehicle i itself i ={x i,1 , x i,2 , ..., x i,n} and the state data sequence X received by vehicle i from vehicle j (i,j) ={x (i,j),1, x (i,j),2 , ..., x (i,j),n};
[0154] S104: Calculate state data sequence X i and X (i,j) Corresponding element difference Δ (0,1),i =|x 0,i -x (0,1),i |, forming a difference sequence Z (0,1) ={Δ (0,1),1 , Δ (0,1),2 , ..., Δ (0,1),n};
[0155] S105: Calculate the node status information consistency trust Tr between vehicle i and vehicle j based on the difference sequence B(i,j) ;
[0156] Among them, the node state information consistency trust Tr between vehicle i and vehicle j is calculated by the following formula: B(i,j) ;
[0157]
[0158]
[0159] Where a1 is a parameter used to control the speed at which trust decreases; n is the number of element differences in the difference sequence.
[0160] S106: Set the trust threshold γ2, and calculate the trust Tr of the node status information consistency between vehicle i and vehicle j by combining the following formula: B(i,j) Perform preprocessing to make the node status information consistent with the trust degree Tr B(i,j) Convert to 0 or 1;
[0161]
[0162] S107: The consistency trust Tr of the behavior information between the pre-processed vehicles i and j A(i,j) The trust degree Tr of the node status information is consistent with B(i,j) Evaluate the trust of vehicle i in vehicle j;
[0163] The formula is:
[0164] γ ij =Tr A(i,j) ·Tr B(i,j) ;
[0165] Where: γ ij It represents the trust of vehicle i to vehicle j.
[0166] 2. System State Space
[0167] In this embodiment, the system state space equation of the vehicle queue is expressed as:
[0168]
[0169] Where: x i (k) is the system status; y i (k) is the system output; u i (k) is the system input; A, B, C, D, E are the system parameter matrices; f i (k) is the fault input; w(k) is the process noise; v(k) is the measurement noise.
[0170] 3. I / O Data Model
[0171] Collect the input and output signals of the system. The input data is the input U of the controller. The output data includes the vehicle's position information, speed information, and acceleration information. The acceleration information is taken as the output data of the I / O system. The input and output data include past (p, past) information and future (f, future) information. f Taking as an example, the specific form of the input and output data is as follows: (where N is not the number of vehicles, but the number of samples, and s is the order of the equivalent vector).
[0172] In this embodiment, the vehicle input is used as the input of the I / O data model, and the acceleration information of the vehicle is used as the output of the I / O data model; the reconstructed I / O data model is expressed as:
[0173] Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ;
[0174]
[0175]
[0176]
[0177] X(k)=[x(k)x(k+1)…x(k+N-1)];
[0178]
[0179] 4. Residual Generator
[0180] Parity Space Approach is a method for Fault Detection and Diagnosis (FDI). Based on the system model, a residual generator is designed. In the absence of a fault, the residual is close to zero. By defining the parity equation, the output of the residual generator is mapped into a lower dimensional space, which is called the parity space. The residual is evaluated to determine whether the system has a fault. If the residual exceeds a certain pre-set threshold, the system may have a fault. Once a fault is detected, the parity space approach can also help determine the location of the fault and estimate the size and characteristics of the fault, such as the magnitude and occurrence time of the fault. The purpose of the parity equation is to provide a framework for detecting and isolating faults in the system through residuals.
[0181] In this embodiment, combined with Figure 4 As shown, the residual generator is constructed through the following steps:
[0182] S301: Extract Y from the I / O data model f , U f , Y p , U p ;
[0183] Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ;
[0184]
[0185]
[0186]
[0187] S302: Obtaining Data Build And perform SVD decomposition;
[0188] The decomposition formula is expressed as:
[0189]
[0190] Where: U z is an orthogonal matrix, the columns are left singular vectors, and the matrix is a transformation of the identity matrix; is a diagonal matrix, where ∑ z,1It is a diagonal matrix of non-zero singular values, where the non-zero singular values are arranged on the diagonal and the other positions are filled with 0; is the transposed matrix of an orthogonal matrix, whose columns are right singular vectors; is the matrix U z The block representation of U z,11 , U z,12 , U z,21 and U z,22 is its submatrix;
[0191] This formula represents the matrix Perform singular value decomposition, the decomposition form is:
[0192]
[0193] Where: U z and V z is an orthogonal matrix, is a diagonal matrix, where ∑ z,1 Contains nonzero singular values.
[0194] S303: Based on The SVD decomposition results of constructing an equivalent space and the correlation matrix
[0195] In this embodiment, from the matrix Select element U from z,12 Constructing equivalence space
[0196] S304: Calculation Get the maximum eigenvalue λ s,max and the eigenvector l s,max ;
[0197] S305: Constructing robust equivalent vectors and
[0198] Based on the system model, a residual generator is designed. The residual is the difference between the actual output of the system and the output predicted by the model. In the absence of faults, the residual should be close to zero. The noise-free state space equation of the system is as follows:
[0199]
[0200] Its equivalent space model is as follows:
[0201] y s (k) = Γ s x(ks)+H s,u u s (k)+H s,w fs (k);
[0202] from Pick any vector from Multiplying both sides of the above formula on the left, we get the following residual vector:
[0203]
[0204] The H2 performance indicators in the equivalent space are as follows:
[0205]
[0206] Calculating the optimal solution of the above performance index function is equivalent to solving the following generalized matrix eigenvalue and eigenvector problem:
[0207]
[0208] S306: Construct a residual generator: r(k)=v s y s (k)-β s u s (k).
[0209] 5. Fault estimation
[0210] The fault estimation part adopts a fault estimation method that combines the odd-even space method and iterative learning. By reconstructing the I / O data model of the state space equation, collecting historical data, including input and output data, and then obtaining the data-driven implementation of the system kernel through QR decomposition and SVD decomposition.
[0211] In this embodiment, combined with Figure 5 As shown, fault estimation is achieved through the following steps:
[0212] S401: Obtain residual generator r(k)=v s y s (k)-β s u s (k);
[0213] S402: Considering that there is a large interference when each vehicle transmits state information, an L2 regularized optimization objective function is introduced. When , we hope to minimize the error between the residual r(k) and the estimated fault, introduce the L2 regularization term, and construct the objective function based on the residual generator;
[0214] The objective function is expressed as:
[0215]
[0216] Where: r represents the residual of the system; Q represents the design matrix; represents the estimated value of the fault signal; ||·|| 2 represents the square of the two norm (L2 norm); λ is the regularization parameter, which controls the weight of the regularization term;
[0217] The objective function consists of two parts:
[0218] Residual Error Term represents the error between the residual and the fault estimate.
[0219] Regularization term Used to deal with interference and noise and prevent overfitting.
[0220] S403: splitting the objective function into two parts: a residual error term and a regularization term;
[0221] The residual error term is expressed as:
[0222]
[0223] The regularization term is expressed as:
[0224]
[0225] S404: solving the gradients of the residual error term and the regularization term respectively;
[0226] 1) Residual error term
[0227] Derivative of J1(f):
[0228] Expand and simplify:
[0229]
[0230] Solve for J 1 The gradient of (f):
[0231]
[0232] 2) Regularization term
[0233] Derivative of J2(f):
[0234] Solve for J 2 The gradient of (f):
[0235] S405: merging and simplifying the gradient of the residual error term and the regularization term;
[0236] The formula is:
[0237]
[0238] To make the gradient zero, solve for the fault estimate f:
[0239]
[0240] Simplified equation:
[0241]
[0242]
[0243]
[0244] S406: Obtaining a closed-form solution for the fault estimation value;
[0245] The formula is:
[0246]
[0247] Where: Q represents the design matrix; r is the residual of the system; λ is the regularization parameter, which is used to control the weight of the regularization term; I is the unit matrix, which is used to ensure that Q T Q+λ, reversible.
[0248] In the iterative process, λ(k+1)=γλ(k) is adaptively adjusted. γ is the adjustment coefficient, and the size of λ is dynamically adjusted according to the estimation error. For problems with complex closed-form solutions, the gradient descent method can be used to iteratively solve the fault estimation value.
[0249] The formula is:
[0250]
[0251] Where: η represents the learning rate; λ represents the regularization parameter; r(k) represents the residual.
[0252] In the present invention, the odd-even space method is combined with an iterative learning algorithm to estimate faults online, which enables the algorithm to adapt to changes in the system operation process. A fault estimation error performance function based on a residual vector is introduced to more accurately evaluate the performance of system fault estimation. Through the iterative learning algorithm, the system can optimally estimate the fault signal online, continuously adjust and improve the fault estimation process, thereby improving the accuracy and robustness of fault estimation. This method can not only detect and diagnose system faults more effectively, but also maintain high efficiency and reliability in the face of complex and dynamic working environments.
[0253] 6. Controller Design
[0254] A compensation-based cooperative fault-tolerant controller is a control strategy used in control systems to handle system failures. When an actuator or sensor in the system fails, the controller is able to maintain system stability and maintain system performance as much as possible by compensating for the effects of the failure. After the fault is detected, it is necessary to estimate the size and characteristics of the fault, such as the magnitude and direction of the fault. Based on the type of fault and the estimated results, a fault-tolerant control strategy is designed. This usually involves redesigning or adjusting controller parameters to compensate for the effects of the fault. Based on the fault-tolerant control strategy, the feedback information of the system is used to adjust the control input to maintain the stability and performance of the system. Under the premise of meeting stability and safety, the system performance is optimized, such as by minimizing the control input or increasing the response speed of the system.
[0255] By introducing a zero-trust architecture and distributed control strategy, the impact of abnormal nodes and fault signals on vehicle platoons is minimized. At the same time, by reducing the permissions of vehicle nodes to the minimum range, it is ensured that the reference signals received by each vehicle node are normal and valid. The driving behaviors and driving strategies between each vehicle node are consistent to ensure the safety of vehicle platoon driving.
[0256] In this embodiment, the controller of vehicle i is represented by the following formula:
[0257]
[0258] Where: x i1 represents the first state quantity of vehicle i, i.e., position, x i2 represents the second state quantity of vehicle i, namely speed, x i3 The third state quantity x represents the acceleration of vehicle i; j1 represents the first state of car j, x j2 represents the second state of car j, x j3 represents the third state quantity of vehicle j (the value of j ranges from 0 to N, and i≠j, j is the total number of vehicles); K p , K v , K a is the controller gain; γ ij B is the trust of vehicle i in vehicle j; * is the parameter matrix to be designed; d represents the distance between vehicles in the vehicle queue; Indicates the fault estimation result.
[0259] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.
Claims
1. A vehicle collaborative fault-tolerant control method based on data consistency under a zero-trust architecture, characterized in that: include: S1: Under the zero-trust architecture, the trust of each vehicle in the vehicle queue towards other vehicles is evaluated based on the data consistency evaluation method; S2: construct system state space equations for the vehicle platoon and reconstruct the system state space equations into an I / O data model; S3: Use the parity space method to build the corresponding residual generator in combination with the I / O data model; S4: Calculate the fault estimate of the actuator by combining the output of the residual generator through an iterative learning method; S5: For each vehicle in the vehicle queue, a compensation-based cooperative fault-tolerant controller is constructed according to the trust evaluation results of other vehicles and the fault estimation value of the actuator; S6: For each vehicle in the vehicle queue, a corresponding control input is generated based on its cooperative fault-tolerant controller to cooperatively adjust the operating state of the vehicle.
2. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture according to claim 1, characterized in that: In step S1, the trust of vehicle i in vehicle j is evaluated through the following steps: S101: Calculate the behavior information consistency trust Tr based on the consistency of lane change frequency, acceleration stability and deceleration stability between vehicle i and vehicle j A(i,j) ; S102: Set the trust threshold γ1, and calculate the trust Tr of the consistency of the behavior information between vehicle i and vehicle j by using the following formula A(i,j) Preprocessing is performed to convert the behavior information consistency trust Tr A(i,j) Convert to 0 or 1; S103: Obtaining the state data sequence X of vehicle i itself i ={x i,1 ,x i,2 ,…,x i,n } and the state data sequence X of vehicle j received by vehicle i (i,j) ={x (i,j),1 ,x (i,j),2 ,…,x (i,j),n }; S104: Calculate state data sequence X i and X (i,j) Corresponding element difference Δ (0,1),i =|x 0,i -x (0,1),i |, forming a difference sequence Z (0,1) ={Δ (0,1),1 ,Δ (0,1),2 ,…,Δ (0,1),n }; S105: Calculate the node status information consistency trust Tr between vehicle i and vehicle j based on the difference sequence B(i,j) ; S106: Set the trust threshold γ2, and calculate the trust Tr of the node status information consistency between vehicle i and vehicle j by combining the following formula B(i,j) Perform preprocessing to make the node status information consistent with the trust degree Tr B(i,j) Convert to 0 or 1; S107: The consistency trust Tr of the behavior information between the pre-processed vehicles i and j A(i,j) The trust degree Tr of the node status information is consistent with B(i,j) Evaluate vehicle i’s trust in vehicle j; The formula is: γ ij =Tr A(i,j) ·Tr B(i,j) ; Where: γ ij represents the trust of vehicle i in vehicle j.
3. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 2, characterized in that: In step S101, the consistency trust degree Tr between the behavior information of vehicle i and vehicle j is calculated by the following formula: A(i,j) : Tr A(i,j) =w1×F C(i,j) +w2×A S(i,j) +w3×D S(i,j) ; F C(i,j) =|F Ci -F Cj |; A S(i,j) =|A Si -A Sj |; D S(i,j) =|D Si -D Sj |; Where: w1, w2, w3 are weight coefficients; F C(i,j) , A S(i,j) , D S(i,j) They represent the consistency of lane change frequency, acceleration stability and deceleration stability between vehicles i and j respectively; F Ci , A Si , D Si They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle i respectively; F Cj , A Sj , D Sj They represent the lane changing frequency, acceleration stability and deceleration stability of vehicle j respectively; a i represents the acceleration value measured at the i-th time; d i represents the deceleration value of the i-th measurement; n represents the number of measurements.
4. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 2, characterized in that: In step S105, the node status information consistency trust Tr between vehicle i and vehicle j is calculated by the following formula: B(i,j) : Where a1 is a parameter used to control the speed at which trust decreases; n is the number of element differences in the difference sequence.
5. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 1, characterized in that: In step S2, the system state space equation of the vehicle queue is expressed as: Where: x i (k) is the system status; y i (k) is the system output; u i (k) is the system input; A, B, C, D, E are the system parameter matrices; f i (k) is the fault input; w(k) is the process noise; v(k) is the measurement noise.
6. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 5, characterized in that: In step S2, the vehicle input is used as the input of the I / O data model, and the acceleration information of the vehicle is used as the output of the I / O data model; The reconstructed I / O data model is expressed as: Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ; X(k)=[x(k)x(k+1)…x(k+N-1)]; 7. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 1, characterized in that: In step S3, the residual generator is constructed by the following steps: S301: Extract Y from the I / O data model f , U f , Y p , U p ; Y f =Γ s X(k)+H s,u U f +H s,w W f +V f ; S302: Obtaining Data Build And perform SVD decomposition; S303: Based on The SVD decomposition results of constructing an equivalent space and the correlation matrix S304: Calculation Get the maximum eigenvalue λ ss,max and the eigenvector l s,max ; S305: Constructing robust equivalent vectors and S306: Construct a residual generator: r(k)=v s y s (k)-β s u s (k).
8. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 1, characterized in that: In step S4, fault estimation is achieved through the following steps: S401: Obtain residual generator r(k)=v s y s (k)-β s u s (k); S402: Introduce L2 regularization term and construct objective function based on residual generator; The objective function is expressed as: Where: r represents the residual of the system; Q represents the design matrix; Indicates the estimated value of the fault signal; ∥·∥ 2 represents the square of the two-norm; λ is the regularization parameter, which controls the weight of the regularization term; S403: splitting the objective function into two parts: a residual error term and a regularization term; The residual error term is expressed as: The regularization term is expressed as: S404: solving the gradients of the residual error term and the regularization term respectively; 1) Gradient of the residual error term J1(f): 2) Gradient of the regularization term J2(f): S405: merging and simplifying the gradient of the residual error term and the regularization term; The formula is: Let the gradient be zero and rewrite the above formula as: Simplified to: S406: Generate a closed-form solution for the fault estimation value; The formula is: Where: represents the fault estimation of the actuator; Q represents the design matrix; r represents the residual of the system; λ represents the regularization parameter; I is the identity matrix, which is used to ensure that Q T Q+λI is reversible.
9. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 8, characterized in that: In step S406, the fault estimation value is iteratively solved using the gradient descent method; The formula is: Where: η represents the learning rate; λ represents the regularization parameter; r(k) represents the residual.
10. The vehicle collaborative fault-tolerant control method based on data consistency under the zero-trust architecture as claimed in claim 1, characterized in that: In step S5, the controller of vehicle i is represented by the following formula: Where: x i1 represents the first state quantity of vehicle i, i.e., position; x i2 represents the second state quantity of vehicle i, namely speed; x i3 The third state quantity x represents the acceleration of vehicle i; j1 represents the first state of vehicle j; x j2 represents the second state of vehicle j; x j3 Represents the third state quantity of vehicle j; K p , K v , K a is the controller gain; d represents the distance between vehicles in the vehicle queue; γ ij B is the trust of vehicle i in vehicle j; * is the parameter matrix to be designed; Indicates the fault estimation result.
