Vehicle running state monitoring method under zero-trust network architecture

Abstract

The invention belongs to the technical field of zero-trust networks, and relates to a vehicle running state monitoring method under a zero-trust network architecture, which comprises the following steps of: establishing a trust level evaluation rule according to the characteristic that the trust level of a vehicle is not transferable under the ZTNA architecture; establishing a communication relationship between the information receiver and the sender according to the trust level evaluation rule and the non-reciprocity of the trust level; establishing a system model of a vehicle node i; establishing a system model of the motorcade ICVs group; designing a distributed unknown input observer capable of realizing an estimation target on the basis that each vehicle only completely trust detection output of an ICVs sensor of the vehicle, control input from a controller of the vehicle and corresponding information trust levels given to information received from adjacent vehicles by a vehicle management system; according to the method, a distributed unknown input observer is designed according to the fact that each vehicle node only utilizes known information of the vehicle node and state estimation information from adjacent vehicles, and the state estimation information of the adjacent vehicles comprises the trust level of the corresponding vehicle, so that an estimation target is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of zero-trust network technology, and specifically relates to a vehicle operation status monitoring method under a zero-trust network architecture. Background Art

[0002] The existing trust level assessment principle under ZTNA is that in the ICV (Intelligent Connected Vehicle) group under the ZTNA (Zero Trust Network Access) architecture, the communication between any vehicle and other vehicles is not trusted or accepted by default. The receiver continuously evaluates the trust level of the sending vehicle based on the ZTNA principle of "never trust, always verify". The acceptance or rejection of subsequent messages will be based on the real-time evaluation results.

[0003] Inspired by the process of building trust relationships between humans, the following analysis evaluates the trust level between vehicles within ZTNA:

[0004] 1) Vehicles with the lowest level of trust can still exchange useful information with each other. Similar to interactions between people, strangers may occasionally exchange information, but this is usually a one-time interaction and does not involve important information. Continuous, useful information exchange usually requires a minimum level of trust. Therefore, we assume that once authorized by management, the level of trust between the vehicle group will reach the minimum level, that is, greater than zero.

[0005] 2) Sending truthful information can increase the level of trust between the sender and the receiver. The trust level is determined by the receiver's evaluation of the sender's behavior, which is known only to the receiver. Between people, trust tends to increase when someone does the right thing or speaks truthfully. Therefore, we assume that the trust level of a vehicle is related to the truthful information it sends and the receiver receives.

[0006] 3) The level of trust is limited. Trust between people rarely reaches an infinite level, and this concept is consistent with the security principles of ZTNA. Therefore, we assume that any vehicle can have a maximum level of trust in its neighboring vehicles. Once this limit is reached, the trust level will no longer increase even if true information is received.

[0007] 4) Obviously, any trust level above the minimum has a time limit. In the absence of long-term contact, people will know less about each other, resulting in a decrease in trust. Therefore, we assume that trust levels are time-sensitive and will decrease over time without continued interaction.

[0008] 5) When a vehicle receives false information, it should quickly reduce the trust level of the sender. The rate of this reduction should be significantly higher than the rate at which the trust level decays naturally over time.

[0009] 6) When the vehicle recognizes that the sender is conducting an attack, it will immediately reduce its trust level to zero and stop receiving any messages from that sender. This is equivalent to a unilateral network disconnection.

[0010] 7) When a vehicle receives conflicting information from multiple sources, it tends to believe the view held by the majority and gives more trust to the information sent by vehicles with higher trust levels.

[0011] From the above analysis, it is obvious that the trust level between the two communicating parties is asymmetric, and only the information receiver has the evaluation result of the sender's trust level. In addition, the trust level is not transferable. In fact, under the ZTNA architecture, the trust evaluation results between vehicles without communication relationship are also meaningless. Summary of the invention

[0012] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a vehicle operation status monitoring method under a zero-trust network architecture.

[0013] In order to achieve the purpose of the present invention, the present invention will be implemented by adopting the technical solution described below.

[0014] A vehicle operation status monitoring method under a zero-trust network architecture comprises the following steps:

[0015] S1. Based on the non-transferable nature of vehicle trust levels under the ZTNA architecture, the initial trust level of all vehicles is set to the lowest trust level by the vehicle management system. To establish a trust level evaluation rule, the trust level evaluation rule includes the accumulation of trust level, natural decay of trust level, accelerated decay of trust level and trust level reduction to zero, wherein:

[0016] The accumulation of trust levels: The trust level of all vehicles is the minimum trust level set by the vehicle management system Every time the i-th vehicle receives a true message from the j-th vehicle, the trust level increases by The value represented by At the same time, the timer for maintaining the trust level is reset; when the trust level reaches the upper limit or the highest level, the trust level of the information recipient will no longer increase when receiving real messages But each authentic message received still resets the timer that maintains the trust level;

[0017] The natural decay of the trust level: When the trust level T is maintainedr The time exceeds the maximum allowed time T r，max , and no new real information is received, the trust level It starts to decrease gradually over time according to linear or parabolic characteristics until it reaches the lowest trust level;

[0018] The accelerated decay rule of the trust level: when receiving false information, the trust level It starts to gradually decrease over time in a linear or parabolic manner, and its rate of decrease is significantly greater than that of natural attenuation;

[0019] The trust level drops to zero: when the information sent by the information sender is identified by the information receiver as an attack, the trust level of the information sender will immediately drop to zero, and the information receiver will stop receiving any information from the information sender until the vehicle management system reinitializes the trust level of the information sender to the lowest trust level;

[0020] S2. According to the trust level evaluation rules under the ZTNA architecture established in step S1 and the non-reciprocity of the trust level, a communication relationship between the information receiver and the information sender is established. The communication relationship is represented as a zero trust directed graph.

[0021] in: is a collection of nodes, is the set of edges; each edge Indicates that information is from node ν j Send to node ν i , node ν i is the information receiver; node v j The sender of the information;

[0022] The information receiver assigns the information trust level a to the received information according to the vehicle management system ij and predefined rules to decide whether to accept information from the sender; among which:

[0023] Information Confidence Level To reflect the credibility of the information received;

[0024] The predefined rules: Based on the zero trust adjacency matrix The definitions are as follows:

[0025]

[0026] In the formula, 0<a min <a max is the preset minimum information trust level, a min <a max<1 is the preset highest information trust level;

[0027] The zero trust adjacency matrix The derivation process is as follows:

[0028] Since the vehicle does not need to send information to itself, we assume that a ii =0; Zero Trust Laplace Matrix and

[0029]

[0030] In order to distinguish between inside and outside the convoy, it is assumed that the convoy satisfies the strong connectivity condition when it is reset by the vehicle management system; the strong connectivity condition means that there is at least one directed path between any two vehicles in the convoy; in addition, since the trust level of the vehicle is continuously evaluated under the ZTNA architecture, a ij Become a variable a ij (t), therefore, the zero trust adjacency matrix and the Zero Trust Laplace Matrix Should be recorded as and Introducing Lemma 1 gives the zero-trust adjacency matrix and the Zero Trust Laplace Matrix relationship;

[0031] Lemma 1. For a strongly connected directed graph with N nodes and its Laplace matrix The following conclusions hold:

[0032] 1) contains a zero eigenvalue, whose corresponding eigenvector is 1 N (Right now ), all other eigenvalues ​​are located in the right semi-closed region of the complex plane;

[0033] 2) There is a vector , so that in The elements satisfy , and satisfy the following conditions:

[0034] 3) Set ,but is a symmetric Laplacian matrix corresponding to an undirected graph;

[0035] S3. Establish vehicle node i System Model

[0036] S31. Establishing a longitudinal dynamics model of the vehicle by analyzing the longitudinal dynamics of the vehicle, wherein the longitudinal dynamics model is expressed as:

[0037]

[0038] Among them, s i (t), v i (t) and a i (t) represent the position, velocity and acceleration of the i-th vehicle, τ i represents the time constant of the engine, u i (t) is the control input;

[0039] S32. Define the system state of vehicle node i as Then the system equation of the i-th vehicle is expressed as:

[0040]

[0041] In the formula, Indicates the system status, n = 3, is the control input, m = 1, and

[0042]

[0043] S33. Since each ICV has a positioning sensor, a speed sensor and an acceleration sensor,

[0044] Each vehicle can output its own position, velocity and acceleration state, so the output matrix of the i-th vehicle is expressed as:

[0045]

[0046] in, represents the sensor output of the i-th vehicle, p = n = 3, and is a known constant matrix;

[0047] S4. Establish a system model for the fleet ICVs group

[0048] S41. The system state of the fleet is expressed as The system input for the corresponding fleet is The system equation of the fleet is expressed as:

[0049]

[0050] In the formula,

[0051] S42. Considering the unknown input disturbance experienced by the fleet In the case of , the system state space equation of the fleet ICVs group is derived as:

[0052]

[0053] In the formula,

[0054] Assume that for i = 1, 2, ..., N, rank(CiD) = q; in order to make the description more concise and clear, (t) will be omitted if there is no ambiguity;

[0055] S5. Based on the fact that each vehicle fully trusts only the detection output of its own ICVs sensor and the control input from its own controller, and the vehicle management system assigns corresponding information trust levels to the information received from neighboring vehicles, a distributed unknown input observer that can achieve the estimation target is designed. The expression of the distributed unknown input observer is:

[0056]

[0057] In the formula, represents the state of the distributed unknown input observer, Represents the estimation of the global state by the i-th vehicle; the matrix as well as They are all determined by solving the distributed unknown input observer coefficient matrix algorithm, and ensuring the stability of the entire residual system so that the distributed unknown input observer can achieve the estimation target; where:

[0058] The algorithm for solving the coefficient matrix of the distributed unknown input observer is as follows: initialize the observers of all vehicles (i=1, 2, ..., N), and perform the following steps for each vehicle i=1, 2, ..., N:

[0059] 1: Set H i ;

[0060] 2: Use formula (9) to solve the matrix P that satisfies the conditions i , and substitute into formula (8) to calculate E i ;

[0061] 3: Calculate formula (11) and find the matrix X that satisfies the conditions i and Y i ;

[0062] 4: Find K io , so that (A io -K io C io ) is the Hurwitz matrix;

[0063] 5: X i and K io Substitute into formula (11) to calculate K i ;

[0064] 6: Use Lemma 1 to find a suitable and To meet the requirements, we can get the appropriate γ i ;

[0065] 7: Y i and γ i Substituting into formula (13) we get M i ;

[0066] 8: Use formulas (15) and (19) to calculate N respectively i and L i ;

[0067] 9: Select H i and the calculated L i 、M i 、E i Substitute into formula (6) and use a ij (t) to obtain the estimation result of the global system state by the first vehicle;

[0068] Where: The formula (8) is: P i =(E i C i +I Nn ),

[0069] Where: Pi is the intermediate matrix used to solve the distributed observer matrix of the i-th vehicle; Ei represents the observer parameter matrix of the i-th vehicle, which is used to process the feedback and correction of the sensor output yi; Ci represents the output matrix of the i-th vehicle, which maps the vehicle state x(t) to the sensor output yi; INn is the unit matrix of (N×n)×(N×n) dimensions;

[0070] The formula (9) is: i B=H i [I m ] i|N , P i D=0,

[0071] Where B is the input matrix associated with the control input u(t), which is used to describe how the input affects the system state; Hi is the parameter matrix in the distributed observer; Where Im is the m×m dimensional identity matrix; D is the disturbance input matrix associated with the unknown input d(t), which is used to describe the impact of the unknown disturbance on the system state;

[0072] The formula (11) is:

[0073]

[0074] Where Θi is the observability matrix of the analysis system related to the i-th vehicle; ζi represents the dimension of the unobservable subspace related to the i-th vehicle; A is the state transition matrix of the system;

[0075] The formula (13) is:

[0076] Where Mi is the parameter matrix related to the i-th vehicle in the distributed observer; γi is the scalar gain coefficient used to adjust the weight of the matrix Mi; Yi is used to describe the unobservable state part of the system, and the column vector constitutes the orthogonal basis of ker(Θi);

[0077] The formula (15) is: i =P i AK i C i ,

[0078] Where Ni is the state matrix of the distributed observer; Ki is the observer gain matrix of the i-th vehicle;

[0079] The formula (19) is: i =X i K io (I+C i E i )-(E i C i +I)AE i ,

[0080] Where Li is the distributed observer parameter matrix associated with the i-th vehicle; Xi is used to extract the observable part in the state block; Kio represents the gain matrix associated with the i-th vehicle and the observable part; I represents the unit matrix; Ei represents the parameter matrix associated with the output feedback of the i-th vehicle in the distributed observer.

[0081] As a preferred solution of the present invention, the evaluation principle of the vehicle trust level under the ZTNA architecture is:

[0082] 1) Vehicles with the lowest trust level can still exchange useful information with each other;

[0083] 2) Sending real information can increase the trust between the sender and the receiver;

[0084] 3) The degree of trust is limited;

[0085] 4) Any trust level above the minimum has a time limit;

[0086] 5) When a vehicle receives false information, it should quickly reduce the trust level of the sender of the information;

[0087] 6) When the vehicle identifies that the sender is conducting an attack, it will immediately reduce its trust level to zero and stop receiving any messages from the sender;

[0088] 7) When a vehicle receives conflicting information from multiple sources, it tends to believe the view held by the majority and give more trust to the information sent by vehicles with higher trust levels.

[0089] As a preferred solution of the present invention, the specific process of solving the distributed unknown input observer coefficient matrix is ​​as follows:

[0090] The state estimation error of the i-th vehicle is defined as From this we can get:

[0091]

[0092] make:

[0093] P i =(E i C i +I Nm ), (8)

[0094] Where: Pi is the intermediate matrix used to solve the distributed observer matrix of the i-th vehicle; Ei represents the observer parameter matrix of the i-th vehicle, which is used to process the feedback and correction of the sensor output yi; Ci represents the output matrix of the i-th vehicle, which maps the vehicle state x(t) to the sensor output yi; INn is the unit matrix of (N×n)×(N×n) dimensions;

[0095] If H is satisfied i u i =P i Bu and P i D=0, which is equivalent to:

[0096] P i B=H i [I m ] i|N , P i D=0, (9)

[0097] Where B is the input matrix associated with the control input u(t), which is used to describe how the input affects the system state; Hi is the parameter matrix in the distributed observer; in, I m is an m×m-dimensional identity matrix; D is the disturbance input matrix associated with the unknown input d(t), which is used to describe the impact of unknown disturbances on the system state;

[0098] Then we have:

[0099]

[0100] By separating the observable and unobservable parts of the system (P i A, C i ), for this incompletely observable system, there exists a non-negative integer Make

[0101]

[0102] Where Θi is the observability matrix of the analysis system related to the i-th vehicle; ζi represents the dimension of the unobservable subspace related to the i-th vehicle; A is the state transition matrix of the system;

[0103] So, It is the system (P i A, C i )of dimensional unobservable subspace;

[0104] Select a matrix Its columns can form ker(Θ i ) is an orthogonal basis, that is, im(Y i )=ker(Θ i );and It is to ensure is a matrix of orthogonal matrices;

[0105]

[0106] in,

[0107] as well as also, C i Y i =0; obviously (A io , C io ) is observable;

[0108] In addition, it is worth noting that as well as Always holds true; because (A io , C io ) is observable, so there must be a matrix K io , so that (A io -K io C io ) is a Hurwitz matrix; such a K can be found as follows io ;

[0109] If we order:

[0110] K i =X i K io , (12)

[0111] Then (P i AK i C i ) will only be affected by the observable part; and let:

[0112]

[0113] Where Mi is the parameter matrix related to the i-th vehicle in the distributed observer; γi is the scalar gain coefficient used to adjust the weight of the matrix Mi; Yi is used to describe the unobservable state part of the system, and the column vector constitutes the orthogonal basis of ker(Θi);

[0114] We have:

[0115]

[0116] If we set:

[0117] N i =P i AK i C i , (15)

[0118] Where Ni is the state matrix of the distributed observer; Ki is the observer gain matrix of the i-th vehicle;

[0119] Then we can get:

[0120]

[0121] Our goals are to achieve:

[0122]

[0123] Right now:

[0124]

[0125] By comparing equation (16) and equation (18), we can see that L i Must be:

[0126] L i =X i K io (I+C i E i )-(E i C i +I)AE i . (19)

[0127] Where Li is the distributed observer parameter matrix related to the i-th vehicle; Xi is used to extract the observable part in the state block; Kio represents the gain matrix related to the i-th vehicle and the observable part; I represents the unit matrix; Ei represents the parameter matrix related to the output feedback of the i-th vehicle in the distributed observer;

[0128] For the unobservable part, we can also deduce

[0129]

[0130] and

[0131]

[0132] Let e ​​= col(ei) 1：N , X = diagX i ) 1：N , Y=diag(Y i ) 1：N , A o =diag(A io ) 1：N , A r =diag(A ir ) 1：N , A u =diag(A iu ) 1：N , C o =diag(C io ) 1：N , K o =diag(K io ) 1：N as well as We can get:

[0133]

[0134] and

[0135]

[0136] Notice We can deduce from equations (22) and (23):

[0137]

[0138] Furthermore, it can be expressed as:

[0139]

[0140] Among them, Θ o =A o -Ko C o ,

[0141] According to Lemma 1, Lemma 2 is given;

[0142] Lemma 2: Assume that the communication network is strongly connected, and (A, col(C i ) 1：N ) is observable; then the matrix It is positive;

[0143] Because (A o -K o C o ) is a Hurwitz matrix, so part is stable; as can be seen from equation (23), we only need to ensure

[0144]

[0145] is stable to ensure that the entire residual system (25) is stable; the following Lyapunov function is selected:

[0146]

[0147] in, In Lemma 1, γ min =min(γ i ) 1：N . It can be deduced that:

[0148]

[0149] According to Lemma 1, we know ;therefore, This means:

[0150]

[0151] in, Because the matrix is positive definite, and γ is defined in Lemma 1;

[0152] Therefore, if but Therefore, equation (26) is stable, that is, the entire residual system is stable, indicating that the distributed observer proposed in equation (6) can achieve the estimation goal.

[0153] Beneficial Effects

[0154] Considering that the states of all vehicles are affected by unknown inputs, this method solves the state estimation problem of each vehicle in the group of intelligent connected vehicles under the ZTNA architecture;

[0155] Inspired by the process of establishing and dissolving human trust relationships, this paper designs the impact verification rules of recursively authenticated vehicle trust level changes in ZTNA;

[0156] The present invention uses a variable directed graph to describe the communication topology of the intelligent connected vehicle in the system structure;

[0157] The invention designs a distributed unknown input observer based on each vehicle node using only known information of itself and state estimation information from neighboring vehicles, and the state estimation information of neighboring vehicles includes the trust level of the corresponding vehicle to achieve the estimation target. BRIEF DESCRIPTION OF THE DRAWINGS

[0158] Figure 1 The trust level evaluation rules of the present invention under ZTNA, wherein: Figure a shows the trust level accumulation; Figure b shows the natural decay of the trust level; Figure c shows the accelerated decay of the trust level; Figure d shows the trust level dropping to zero;

[0159] Figure 2 Trust level of Case 1 of the present invention under ZTNA;

[0160] Figure 3 It is the trust level of Case 2 of the present invention under ZTNA;

[0161] Figure 4 It is the trust level of Case 3 of the present invention under ZTNA;

[0162] Figure 5 All vehicle states and their estimates under ZTNA in Case 1 of the present invention;

[0163] Figure 6 All vehicle states and their estimates under ZTNA in Case 2 of the present invention;

[0164] Figure 7 All vehicle states and their estimates under ZTNA in Case 3 of the present invention;

[0165] Figure 8 is the estimated result of the single state of the first vehicle in the embodiment of the present invention (γ i =50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0166] Fig. 9 is the estimated result of the single state of the second vehicle in the embodiment of the present invention (γ i=50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0167] Fig.10 is the estimated result of the independent state of the third vehicle in the embodiment of the present invention (γ i =50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0168] Fig.11 is the estimated result of the individual state of the fourth vehicle in the embodiment of the present invention (γ i =50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0169] Fig.12 is the estimated result of the individual state of the fifth vehicle in the embodiment of the present invention (γ i =50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0170] Fig.13 is the estimated result of the individual state of the sixth vehicle in the embodiment of the present invention (γ i =50), where: Figure a is a comparison diagram of the position of the vehicle and the estimated position; Figure b is a comparison diagram of the speed of the vehicle and the estimated speed; Figure c is a comparison diagram of the acceleration of the vehicle and the estimated acceleration;

[0171] Fig.14 is the third state of the sixth vehicle in the embodiment of the present invention and its corresponding estimation result (γ i =5);

[0172] Fig.15 is the third state of the sixth vehicle in the embodiment of the present invention and its corresponding estimation result (γ i =20);

[0173] Fig.16 is the third state of the sixth vehicle in the embodiment of the present invention and its corresponding estimation result (γ i =100);

[0174] Fig.17 The communication topology diagram of the experimental vehicle of the present invention;

[0175] Fig.18It is a comparison diagram of the estimation results of the present invention under NATA, wherein: Figure a is a comparison diagram of the position of the fleet and the estimated position; Figure b is a comparison diagram of the speed of the fleet and the estimated speed; Figure c is a comparison diagram of the acceleration of the fleet and the estimated acceleration. DETAILED DESCRIPTION

[0176] The present invention is further described in conjunction with the accompanying drawings and embodiments.

[0177] As an embodiment of the present invention, Figures 1 to 18 As shown, a vehicle operation status monitoring method under a zero-trust network architecture includes the following steps:

[0178] 1. Trust Level Assessment Rules under ZTNA

[0179] Since the propagation of trust levels is eliminated under the ZTNA architecture, all trust levels start from the lowest trust level set by the vehicle management system. The trust level will increase based on the received true information and decrease due to the received false information.

[0180] 1) Accumulation of trust level: The process of increasing trust level starts from the initial trust level set by the manager. Initially, this level represents the minimum trust level required to maintain communication. Whenever the i-th vehicle receives a genuine message from the j-th vehicle, the trust level increases by The value represented by At the same time, the timer for maintaining the trust level will be reset. When the trust level reaches the upper limit or the maximum level, further received real messages will no longer increase the trust level However, each genuine message received will still reset the timer to maintain the trust level. May not be a fixed constant, but rather a function that reflects factors such as the importance of the message and the identity of the sender. Figure 1 As shown in Figure a.

[0181] 2) Natural decay of trust level: When the trust level T is maintained r The time exceeds the maximum allowed time T r，max , and no new real information is received, the trust level It starts to gradually decrease over time according to a certain pattern (linear or other curve) until it reaches the lowest trust level. If the trust level continues to drop to the lowest level and no real information is received, it will remain at the lowest trust level and will not change. See Figure 1 As shown in Figure b.

[0182] 3) Accelerated decay of trust level: When receiving false information, the trust level will decrease faster, such as Figure 1As shown in Figure c.

[0183] 4) Trust level drops to zero: When the sender of information is identified as carrying out an attack, its trust level will immediately drop to zero, and the receiver will stop receiving any information from the sender until the administrator reinitializes the trust level to the lowest trust level, such as Figure 1 As shown in Figure d.

[0184] 2. Establishing the Communication Topology under ZTNA

[0185] Due to the non-reciprocity of trust levels and the trust level evaluation rules under the ZTNA architecture, the trust evaluation results between two communicating vehicles may be different. This communication relationship can be represented as a directed graph in ZTNA. Zero Trust Directed Graph in is a collection of nodes, is a set of edges. Each edge Indicates that information is from node ν j Send to node ν i . Node v i According to the assignment to node v j The trust level is used to evaluate whether to receive j The received information is also assigned a trust level. To reflect the credibility of the received information. Zero Trust Adjacency Matrix The definitions are as follows:

[0186]

[0187] In the formula, 0<a min <a max is the preset minimum trust level, a min <a max <1 is the preset highest trust level. Since the vehicle does not need to send information to itself, it is assumed that a ii = 0. Zero Trust Laplace Matrix and

[0188]

[0189] In order to distinguish between inside and outside the fleet, it is assumed that the fleet satisfies the strong connectivity condition when it is reset by the management party. This condition means that there is at least one directed path between any two vehicles in the fleet. In addition, since the trust level of the vehicle is continuously evaluated under the ZTNA architecture, a ij Become a variable a ij (t), therefore, the zero trust adjacency matrix and the Zero Trust Laplace Matrix Should be recorded as and

[0190] Lemma 1. For a strongly connected directed graph with N nodes and its Laplace matrix The following conclusions hold:

[0191] 1) contains a zero eigenvalue, whose corresponding eigenvector is 1 N (Right now ), all other eigenvalues ​​are located in the right semi-closed region of the complex plane.

[0192] 2) There is a vector , so that in The elements satisfy , and satisfy the following conditions:

[0193] 3) Set ,but is a symmetric Laplacian matrix corresponding to an undirected graph.

[0194] 3. Establishing System Modeling of Vehicle Nodes

[0195] After analyzing the longitudinal dynamics of the vehicle, the following longitudinal dynamics model can be established:

[0196]

[0197] Among them, s i (t),υ i (t) and a i (t) represent the position, velocity and acceleration of the i-th vehicle respectively. τ i represents the time constant of the engine, u i (t) is the control input.

[0198] Define the system status as Then the system equation of the i-th vehicle can be expressed as:

[0199]

[0200] in, represents the system status (here n=3), is the control input (here m=1), and

[0201]

[0202] ICVs are equipped with positioning sensors, velocity sensors, and acceleration sensors, allowing any vehicle to easily output its own position, velocity, and acceleration state without considering possible interference from these sensors. Therefore, the output matrix of the i-th vehicle can be expressed as:

[0203]

[0204] in, Indicates i The sensor output of the vehicle (here p=n=3), and is a known constant matrix. Obviously, is observable.

[0205] 4. Establishing the system model of the fleet ICVs group

[0206] The system state of the entire fleet can be expressed as The system input for the corresponding fleet is Then the system of the entire fleet can be expressed as

[0207]

[0208] in,

[0209] Considering the unknown input disturbances experienced by the fleet We can derive the state space equation of the entire system as follows:

[0210]

[0211] in,

[0212] Assumption 1. For i=1, 2, ..., N, rank(CiD)=q.

[0213] To make the description more concise and clear, (t) will be omitted when there is no ambiguity. For example, x is equivalent to x(t).

[0214] 5. Distributed Global State Estimation

[0215] In order to present the method more concisely, only the final results are provided in this section, and the derivation of more complex formulas is summarized in the Appendix.

[0216] Based on the premise that each vehicle fully trusts only the detection outputs of its own sensors and the control inputs from its own controller, and assigns corresponding trust levels to the information received from neighboring vehicles, we propose a distributed observer of the following form:

[0217]

[0218] in, represents the state of the distributed observer, Represents the i-th vehicle's estimate of the global state. The matrix as well as is the observer parameter matrix that needs to be determined.

[0219] For the i-th vehicle, the state estimation error is defined as From this we can get:

[0220]

[0221] Order: P i =(E i C i +I Nn ), (8)

[0222] Where: Pi is the intermediate matrix used to solve the distributed observer matrix of the i-th vehicle; Ei represents the observer parameter matrix of the i-th vehicle, which is used to process the feedback and correction of the sensor output yi; Ci represents the output matrix of the i-th vehicle, which maps the vehicle state x(t) to the sensor output yi; INn is the unit matrix of (N×n)×(N×n) dimensions;

[0223] If H is satisfied i u i =P i Bu and P i D=0, which is equivalent to

[0224] P i B=H i [I m ] i|N , P i D=0, (9)

[0225] Where B is the input matrix associated with the control input u(t), which is used to describe how the input affects the system state; Hi is the parameter matrix in the distributed observer; Among them, I m is an m×m-dimensional identity matrix; D is the disturbance input matrix associated with the unknown input d(t), which is used to describe the impact of unknown disturbances on the system state;

[0226] Then we have:

[0227] By separating the observable and unobservable parts of the system (P i A, C i ), for this incompletely observable system, there exists a non-negative integer Make

[0228]

[0229] Where Θi is the observability matrix of the analysis system related to the i-th vehicle; ζi represents the dimension of the unobservable subspace related to the i-th vehicle; A is the state transition matrix of the system;

[0230] So, It is the system (P i A, C i )of dimensional unobservable subspace.

[0231] Select a matrix Its columns can form ker(θ i ) is an orthogonal basis, that is, im(Y i )=ker(θ i ).and It is to ensure is an orthogonal matrix.

[0232]

[0233] in, as well as also, C i Y i =0. Obviously, (A io , C io ) is observable.

[0234] In addition, it is worth noting that as well as Always holds true. io , C io ) is observable, so there must be a matrix K io , so that (A io -K io C io ) is a Hurwitz matrix. Such a K can be found in several ways io .

[0235] If we set: K i =X i K io , (12)

[0236] Then (P i AK i C i ) will only be affected by the observable part. And let:

[0237] Where Mi is the parameter matrix related to the i-th vehicle in the distributed observer; γi is the scalar gain coefficient used to adjust the weight of the matrix Mi; Yi is used to describe the unobservable state part of the system, and the column vector constitutes the orthogonal basis of ker(Θi);

[0238] We have:

[0239] If we set: N i =P i AK i C i , (15)

[0240] Where Ni is the state matrix of the distributed observer; Ki is the observer gain matrix of the i-th vehicle;

[0241] Then we can get:

[0242]

[0243] Our goals are to achieve:

[0244] Right now:

[0245]

[0246] By comparing equation (16) and equation (18), we can see that L i Must be:

[0247] L i =X i K io (I+C i E i )-(E i C i +I)AE i . (19)

[0248] Where Li is the distributed observer parameter matrix related to the i-th vehicle; Xi is used to extract the observable part in the state block; Kio represents the gain matrix related to the i-th vehicle and the observable part; I represents the unit matrix; Ei represents the parameter matrix related to the output feedback of the i-th vehicle in the distributed observer;

[0249] For the unobservable part, we can also deduce:

[0250] and

[0251]

[0252] Let e ​​= col(e i )1：N , X=diag(X i ) 1：N , Y=diag(Y i ) 1：N , A o =diag(A io ) 1：N , A r =diag(A ir ) 1：N , A u =diag(A iu ) 1：N , C o =diag(C io ) 1：N , K o =diag(K io ) 1：N as well as We can get:

[0253] and

[0254]

[0255] Notice We can deduce from equations (22) and (23):

[0256]

[0257] Furthermore, it can be expressed as:

[0258] Among them, Θ o =A o -K o C o ,

[0259] According to Lemma 1, the following lemma is given.

[0260] Lemma 2. Assume that the communication network is strongly connected, and (A, col(C i ) 1：N ) is observable; then the matrix It is positive.

[0261] Because (A o -K o C o ) is a Hurwitz matrix, so As can be seen from equation (23), we only need to ensure that

[0262]

[0263] is stable to ensure that the entire residual system (25) is stable. Select the following Lyapunov function:

[0264]

[0265] in, In Lemma 1, γ min =min(γ i ) 1：N . It can be deduced that:

[0266]

[0267] According to Lemma 1, we know .therefore, This means:

[0268]

[0269] Table 1 Algorithm for solving the coefficient matrix of the distributed unknown input observer

[0270]

[0271]

[0272] in, Because the matrix is positive definite, and It is defined in Lemma 1.

[0273] Therefore, if but Therefore, Equation (26) is stable, which means that the entire residual system is stable. This shows that the distributed observer proposed in Equation (6) can achieve the estimation goal.

[0274] 6. Simulation Analysis

[0275] In order to verify the correctness and effectiveness of the proposed method, several potential communication scenarios in ZTNA are explored and simulated in this section. Subsequently, a simplified experiment is conducted using a fleet of three intelligent robot cars.

[0276] Consider an intelligent fleet of six cars. During the simulation, the following parameters are used: τ i = 0.5, and

[0277] D = col(D i ) 1：6 ,

[0278] and And u i =1+0.4sin(1.2t), d=0.5sin(t).

[0279] Taking the fourth vehicle as an example, the relevant parameter matrix is ​​as follows (elements not mentioned are all 0):

[0280] Among them, p 10，11 =0.3, p 10，12 =0.3, p 11，11 =0.25, p 11，12 =0.25, p 12，11 =0.2, p 12，12 =0.2. It can be verified that

[0281] in,

[0282] xy 10，1 =xy 11，2 =xy 12，3 =xy 12，4 =

[0283] xy 1，4 =xy 2，5 =xy 3，6 =xy 4，7 =xy 5，8 =xy 6，9 =xy 7，10 =xy 8，11 =xy 9，12 =1.

[0284]

[0285] Among them, n 10，10 =-0.5n 10，12 =-0.15, n 11，11 =-0.5, n 11，12 =-0.125, n 12，10 =0.15, n 12，11 =0.125, n 12，12 =-0.5.

[0286] Among them l 10，2 =0.18,l 10，3 =-0.12, l 11，2 =0.15, l 11，3 =-0.1, l 12，2 =0.0238, l 12，3=-0.1762.

[0287] VII. Impact of Changes in Different Trust Levels

[0288] First, the performance of the present invention under three different trust level change trends is studied, such as Figures 2 to 7 As shown in Figure 1, the trust level varies between 0.2 and 0.8. In cases 2 and 3, the trust level varies between 0.1 and 0.9, and the trust level in case 3 changes more frequently. All three cases involve directed graphs. It can be observed that different trust level changes have little effect on the performance of the proposed method. Overall, the method can provide satisfactory estimation results.

[0289] 8. All individual states and their estimates

[0290] In order to show the details of the proposed method more clearly, Figures 8 to 13 As shown in the figure, the estimated results of the individual states of the six smart cars are given respectively (γ i =50). The real state of the system is represented by the blue dashed lines. Obviously, even if it is affected by the unknown input d, the proposed method can still eliminate its influence. Although the initial values ​​of all vehicles are different, all vehicles can still accurately estimate the state of each vehicle in the end.

[0291] 9. Different i Impact

[0292] It can be observed that as γ i As increases, the weights of the estimated results of other vehicles become more prominent. This indicates that the state estimation results of all vehicles become more uniform. Figures 14 to 16 As shown, γ i The larger the value, the higher the overlap of the estimation results, thus improving the estimation accuracy. However, as the accuracy increases, the calculation time also increases. Figures 14 to 16 ,Under the same computing environment, the processing times are 0.28 seconds, 0.78 seconds, and 4.83 seconds, respectively.

[0293] 10. Simplify the experiment

[0294] The experimental verification fleet consisted of three intelligent experimental vehicles, numbered 1, 2, and 3. Due to the limitations of experimental conditions and in order to clearly observe the performance of the method, the experiment was only conducted at low speed and low communication frequency (10Hz). The experiment lasted for 1 minute, with the first 10 seconds designated as the acceleration phase, followed by the constant speed phase. Fig.17 and Fig.18As shown, the method obtains accurate estimation results. For comparison, on the right, when a 20 cm deviation is deliberately injected into the position signal of vehicle 1, the continuous zero-trust verification mechanism is activated. By reducing the weight of vehicle 1 in the estimation process, the overall estimation result is basically unaffected.

Claims

1. A vehicle operation status monitoring method under a zero-trust network architecture, characterized in that: The steps include: S1. Based on the non-transferable nature of vehicle trust levels under the ZTNA architecture, the initial trust level of all vehicles is set to the lowest trust level by the vehicle management system. To establish a trust level evaluation rule, the trust level evaluation rule includes the accumulation of trust level, natural decay of trust level, accelerated decay of trust level and trust level reduction to zero, wherein: The accumulation of trust levels: The trust level of all vehicles is the minimum trust level set by the vehicle management system Every time the i-th vehicle receives a true message from the j-th vehicle, the trust level increases by The value represented by At the same time, the timer for maintaining the trust level is reset; when the trust level reaches the upper limit or the highest level, the trust level of the information recipient will no longer increase when receiving real messages But each authentic message received still resets the timer that maintains the trust level; The natural decay of the trust level: When the trust level T is maintained r The time exceeds the maximum allowed time T r，max , and no new real information is received, the trust level It starts to decrease gradually over time according to linear or parabolic characteristics until it reaches the lowest trust level; The accelerated decay rule of the trust level: when receiving false information, the trust level It starts to gradually decrease over time in a linear or parabolic manner, and its rate of decrease is significantly greater than that of natural attenuation; The trust level drops to zero: when the information sent by the information sender is identified by the information receiver as an attack, the trust level of the information sender will immediately drop to zero, and the information receiver will stop receiving any information from the information sender until the vehicle management system reinitializes the trust level of the information sender to the lowest trust level; S2. According to the trust level evaluation rules under the ZTNA architecture established in step S1 and the non-reciprocity of the trust level, a communication relationship between the information receiver and the information sender is established. The communication relationship is represented as a zero trust directed graph. in: is a collection of nodes, is the set of edges; each edge ε ij =ε means the information is from node v j Send to node v i , node v i is the information receiver; node v j The sender of the information; The information receiver assigns the information trust level a to the received information according to the vehicle management system ij and predefined rules to decide whether to accept information from the sender; among which: Information Confidence Level To reflect the credibility of the information received; The predefined rules: Based on the zero trust adjacency matrix The definitions are as follows: In the formula, 0<a min <a max is the preset minimum information trust level, a min <a max <1 is the preset highest information trust level; The zero trust adjacency matrix The derivation process is as follows: Since the vehicle does not need to send information to itself, we assume that a ii =0; Zero Trust Laplace Matrix and In order to distinguish between inside and outside the convoy, it is assumed that the convoy satisfies the strong connectivity condition when it is reset by the vehicle management system; the strong connectivity condition means that there is at least one directed path between any two vehicles in the convoy; in addition, since the trust level of the vehicle is continuously evaluated under the ZTNA architecture, a ij Become a variable a ij (t), therefore, the zero trust adjacency matrix and the Zero Trust Laplace Matrix Should be recorded as and Introducing Lemma 1 gives the zero-trust adjacency matrix and the Zero Trust Laplace Matrix relationship; Lemma 1. For a strongly connected directed graph with N nodes and its Laplace matrix The following conclusions hold: 1) contains a zero eigenvalue, whose corresponding eigenvector is 1 N (Right now ), all other eigenvalues ​​are located in the right semi-closed region of the complex plane; 2) There exists a vector θ=col(θ i ) 1:N , so that The elements of θ satisfy 0<θ i ≤1, and the following conditions are met: 3) Let Ψ = diag(θ i ) i:N ,but is a symmetric Laplacian matrix corresponding to an undirected graph; S3. Establishing the system model of vehicle node i S31. Establishing a longitudinal dynamics model of the vehicle by analyzing the longitudinal dynamics of the vehicle, wherein the longitudinal dynamics model is expressed as: Among them, s i (t), v i (t) and a i (t) represent the position, velocity and acceleration of the i-th vehicle, τ i represents the time constant of the engine, u i (t) is the control input; S32. Define the system state of vehicle node i as Then the system equation of the i-th vehicle is expressed as: In the formula, Indicates the system status, n = 3, is the control input, m = 1, and S33. Since each ICV has a positioning sensor, a speed sensor and an acceleration sensor, each vehicle can output its own position, speed and acceleration status. Therefore, the output matrix of the i-th vehicle is expressed as: in, represents the sensor output of the i-th vehicle, p = n = 3, and is a known constant matrix; S4. Establish a system model for the fleet ICVs group S41. The system state of the fleet is expressed as The system input for the corresponding fleet is The system equation of the fleet is expressed as: In the formula, S42. Considering the unknown input disturbance experienced by the fleet In the case of , the system state space equation of the fleet ICVs group is derived as: In the formula, Assume that, for i=1, 2, ..., N, rank(C i D) = q; In order to make the description more concise and clear, (t) will be omitted if there is no ambiguity; S5. Based on the fact that each vehicle fully trusts only the detection output of its own ICVs sensor and the control input from its own controller, and the vehicle management system assigns corresponding information trust levels to the information received from neighboring vehicles, a distributed unknown input observer that can achieve the estimation target is designed. The expression of the distributed unknown input observer is: In the formula, represents the state of the distributed unknown input observer, Represents the estimation of the global state by the i-th vehicle; the matrix as well as They are all determined by solving the distributed unknown input observer coefficient matrix algorithm, and ensuring the stability of the entire residual system so that the distributed unknown input observer can achieve the estimation target; where: The algorithm for solving the coefficient matrix of the distributed unknown input observer is as follows: initialize the observers of all vehicles (i=1, 2, ..., N), and perform the following steps for each vehicle i=1, 2, ..., N: 1: Set H i ; 2: Use formula (9) to solve the matrix P that satisfies the conditions i , and substitute into formula (8) to calculate E i ; 3: Calculate formula (11) and find the matrix X that satisfies the conditions i and Y i ; 4: Find K io , so that (A io -K io C io ) is the Hurwitz matrix; 5: X i and K io Substitute into formula (11) to calculate K i ; 6: Use Lemma 1 to find suitable θ and γ to meet the requirements, thereby obtaining an appropriate γ i ; 7: Y i and γ i Substituting into formula (13) we get M i ; 8: Use formulas (15) and (19) to calculate N respectively i and L i ; 9: Select H i and the calculated L i 、M i 、E i Substitute into formula (6) and use a ij (t) to obtain the estimation result of the global system state by the first vehicle; Where: The formula (8) is: P i =(E i C i +I Nn ), Where: Pi is the intermediate matrix used to solve the distributed observer matrix of the i-th vehicle; Ei represents the observer parameter matrix of the i-th vehicle, which is used to process the feedback and correction of the sensor output yi; Ci represents the output matrix of the i-th vehicle, which maps the vehicle state x(t) to the sensor output yi; INn is the unit matrix of (N×n)×(N×n) dimensions; The formula (9) is: i B=H i [I m ]i |N , P i D=0, Where B is the input matrix associated with the control input u(t), which is used to describe how the input affects the system state; Hi is the parameter matrix in the distributed observer; Among them, I m is an m×m-dimensional identity matrix; D is the disturbance input matrix associated with the unknown input d(t), which is used to describe the impact of the unknown disturbance on the system state; The formula (11) is: Where Θi is the observability matrix of the analysis system related to the i-th vehicle; ζi represents the dimension of the unobservable subspace related to the i-th vehicle; A is the state transition matrix of the system; The formula (13) is: Where Mi is the parameter matrix related to the i-th vehicle in the distributed observer; γi is the scalar gain coefficient used to adjust the weight of the matrix Mi; Yi is used to describe the unobservable state part of the system, and the column vector constitutes the orthogonal basis of ker(Θi); The formula (15) is: i =P i AK i C i , Where Ni is the state matrix of the distributed observer; Ki is the observer gain matrix of the i-th vehicle; The formula (19) is: i =X i K io (I+C i E i )-(E i C i +I)AE i , Where Li is the distributed observer parameter matrix associated with the i-th vehicle; Xi is used to extract the observable part in the state block; Kio represents the gain matrix associated with the i-th vehicle and the observable part; I represents the unit matrix; Ei represents the parameter matrix associated with the output feedback of the i-th vehicle in the distributed observer.

2. The vehicle operation status monitoring method under a zero-trust network architecture according to claim 1 is characterized in that: The evaluation principles of vehicle trust level under the ZTNA architecture are: 1) Vehicles with the lowest trust level can still exchange useful information with each other; 2) Sending real information can increase the trust between the sender and the receiver; 3) The degree of trust is limited; 4) Any trust level above the minimum has a time limit; 5) When a vehicle receives false information, it should quickly reduce the trust level of the sender of the information; 6) When the vehicle identifies that the sender is conducting an attack, it will immediately reduce its trust level to zero and stop receiving any messages from the sender; 7) When a vehicle receives conflicting information from multiple sources, it tends to believe the view held by the majority and give more trust to the information sent by vehicles with higher trust levels.

3. The vehicle operation status monitoring method under a zero-trust network architecture according to claim 1 is characterized in that: The specific process of solving the coefficient matrix of the distributed unknown input observer is as follows: The state estimation error of the i-th vehicle is defined as From this we can get: make: P i =(E i C i +I Nn ), (8) If H is satisfied i u i =P i Bu and P i D=0, which is equivalent to: P i B=H i [Im] i|N ,P i D=0 (9) Then we have: By separating the observable and unobservable parts of the system (P i A, C i ), for this incompletely observable system, there exists a non-negative integer Make So, It is the system (P i A, C i )of dimensional unobservable subspace; Select a matrix Its columns can form ker(Θ i ), that is, im(Y i )=ker(Θ i );and It is to ensure is a matrix of orthogonal matrices; in, as well as also, C i Y i =0; obviously (A io , C io ) is observable; In addition, it is worth noting that as well as Always holds true; because (A io , C io ) is observable, so there must be a matrix K io , so that (A io -K io C io ) is a Hurwitz matrix; such a K can be found in a number of ways io ; If we order: K i =X i K io (12) Then (P i AK i C i ) will only be affected by the observable part; and let: We have: If we set: N i =P i A-K i C i , (15) Then we can get: Our goals are to achieve: Right now: By comparing equation (16) and equation (18), we can see that L i Must be: THE i =X i K io (I+C i AND i )-(AND i C i +I)AE i . (19) For the unobservable part, we can also deduce and Let e ​​= col(e i ) 1：N , X=diag(X i ) 1：N , Y=diag(Y i ) 1：N , A o =diag(Y io ) 1：N , A r =diag(A ir ) 1：N , A u =diag(A iu ) 1:N , C o =diag(C io ) 1:N , K o =diag(K io ) 1:N as well as We can get: and Notice We can deduce from equations (22) and (23): Furthermore, it can be expressed as: Among them, Th o =A o -K o C o , According to Lemma 1, Lemma 2 is given; Lemma 2: Assume that the communication network is strongly connected, and (A,col(C i ) 1:N ) is observable; then the matrix It is positive; Because (A o -K o C o ) is a Hurwitz matrix, so As can be seen from equation (23), we only need to ensure that is stable to ensure that the entire residual system (25) is stable. Select the following Lyapunov function: in, θ i In Lemma 1, γ min =min(γ i ) 1：N . It can be deduced that: According to Lemma 1, we know that 0<θ i ≤1; therefore, This means: in, Because the matrix is positive definite, and Υ is defined in Lemma 1; Therefore, if but Therefore, equation (26) is stable, that is, the entire residual system is stable, indicating that the distributed observer proposed in equation (6) can achieve the estimation goal.

Images