AGV vehicle physical layer authentication method based on path loss and speed and position estimation

Abstract

The invention discloses an AGV (Automatic Guided Vehicle) physical layer authentication method based on path loss and speed and position estimation. The method comprises the following steps that: a plurality of edge computing units receive IDs (Identity) and speeds and positions thereof reported by to-be-authenticated AGVs (Automatic Guided Vehicles); each edge calculation unit is used for extracting a channel frequency response matrix from the signal sent by the trolley and inputting the channel frequency response matrix into a Gaussian process regression model to obtain predicted path loss, position and speed; calculating acceptance domains of the path loss, the position and the speed according to the reported ID and the predicted path loss, the position and the speed; calculating path loss at the current moment according to the reported position; judging whether path loss and speed are in an acceptance domain or not, if so, authenticating that the trolley is legal, judging whether the position is in the acceptance domain or not, and if so, authenticating that the trolley does not encounter position forgery / spoofing attack; otherwise, the trolley is authenticated to be illegal; the authentication result is reported to other edge computing units; and finally authenticating the legality of the trolley according to all authentication results. According to the invention, the authentication accuracy and robustness are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless information security, and in particular relates to an AGV (Automated Guided Vehicle) vehicle physical layer authentication method based on path loss and speed and position estimation. Background Art

[0002] The Industrial Internet is the product of the deep integration of the new generation of information and communication technology and the real economy. Through the interconnection of equipment and the sharing of data resources, it has spawned new business models and organizational forms in the fields of manufacturing, energy, transportation, etc., and promoted the transformation and upgrading of industries to networking, intelligence, and greenness. In the Industrial Internet scenario, AGVs are widely connected to warehousing and logistics transportation systems. Through intelligent scheduling of AGVs for collaborative operations, they can efficiently complete various logistics and transportation tasks, playing a huge role in improving the level of system intelligence and reducing warehousing costs.

[0003] However, the mobile access of large-scale AGVs has led to a rapid increase in the number of communication entities in the system, a relative reduction in communication resources, and a significant increase in authentication delay. How to ensure successful communication when attacked has become an urgent problem to be solved. Traditional security authentication methods are mainly implemented using encryption-based authentication mechanisms, but this security mechanism has some security vulnerabilities. For example, there is a risk of key leakage during management, and it may be intercepted by opponents during distribution. Secondly, the security of the encryption authentication mechanism mainly depends on the length of the key. In the past, when the computing power of computers was limited, this method could ensure a certain degree of security. However, with the research and development of computer computing power and cryptography, the risk of encryption authentication mechanisms being cracked has increased, and it has gradually been unable to meet the increasing security needs. In addition, the resources of AGVs themselves are relatively insufficient, and they are not suitable for a large number of encryption and decryption calculations. In addition, the explicit authentication mechanism has a large authentication delay and low communication efficiency. Therefore, a new security authentication mechanism is urgently needed to solve these problems, and the physical layer authentication method can effectively avoid the security vulnerabilities of traditional encryption-based authentication schemes. Different RF components will inevitably have some deviations during the manufacturing process due to different production processes. These deviations will cause the signals they send to have different characteristics, which are mainly reflected in some equipment and channel characteristics such as carrier frequency offset, I / Q imbalance, channel frequency response, etc. Since these characteristics are not forgeable and cannot be completely cloned, different RF components can be distinguished by identifying the differences between these characteristics.

[0004] However, existing physical layer authentication methods mainly focus on authentication scenarios with fixed locations. When external attackers attack vehicles through location spoofing and forgery attacks, these existing physical layer authentication methods are unable to detect and resist these unknown attacks, which affects the performance of the authentication scheme. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides an AGV vehicle physical layer authentication method based on path loss and speed and position estimation. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0006] An embodiment of the present invention provides an AGV vehicle physical layer authentication method based on path loss and speed and position estimation, the method comprising:

[0007] Multiple edge computing units respectively receive the ID information reported by the AGV to be authenticated and its current speed and location information;

[0008] For each edge computing unit, the execution steps include: the edge computing unit extracts a channel frequency response matrix from the signal sent by the AGV to be authenticated; the channel frequency response matrix is ​​input into a trained Gaussian process regression model to obtain the path loss, position information and speed information predicted at the current moment; the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed are calculated according to the ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information; the path loss at the current moment is calculated according to the position information reported by the AGV to be authenticated; it is determined whether the path loss and speed information at the current moment are within In the corresponding acceptance domain, if they are all in: the edge computing unit authenticates that the AGV car to be authenticated is legal, and continues to determine whether the current location information is in the acceptance domain of the location. If so, the edge computing unit authenticates that the AGV car to be authenticated has not encountered a location forgery / deception attack; otherwise: the edge computing unit authenticates that the AGV car to be authenticated is illegal; the authentication result is reported to other edge computing units; the number of edge computing units that authenticate the AGV car to be authenticated as legal is counted, and the overall authentication rate at the current moment is calculated based on the counted number, and it is determined whether the overall authentication rate reaches the overall authentication rate threshold. If so, the AGV car to be authenticated is finally authenticated as legal.

[0009] In one embodiment of the present invention, the channel frequency response matrix extracted by the edge computing unit from the signal sent by the AGV to be authenticated is expressed as:

[0010]

[0011] Among them, H A represents the channel frequency response matrix extracted by the edge computing unit, Z ρ Represents the signal sent by the AGV to be authenticated, χ ρrepresents the pilot signal for estimating the channel frequency response matrix and path loss that can be obtained by both the edge computing unit and the AGV to be authenticated, m = 1, 2…, M, M represents the number of subcarriers, represents the channel frequency response of the mth subcarrier, represents the channel frequency response of the mth subcarrier at time t, () T Indicates the transpose operation, () -1 Represents the inverse operation.

[0012] In one embodiment of the present invention, the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment are calculated according to the self ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information, including:

[0013] According to the ID information reported by the AGV to be authenticated, the path loss error, position error and speed error at multiple historical moments are obtained, and the average path loss error, average position error and average speed error at the current moment are calculated based on the obtained path loss error, position error and speed error;

[0014] The acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment are calculated according to the predicted path loss, the position information and the speed information, as well as the average path loss error, the average position error and the average speed error.

[0015] In one embodiment of the present invention, the acceptance domain of the path loss at the current moment, the acceptance domain of the position, and the acceptance domain of the speed are calculated based on the predicted path loss, position information, and speed information, as well as the average path loss error, the average position error, and the average speed error. The formula is expressed as follows:

[0016]

[0017] in, represents the acceptance domain of the path loss at time t, represents the predicted path loss at time t, represents the average path loss error at time t, z PL represents a first control parameter;

[0018]

[0019]

[0020] in, represents the receptive field of the position on the x-axis at time t, Represents the predicted position information on the x-axis at time t, represents the average position error on the x-axis at time t, z loc represents the second control parameter, represents the receptive field of the position on the y-axis at time t, Represents the predicted position information on the y-axis at time t, Represents the average position error on the y-axis at time t;

[0021]

[0022] in, represents the receptive domain of the velocity at time t, represents the speed information predicted at time t, represents the average speed error at time t, z sp represents the third control parameter.

[0023] In one embodiment of the present invention, determining whether the path loss and speed information at the current moment are within the corresponding acceptance domain includes:

[0024] Determine whether the path loss at the current moment is within the acceptance domain of the path loss. If not, the edge computing unit authenticates that the AGV to be authenticated is illegal. If so, continue to determine whether the speed information at the current moment is within the acceptance domain of the speed. If not, the edge computing unit authenticates that the AGV to be authenticated is illegal. If so, the edge computing unit authenticates that the AGV to be authenticated is legal, and continues to determine whether the position information at the current moment is within the acceptance domain of the position. If not, the edge computing unit authenticates that the AGV to be authenticated has encountered a position forgery / deception attack. If so, the edge computing unit authenticates that the AGV to be authenticated has not encountered a position forgery / deception attack.

[0025] In one embodiment of the present invention, the process of training the Gaussian process regression model includes:

[0026] Each edge computing unit has multiple tables, each of which is used to record the historical information of different legal AGVs; the historical information includes the path frequency response matrix, path loss value, speed information and location information corresponding to multiple historical moments;

[0027] Using the historical information as a training data set, and assigning a label to each training data in the training data set;

[0028] Gaussian process regression is performed according to the training data set and the labels to obtain an optimal solution for the hyperparameters, and the Gaussian process model corresponding to the optimal solution is used as the trained Gaussian process model.

[0029] In one embodiment of the present invention, performing Gaussian process regression according to the training data set and the label to obtain an optimal solution for the hyperparameter includes:

[0030] Calculate the covariance value between any two training data in the training data set according to the covariance function, and form a covariance matrix with all covariance values;

[0031] A marginal log-likelihood function is constructed according to the label and the covariance matrix, and a partial derivative of the marginal log-likelihood function is calculated to maximize the marginal log-likelihood function related to the hyperparameter to obtain an optimal solution for the hyperparameter.

[0032] In one embodiment of the present invention, the covariance value between any two training data in the training data set is calculated according to the covariance function, and the formula is expressed as follows:

[0033]

[0034] Among them, (h tri ,h trj ) represents any two training data in the training data set, i, j = 1, 2, ..., N, N represents the number of training data in the training data set, k(h tri ,h trj ) represents the covariance function, θ={σ f ,σ l ,σ n} represents the hyperparameter of the Gaussian process regression model, σ f 2 represents the signal variance, σ l represents the characteristic length, σ n 2 The variance of the input noise that characterizes the randomness in wireless communications, Indicates h tri 、h trj The autocorrelation coefficient between T represents the transpose operation, Means defined as.

[0035] In one embodiment of the present invention, the partial derivative of the marginal log-likelihood function is calculated to maximize the marginal log-likelihood function associated with the hyperparameter to obtain the optimal solution of the hyperparameter, which is expressed as follows:

[0036]

[0037] Among them, logp(y|H tr ) represents the marginal log-likelihood function, represents logp(y|H tr ) Find the hyperparameter θ p The partial derivative of , y represents the label set of all legal AGV cars in the training data set, K represents the covariance matrix, It means to find the hyperparameter θ for Kp The partial derivative of p ∈θ, tr() represents the trace operation, () T Indicates the transpose operation, () -1 Represents the inverse operation.

[0038] In one embodiment of the present invention, for each edge computing unit, the execution step further includes:

[0039] If the edge computing unit authenticates that the AGV to be authenticated is legal, the path frequency response matrix, path loss value, speed and position information of the AGV to be authenticated are updated to the table corresponding to the edge computing unit; wherein the updated table is used to periodically retrain the Gaussian process regression model to obtain a new trained Gaussian process regression model.

[0040] Beneficial effects of the present invention:

[0041] The present invention proposes an AGV vehicle physical layer authentication method based on path loss and speed and position estimation, which inputs the extracted physical layer channel characteristics, namely the channel frequency response matrix, into a Gaussian process regression model to obtain predicted path loss, speed information and position information, and calculates the acceptance domain of path loss, the acceptance domain of speed and the acceptance domain of position according to the current actual path loss, speed information and position information, as well as the predicted path loss, speed information and position information, and then judges the legitimacy of the AGV car to be authenticated according to the acceptance domain, thereby improving the authentication accuracy of the AGV car, and further judges whether the AGV car to be authenticated has suffered a position forgery / deception attack. Since the legitimacy of the AGV car has been successfully authenticated when it suffers a position forgery / deception attack, the possibility of a legitimate AGV car being erroneously rejected is reduced, thereby improving the communication efficiency of the AGV car to be authenticated; the present invention utilizes multiple edge computing units for collaborative authentication, and the AGV car to be authenticated sends signals to the surrounding edge computing units, and the multiple edge computing units jointly authenticate it, which can not only enhance the robustness of the authentication system, avoid the situation where a single edge computing unit is destroyed and the authentication system fails, but also improve the authentication accuracy. In general, the physical layer authentication method based on path loss and the speed and position characteristics of the AGV proposed in the present invention solves the problems of low accuracy, low system robustness, and difficulty in detecting and resisting location forgery / deception attacks in existing vehicle authentication, and can be widely used in smart factories, smart manufacturing, smart homes and other fields.

[0042] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1It is a flow chart of a method for physical layer authentication of an AGV vehicle based on path loss and speed and position estimation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0045] See also Figure 1 The embodiment of the present invention provides an AGV vehicle physical layer authentication method based on path loss and speed and position estimation, which specifically includes the following steps:

[0046] S10. Multiple edge computing units respectively receive the ID information of the AGV to be authenticated and its current speed information and location information.

[0047] For each edge computing unit, the execution steps include:

[0048] S20. The edge computing unit extracts a channel frequency response matrix from the signal sent by the AGV to be authenticated.

[0049] The channel frequency response matrix extracted by the edge computing unit of the embodiment of the present invention from the signal sent by the AGV to be authenticated is expressed as follows:

[0050]

[0051] Among them, H A represents the channel frequency response matrix extracted by the edge computing unit, Z ρ Represents the signal sent by the AGV to be authenticated, χ ρ represents the pilot signal for estimating the channel frequency response matrix and path loss that can be obtained by both the edge computing unit and the AGV to be authenticated, m = 1, 2…, M, M represents the number of subcarriers, represents the channel frequency response of the mth subcarrier, represents the channel frequency response of the mth subcarrier at time t, () T Indicates the transpose operation, () -1 Represents the inverse operation.

[0052] S30, inputting the channel frequency response matrix into the trained Gaussian process regression model to obtain the path loss, position information and speed information predicted at the current moment.

[0053] Since the Gaussian process regression model is a non-parametric regression model, it has no specific function form, but performs model regression by learning the distribution of data. It can handle complex nonlinear relationships, so that the model curve can continuously fit the training data, and finally obtain an ideal Gaussian process regression model. However, since the speed of the AGV car is uncertain and there is no obvious pattern, it is more appropriate to choose the Gaussian process regression model for prediction. The process of training the Gaussian process regression model proposed in the embodiment of the present invention includes: each edge computing unit establishes multiple tables, each table is used to record the historical information of different legal AGV cars; the historical information includes the path frequency response matrix, path loss value, speed information and position information corresponding to multiple historical moments; the historical information is used as a training data set, and a label is assigned to each legal AGV car; Gaussian process regression is performed based on the training data set and the label to obtain the optimal solution of the hyperparameters, and the Gaussian process model corresponding to the optimal solution is used as the trained Gaussian process model. More specifically:

[0054] On each edge computing unit, the channel frequency response matrix, path loss value, speed information, and location information of all legal AGVs at multiple historical moments are pre-established and mapped into a table. The historical information of all legal AGVs in the table is used as the training data set H. tr =(h tr1 ,…,h tri ,…,h trN ) T ,h tri represents the i-th training data in the training data set, i = 1, 2, ..., N, N represents the number of training data in the training data set, and each training sample includes the channel frequency response matrix, path loss value, speed information and location information, denoted as h tri =(CFR i ,PL i ,sp i ,loc i ) T , CFR i represents the channel frequency response matrix of the i-th training sample, PL i represents the path loss value of the i-th training sample, sp i Indicates the speed information of the i-th training sample, loc i represents the location information of the i-th training sample; and assigns a label to each training data in the training sample set, denoted as y = (ID1,…,ID i ,…,ID N ) T , ID i Represents the label assigned to the i-th training data in the training data set.

[0055] Furthermore, the embodiment of the present invention performs Gaussian process regression based on the training data set and the label to obtain the optimal solution of the hyperparameter, including: calculating the covariance value between any two training data in the training data set according to the covariance function, and forming a covariance matrix from all covariance values; constructing a marginal log-likelihood function according to the label and the covariance matrix, and calculating the partial derivative of the marginal log-likelihood function to maximize the marginal log-likelihood function related to the hyperparameter to obtain the optimal solution of the hyperparameter. Among them,

[0056] The covariance value between any two training data in the training data set is calculated according to the covariance function. The formula is expressed as:

[0057]

[0058] Among them, (h tri ,h trj ) represents any two training data in the training data set, i, j = 1, 2, ..., N, N represents the number of training data in the training data set, k(h tri ,h trj ) represents the covariance function, θ={σ f ,σ l ,σ n} represents the hyperparameter of the Gaussian process regression model, σ f 2 represents the signal variance, σ l represents the characteristic length, σ n 2 The variance of the input noise that characterizes the randomness in wireless communications, Indicates h tri 、h trj The autocorrelation coefficient between T represents the transpose operation, It is defined as. The covariance matrix is ​​composed of all covariance values, and the formula is expressed as:

[0059]

[0060] Among them, K(H tr ,H tr ) represents the covariance matrix, which is denoted as K in subsequent use.

[0061] Next, a marginal log-likelihood function is constructed according to the label and the covariance matrix, and the partial derivative of the marginal log-likelihood function is calculated to maximize the marginal log-likelihood function related to the hyperparameter to obtain the optimal solution of the hyperparameter. The formula is expressed as follows:

[0062]

[0063] Among them, logp(y|H tr) represents the marginal log-likelihood function, represents logp(y|H tr ) Find the hyperparameter θ p The partial derivative of , y represents the label set of all legal AGV cars in the training data set, K represents the covariance matrix, It means to find the hyperparameter θ for K p The partial derivative of p ∈θ, tr() represents the trace operation, () T Indicates the transpose operation, () -1 In this embodiment of the present invention, for the hyperparameter θ={σ f ,σ l ,σ n}In the solution, each time a parameter σ in the hyperparameter θ is f or l or n Perform optimization and solution, that is, each time the optimization solution is performed, θ p is f or l or n , after optimizing and solving the three parameters, the optimal solution of the hyperparameters is obtained, and the Gaussian process model corresponding to the optimal solution is used as the trained Gaussian process model.

[0064] Then, the channel frequency response matrix H extracted by S20 is A Input the trained Gaussian process regression model to obtain the path loss, speed information and location information predicted at the current moment. Among them, the path loss predicted at time t is recorded as The predicted speed information at time t is recorded as The predicted position information at time t is recorded as Represents the predicted position information on the x-axis at time t, Represents the predicted position information on the y-axis at time t.

[0065] S40, calculating the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment according to the self ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information.

[0066] The embodiment of the present invention provides an optional solution, which calculates the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment according to the self-ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information, including: obtaining the path loss error, position error and speed error of multiple historical moments according to the self-ID information reported by the AGV to be authenticated, and calculating the average path loss error, average position error and average speed error at the current moment according to the obtained path loss error, position error and speed error; calculating the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment according to the predicted path loss, position information and speed information, as well as the average path loss error, average position error and average speed error. More specifically:

[0067] According to the ID information reported by the AGV to be authenticated, the path loss error, position error and speed error of the t-1 historical moment are obtained; the corresponding cumulative error is calculated based on the path loss error, position error and speed error of the t-1 historical moment; then the average path loss error, average position error and average speed error of the current moment are calculated based on the accumulated error; on the edge computing unit, a table will be established to record the path loss error, position error and speed error of multiple historical moments. When calculating the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment, the path loss error, position error and speed error of all moments before this moment are obtained from the table. Among them,

[0068] Average path loss error at the current moment The formula is:

[0069]

[0070] in, represents the path loss error at the vth historical moment, It means to find the cumulative error of the path loss at the t-1 historical moment.

[0071] Average position error at the current moment The formula is:

[0072]

[0073]

[0074] in, represents the position error on the x-axis at the vth historical moment, It means to find the cumulative error on the x-axis at t-1 historical moments. represents the position error on the y-axis at the vth historical moment, It means to find the cumulative error on the y-axis at t-1 historical moments.

[0075] Average speed error at the current moment The formula is:

[0076]

[0077] in, represents the velocity error at the vth historical moment, It means to find the cumulative error of the speed information at the t-1 historical moment.

[0078] Finally, the acceptance domain of the path loss, the acceptance domain of the position, and the acceptance domain of the speed at the current moment are calculated based on the predicted path loss, position information, and speed information, as well as the average path loss error, the average position error, and the average speed error. The formula is expressed as follows:

[0079]

[0080] in, represents the acceptance domain of the path loss at time t, represents the predicted path loss at time t, represents the average path loss error at time t, z PL represents the first control parameter, the first control parameter z PL The value is set according to the actual situation;

[0081]

[0082]

[0083] in, represents the receptive field of the position on the x-axis at time t, Represents the predicted position information on the x-axis at time t, represents the average position error on the x-axis at time t, z loc represents the second control parameter, the second control parameter z loc The value is set according to the actual situation. represents the receptive field of the position on the y-axis at time t, Represents the predicted position information on the y-axis at time t, Represents the average position error on the y-axis at time t;

[0084]

[0085] in, represents the receptive domain of the velocity at time t, represents the speed information predicted at time t, represents the average speed error at time t, z sp represents the third control parameter, the third control parameter z sp The value is set according to the actual situation.

[0086] S50: Calculate the path loss at the current moment according to the position information reported by the AGV to be authenticated.

[0087] In the process of channel estimation, the path loss between the edge computing unit and the AGV to be authenticated at the current moment is calculated, and the formula is expressed as:

[0088] PL ID (t) = E × log 10 (d(t))+F;

[0089] Among them, PL ID (t) represents the path loss at time t, E and F represent the path loss related distance between the AGV to be authenticated and the edge computing unit and the reference path loss respectively; d(t) represents the distance between the AGV to be authenticated and the edge computing unit at the current moment, in meters, and the formula is expressed as:

[0090]

[0091] Among them, loc A,x (t) represents the position information of the edge computing unit on the x-axis, loc A,y (t) represents the position information of the edge computing unit on the y-axis.

[0092] S60. Determine whether the path loss and speed information at the current moment are in the corresponding acceptance domain. If so, the edge computing unit authenticates that the AGV to be authenticated is legal, and continues to determine whether the position information at the current moment is in the acceptance domain of the position. If so, the edge computing unit authenticates that the AGV to be authenticated has not encountered a position forgery / deception attack; otherwise, the edge computing unit authenticates that the AGV to be authenticated is illegal.

[0093] The embodiment of the present invention determines whether the path loss and speed information at the current moment are in the corresponding acceptance domain, including: determining whether the path loss at the current moment is in the acceptance domain of the path loss, if not, the edge computing unit authenticates that the AGV car to be authenticated is illegal, if yes, then continue to determine whether the speed information at the current moment is in the acceptance domain of the speed, if not, the edge computing unit authenticates that the AGV car to be authenticated is illegal, if yes, then the edge computing unit authenticates that the AGV car to be authenticated is legal, and continue to determine whether the position information at the current moment is in the acceptance domain of the position, if not, the edge computing unit authenticates that the AGV car to be authenticated encounters a position forgery / deception attack, if yes, the edge computing unit authenticates that the AGV car to be authenticated does not encounter a position forgery / deception attack. More specifically:

[0094] S601, determine the path loss: if the path loss PL at time t is ID (t) In the acceptance domain of path loss Then calculate the path loss error at the current moment At this time According to the path loss PL at the current moment ID (t) and the path loss predicted at the current time Calculate and execute S 602; if the path loss PL at time t is ID (t) Not in the acceptance domain of the path loss If the AGV to be authenticated is illegal, the current state is retained. That is, the path loss error at the current moment is taken as the path loss error at the previous moment, and S70 is executed;

[0095] S602, determine the speed information: if the speed information sp at time t ID (t) in the receptive field of velocity If , the AGV to be authenticated is determined to be legal, and the speed error at the current moment is calculated. At this time According to the current speed information sp ID (t) and the speed information predicted at the current moment Calculate and execute S603; if the speed information sp at time t ID (t) is not in the velocity acceptance domain If the AGV to be authenticated is illegal, the current εsp is retained. ID (t) = εsp ID (t-1), execute S70;

[0096] S603, determine the location information: if the location information at time t is loc i,x (t), loc i,y (t) The receptive field at the position If the AGV to be authenticated is not attacked by location fraud / forgery, the current location information is calculated. At this time According to the current loc i,x (t), loc i,y (t) and the current predicted Calculated; if the position information loc at time t i,x (t), loc i,y (t) The receptive fields that are not at the position If the AGV to be authenticated is attacked by location fraud / forgery, the current moment is retained.

[0097] Among them, the updated or retained by S601 to S603 The table stored in the edge computing unit for recording path loss error, position error and speed error is used to periodically recalculate the corresponding acceptance domain.

[0098] Furthermore, for each edge computing unit, the execution step further includes:

[0099] If the edge computing unit authenticates the AGV to be authenticated as legitimate, the path frequency response matrix, path loss value, speed and position information of the AGV to be authenticated are updated to the table corresponding to the edge computing unit; wherein the updated table is used to periodically retrain the Gaussian process regression model to obtain a new trained Gaussian process regression model. It can be seen that adding more information about legitimate AGVs to the table, because samples are added during the training process, makes the Gaussian process regression model obtained after periodic retraining have a better fitting effect.

[0100] S70. Report the authentication result to other edge computing units; used for subsequent statistical authentication of each edge computing unit.

[0101] S80, count the number of edge computing units that authenticate the AGV to be authenticated as legal, calculate the overall authentication rate at the current moment based on the counted number, and determine whether the overall authentication rate reaches the overall authentication rate threshold. If so, the AGV to be authenticated is finally authenticated as legal.

[0102] The embodiment of the present invention pre-sets an overall authentication rate threshold to evaluate the authentication results of all edge computing units as a whole. Assuming that the overall authentication rate threshold is set to β, and the overall authentication rate at the current moment is α, Among them, M represents the number of edge computing units participating in the authentication of the AGV to be authenticated, and m represents the number of edge computing units that have successfully authenticated. It is compared with the overall authentication rate threshold β: if α≥β, the AGV to be authenticated is finally judged to be legal; if α<β, the AGV to be authenticated is finally judged to be illegal.

[0103] In summary, the AGV vehicle physical layer authentication method based on path loss and speed and position estimation proposed in the embodiment of the present invention inputs the extracted physical layer channel characteristics, namely the channel frequency response matrix, into the Gaussian process regression model to obtain the predicted path loss, speed information and position information, and calculates the acceptance domain of the path loss, the acceptance domain of the speed and the acceptance domain of the position according to the current actual path loss, speed information and position information, and the predicted path loss, speed information and position information, and then judges the legitimacy of the AGV to be authenticated according to the acceptance domain, thereby improving the authentication accuracy of the AGV, and further judging the legitimacy of the AGV to be authenticated. Whether the AGV car has been subjected to a location forgery / deception attack, since the legitimacy of the AGV car has been successfully authenticated when it is subjected to a location forgery / deception attack, the possibility of a legitimate AGV car being rejected by mistake is reduced, and the communication efficiency of the AGV car to be authenticated is improved; the embodiment of the present invention makes full use of multiple edge computing units for collaborative authentication, and the AGV car to be authenticated sends a signal to the surrounding edge computing units, and multiple edge computing units jointly authenticate it, which can not only enhance the robustness of the authentication system, avoid the situation where a single edge computing unit is destroyed and the authentication system fails, but also improve the authentication accuracy. In general, the physical layer authentication method based on path loss and the speed and position characteristics of the AGV car proposed in the embodiment of the present invention solves the problems of low accuracy, low system robustness, and difficulty in detecting and resisting location forgery / deception attacks in existing vehicle authentication, and can be widely used in smart factories, smart manufacturing, smart homes and other fields.

[0104] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0105] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the specification and its drawings. In the specification, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude multiple situations. Certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0106] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A physical layer authentication method for AGV vehicles based on path loss and speed and position estimation, characterized in that: The method comprises: Multiple edge computing units respectively receive the ID information reported by the AGV to be authenticated and its current speed and location information; For each edge computing unit, the execution steps include: the edge computing unit extracts a channel frequency response matrix from the signal sent by the AGV to be authenticated; the channel frequency response matrix is ​​input into a trained Gaussian process regression model to obtain the path loss, position information and speed information predicted at the current moment; the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed are calculated according to the ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information; the path loss at the current moment is calculated according to the position information reported by the AGV to be authenticated; it is determined whether the path loss and speed information at the current moment are within In the corresponding acceptance domain, if they are all in: the edge computing unit authenticates that the AGV car to be authenticated is legal, and continues to determine whether the current location information is in the acceptance domain of the location. If so, the edge computing unit authenticates that the AGV car to be authenticated has not encountered a location forgery / deception attack; otherwise: the edge computing unit authenticates that the AGV car to be authenticated is illegal; the authentication result is reported to other edge computing units; the number of edge computing units that authenticate the AGV car to be authenticated as legal is counted, and the overall authentication rate at the current moment is calculated based on the counted number, and it is determined whether the overall authentication rate reaches the overall authentication rate threshold. If so, the AGV car to be authenticated is finally authenticated as legal.

2. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 1 is characterized in that: The channel frequency response matrix extracted by the edge computing unit from the signal sent by the AGV to be authenticated is expressed as: Among them, H A represents the channel frequency response matrix extracted by the edge computing unit, Z ρ Represents the signal sent by the AGV to be authenticated, χ ρ represents the pilot signal for estimating the channel frequency response matrix and path loss that can be obtained by both the edge computing unit and the AGV to be authenticated, m = 1, 2…, M, M represents the number of subcarriers, represents the channel frequency response of the mth subcarrier, represents the channel frequency response of the mth subcarrier at time t, () T Indicates the transpose operation, () -1 Represents the inverse operation.

3. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 1 is characterized in that: According to the ID information reported by the AGV to be authenticated, as well as the predicted path loss, position information and speed information, the acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment are calculated, including: According to the ID information reported by the AGV to be authenticated, the path loss error, position error and speed error at multiple historical moments are obtained, and the average path loss error, average position error and average speed error at the current moment are calculated based on the obtained path loss error, position error and speed error; The acceptance domain of the path loss, the acceptance domain of the position and the acceptance domain of the speed at the current moment are calculated according to the predicted path loss, the position information and the speed information, as well as the average path loss error, the average position error and the average speed error.

4. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 3 is characterized in that: The acceptance domain of the path loss, the acceptance domain of the position, and the acceptance domain of the speed at the current moment are calculated according to the predicted path loss, the position information, and the speed information, as well as the average path loss error, the average position error, and the average speed error. The formula is expressed as follows: in, represents the acceptance domain of the path loss at time t, represents the predicted path loss at time t, represents the average path loss error at time t, z PL represents a first control parameter; in, represents the receptive field of the position on the x-axis at time t, Represents the predicted position information on the x-axis at time t, represents the average position error on the x-axis at time t, z loc represents the second control parameter, represents the receptive field of the position on the y-axis at time t, Represents the predicted position information on the y-axis at time t, Represents the average position error on the y-axis at time t; in, represents the receptive domain of the velocity at time t, represents the speed information predicted at time t, represents the average speed error at time t, z sp represents the third control parameter.

5. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 1 is characterized in that: Determine whether the path loss and speed information at the current moment are within the corresponding acceptance domain, including: Determine whether the path loss at the current moment is within the acceptance domain of the path loss. If not, the edge computing unit authenticates that the AGV to be authenticated is illegal. If so, continue to determine whether the speed information at the current moment is within the acceptance domain of the speed. If not, the edge computing unit authenticates that the AGV to be authenticated is illegal. If so, the edge computing unit authenticates that the AGV to be authenticated is legal, and continues to determine whether the position information at the current moment is within the acceptance domain of the position. If not, the edge computing unit authenticates that the AGV to be authenticated has encountered a position forgery / deception attack. If so, the edge computing unit authenticates that the AGV to be authenticated has not encountered a position forgery / deception attack.

6. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 1 is characterized in that: The process of training a Gaussian process regression model includes: Each edge computing unit has multiple tables, each of which is used to record the historical information of different legal AGVs; the historical information includes the path frequency response matrix, path loss value, speed information and location information corresponding to multiple historical moments; Using the historical information as a training data set, and assigning a label to each training data in the training data set; Gaussian process regression is performed according to the training data set and the labels to obtain an optimal solution for the hyperparameters, and the Gaussian process model corresponding to the optimal solution is used as the trained Gaussian process model.

7. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 6 is characterized in that: Performing Gaussian process regression according to the training data set and the label to obtain an optimal solution for the hyperparameters includes: Calculate the covariance value between any two training data in the training data set according to the covariance function, and form a covariance matrix with all covariance values; A marginal log-likelihood function is constructed according to the label and the covariance matrix, and a partial derivative of the marginal log-likelihood function is calculated to maximize the marginal log-likelihood function related to the hyperparameter to obtain an optimal solution for the hyperparameter.

8. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 7 is characterized in that: The covariance value between any two training data in the training data set is calculated according to the covariance function. The formula is expressed as: Among them, (h tri ,h trj ) represents any two training data in the training data set, i, j = 1, 2, ..., N, N represents the number of training data in the training data set, k(h tri ,h trj ) represents the covariance function, θ={σ f ,σ l ,σ n } represents the hyperparameter of the Gaussian process regression model, σ f 2 represents the signal variance, σ l represents the characteristic length, σ n 2 The variance of the input noise that characterizes the randomness in wireless communications, Indicates h tri 、h trj The autocorrelation coefficient between T represents the transpose operation. Means defined as.

9. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 7 is characterized in that: The partial derivative of the marginal log-likelihood function is calculated to maximize the marginal log-likelihood function related to the hyperparameters to obtain the optimal solution of the hyperparameters. The formula is expressed as: Among them, logp(y|H tr ) represents the marginal log-likelihood function, represents logp(y|H tr ) Find the hyperparameter θ p The partial derivative of , y represents the label set of all legal AGV cars in the training data set, K represents the covariance matrix, It means to find the hyperparameter θ for K p The partial derivative of p ∈θ, tr() represents the trace operation, () T Indicates the transpose operation, () -1 Represents the inverse operation.

10. The AGV vehicle physical layer authentication method based on path loss and speed and position estimation according to claim 6, characterized in that: For each edge computing unit, the execution steps further include: If the edge computing unit authenticates that the AGV to be authenticated is legal, the path frequency response matrix, path loss value, speed and position information of the AGV to be authenticated are updated to the table corresponding to the edge computing unit; wherein the updated table is used to periodically retrain the Gaussian process regression model to obtain a new trained Gaussian process regression model.

Images