Radar anti-deception system for autonomous vehicles to identify ghost vehicles

Abstract

A radar anti-deception system of an autonomous vehicle includes a plurality of radar sensors that generate a plurality of input detections representing radio frequency (RF) signals reflected from an object, and a controller in electronic communication with the plurality of radar sensors. The one or more controllers execute instructions to determine a signal-to-noise ratio (SNR) distance ratio of the input detections generated by the plurality of radar sensors, wherein a value of the SNR distance ratio indicates that the object is a ghost vehicle. The one or more controllers also determine an effective number of particles that indicates a degree of degradation of importance sampling of each variable as part of a state variable. In response to determining that the effective number of particles is equal to or less than a predetermined threshold, the one or more controllers estimate a ghost location of the ghost vehicle.

Description

Technical Field

[0001] This disclosure relates to systems and methods for identifying ghost vehicles created when the radar sensors of an autonomous vehicle are spoofed. Background Technology

[0002] Autonomous vehicles can utilize various sensors, such as radar sensors, vision sensors, and lidar sensors, to sense their environment. While any type of autonomous sensor can be fooled, radar sensors tend to be the most easily fooled compared to other autonomous sensors. Although radar sensors are easily fooled, they are also the only autonomous sensors capable of operating in a wide range of weather conditions, making the development of anti-spoofing techniques for radar sensors advantageous.

[0003] A relatively simple method to deceive radar sensors is to use digital radio frequency (RF) memory to create ghost vehicles. Ghost vehicles can confuse or mislead autonomous vehicles. Specifically, the digital storage device can be used to generate false return RF signals by receiving, delaying, and forwarding RF signals emitted by the autonomous vehicle. The autonomous vehicle's radar system processes and manipulates these false RF signals as actual return RF signals from another vehicle in the surrounding environment, thus creating the ghost vehicle effect.

[0004] Therefore, although current anti-spoofing technologies have achieved their intended purpose, new and improved methods for identifying ghost vehicles are still needed in the field. Summary of the Invention

[0005] According to several aspects, a radar anti-spoofing system for an autonomous vehicle is disclosed. The radar anti-spoofing system includes: multiple radar sensors generating multiple input detection points representing radio frequency (RF) signals reflected from an object; and one or more controllers electronically communicating with the multiple radar sensors, wherein the one or more controllers execute instructions to determine a signal-to-noise ratio (SNR) distance ratio of the multiple input detection points generated by the multiple radar sensors, wherein the value of the SNR distance ratio indicates that the object is a ghost vehicle. The one or more controllers execute instructions to determine an importance sample of each variable as part of a state variable based on the multiple input detection points and the SNR distance ratio. The one or more controllers execute instructions to weight the importance samples of each variable as part of the state variable. The one or more controllers execute instructions to determine an effective particle count, which indicates the degree of particle degradation of the importance samples of each variable as part of the state variable. Finally, in response to determining that the effective particle count is equal to or less than a predetermined threshold, the one or more controllers estimate the ghost position of the ghost vehicle based on the state variable.

[0006] In one respect, the SNR distance ratio represents the difference in the SNR constraint variable of an object over time, where the SNR constraint variable represents the physical constraint between the measured SNR and the distance to the object.

[0007] In another aspect, the SNR distance ratio is determined by the following formula:

[0008]

[0009] Where t and t+1 represent continuous time steps, γ is the SNR constraint variable, and d T Let d represent the actual distance, and d M This indicates the measured distance.

[0010] In yet another aspect, state variables are expressed as:

[0011] X t =[x(t), y(t), v x (t), v y (t), γ(t)] T

[0012] Among them, X t These are state variables, x(t) and y(t) represent the x and y positions, and v... x (t), v y γ(t) represents the x and y velocity coordinates, and γ(t) is the SNR distance ratio.

[0013] In one aspect, the importance sampling of each variable, which is part of the state variable, is weighted by determining the following values:

[0014]

[0015]

[0016] in, This represents the weight of the sampling point at time step t. It is a posterior distribution. It is a set of sampling points, and σ j This represents the variance that depends on the dynamic range of the corresponding component.

[0017] In another respect, the effective number of particles is determined by the following formula:

[0018]

[0019] Where, N eff It is the effective number of particles, N s It is the total number of particles, and This represents the weight of the sampling point at time step t.

[0020] In another respect, the predetermined threshold is equal to half the total number of particle points.

[0021] In one aspect, one or more controllers execute instructions to re-perform a resampling operation in response to determining that the effective number of particles is equal to or less than a predetermined threshold, in order to improve the particle distribution.

[0022] In another aspect, one or more controllers execute instructions to estimate the SNR distance ratio based on state variables, thereby determining the estimated SNR distance ratio.

[0023] In another aspect, one or more controllers execute instructions to compare the estimated SNR distance ratio with a predetermined value less than 1, and determine that the ghost location is a valid ghost point in response to determining that the estimated SNR distance ratio is less than or equal to the predetermined value less than 1.

[0024] In one aspect, one or more controllers execute instructions to determine that a ghost location is a non-ghost point in response to determining that the estimated SNR distance ratio is greater than or less than 1.

[0025] In one aspect, a method for detecting and tracking ghost vehicles using a radar anti-spoofing system is disclosed. The method includes: one or more controllers determining an SNR (Side-to-Noise Ratio) ratio of input detection points generated by multiple radar sensors, wherein the value of the SNR ratio indicates that the object is a ghost vehicle. The method further includes: one or more controllers determining an importance sample of each variable as part of a state variable based on the multiple input detection points and the SNR ratio. The method further includes: weighting the importance samples of each variable as part of the state variable. The method includes: determining an effective particle count, which indicates the degree of particle degradation of the importance samples of each variable as part of the state variable. In response to determining that the effective particle count is equal to or less than a predetermined threshold, the method includes: estimating the ghost position of the ghost vehicle based on the state variable.

[0026] In another aspect, in response to determining that the effective number of particles is equal to or less than a predetermined threshold, the method includes: re-performing a resampling operation to improve the particle distribution.

[0027] In another aspect, the method includes: estimating the SNR distance ratio based on state variables, thereby determining the estimated SNR distance ratio.

[0028] In one aspect, the method includes: comparing an estimated SNR distance ratio with a predetermined value less than 1, and determining that a ghost location is a valid ghost point in response to determining that the estimated SNR distance ratio is less than or equal to the predetermined value less than 1.

[0029] In another aspect, in response to determining that the estimated SNR distance ratio is greater than or less than a predetermined value of 1, the method includes: determining that the ghost location is a non-ghost point.

[0030] In one aspect, a radar anti-spoofing system for an autonomous vehicle is disclosed. The radar anti-spoofing system includes: multiple radar sensors that generate multiple input detection points representing RF signals reflected from an object; and one or more controllers that electronically communicate with the multiple radar sensors, wherein the one or more controllers execute instructions to determine an SNR deviation coefficient of the object based on the multiple input detection points from the multiple radar sensors. The one or more controllers execute instructions to modify the innovation covariance matrix of a Kalman filter by combining the measured noise covariance matrix with the SNR deviation coefficient. The one or more controllers execute instructions to determine a spoofing detection metric that quantifies the relationship between the updated state covariance matrix determined by the Kalman filter and a measured detection distance between the autonomous vehicle and the object, wherein the value of the spoofing detection metric changes with time when the object is a ghost vehicle.

[0031] In another aspect, one or more controllers execute instructions to determine the spoofing detection metric ratio based on the standard deviation of the spoofing detection metric determined within a defined time window divided by the mean of the spoofing detection metric determined within the defined time window.

[0032] In another aspect, one or more controllers execute instructions to compare a spoofing detection metric ratio with a threshold, and determine that the object is a ghost vehicle in response to determining that the spoofing detection metric ratio is greater than the threshold. In response to determining that the spoofing detection metric ratio is less than or equal to the threshold, one or more controllers determine that the object is a real vehicle.

[0033] In another aspect, one or more controllers execute instructions to calculate the Doppler deviation coefficient of an object by determining the difference between the expected Doppler frequency and the measured Doppler frequency of the object.

[0034] Other applicable areas will become apparent from the description provided herein. It should be understood that this description and specific examples are for illustrative purposes only and are not intended to limit the scope of this disclosure. Attached Figure Description

[0035] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this disclosure in any way.

[0036] Figure 1 This is a schematic diagram of an exemplary autonomous vehicle according to an exemplary embodiment, the autonomous vehicle including an autonomous driving system and a disclosed radar anti-spoofing system for identifying and tracking ghost vehicles;

[0037] Figure 2 According to an exemplary implementation scheme Figure 1 The block diagram of the disclosed radar anti-spoofing system is shown below;

[0038] Figure 3 This is a flowchart illustrating a method for identifying and tracking ghost vehicles using a radar anti-spoofing system according to an exemplary embodiment;

[0039] Figure 4 This is a schematic diagram of another embodiment of a radar anti-spoofing system according to an exemplary embodiment; and

[0040] Figure 5 This illustrates a method for using an exemplary embodiment based on... Figure 4 The diagram shows the process flow of a radar anti-spoofing system for identifying ghost vehicles. Detailed Implementation

[0041] The following description is exemplary in nature and is not intended to limit this disclosure, its application or use.

[0042] See Figure 1 An exemplary autonomous vehicle 10 is shown. The autonomous vehicle 10 has an autonomous driving system 12, which includes a plurality of autonomous sensors 14 that communicate electronically with one or more autonomous controllers 16. In such a way... Figure 1 In the example shown, the multiple autonomous sensors 14 include multiple radar sensors 20, one or more cameras 22, an inertial measurement unit (IMU) 24, a global positioning system (GPS) 26, and a lidar 28. However, it should be understood that additional sensors may also be employed. One or more autonomous controllers 16 include an autonomous driving system 30 and a radar anti-spoofing system 32. The autonomous driving system 30 includes a vehicle detection block 34, a vehicle tracking block 36, a sensor fusion block 38, and a vehicle control block 40, which determines one or more action commands 44 based on inputs from the autonomous sensors 14 to guide the autonomous vehicle 10. The radar anti-spoofing system 32 includes a ghost vehicle detection block 46 and a ghost vehicle tracking block 48.

[0043] like Figure 1As shown, the radar anti-spoofing system 32 can be implemented as a standalone module, requiring no modification to the autonomous driving system 30. Multiple radar sensors 20 generate multiple demodulated radio frequency (RF) signals representing RF signals reflected from objects in the environment surrounding the autonomous vehicle 10, and are represented as multiple input detection points 50 received as input by the vehicle detection block 34 of the radar anti-spoofing system 32. The ghost vehicle detection block 46 of the radar anti-spoofing system 32 determines the signal-to-noise ratio (SNR) distance ratio 52 and the ghost vehicle detection points for identifying ghost vehicles, and the ghost vehicle tracking block 48 determines the ghost position 54 by tracking the motion of the ghost vehicle based on a particle filter. The ghost vehicle tracking block 48 sends the ghost position 54 to the vehicle tracking block 36 of the autonomous driving system 30. Therefore, the autonomous driving system 30 can prevent or mitigate the effects of radar sensor spoofing.

[0044] Figure 2 This is a block diagram of the radar anti-spoofing system 32. The ghost vehicle detection block 46 determines the SNR distance ratio 52 of objects in the environment based on the input detection point 50. The ghost vehicle detection block 46 determines that the object is a ghost vehicle based on the value of the SNR distance ratio 52. The SNR distance ratio 52 represents the difference of the object's SNR constraint variable c over time, where the SNR constraint variable c represents the physical constraint between the measured SNR and the object's distance. Specifically, the object's measured SNR is inversely proportional to the fourth power of the object's distance. According to Equation 1, the SNR constraint variable c is determined as:

[0045]

[0046] Where A is a constant derived from the parameters of a specific radar sensor and is independent of distance, d T d represents the actual distance. M Let represent the measured distance, and Δd be the difference between the actual distance and the measured distance. For a real vehicle in motion, Δd << d T And c≈A. However, for ghost vehicles, the measured distance d M It is determined based on the time delay caused by the deception device, but the actual distance d T It is a deception device and an autonomous vehicle 10 ( Figure 1 The distance between the ghost vehicle and the autonomous vehicle 10. If the ghost vehicle is between the deceiving vehicle and the autonomous vehicle 10, the measured distance is always less than the true distance (Δd > 0); and c < A. As the ghost vehicle approaches the autonomous vehicle 10, the SNR constraint variable c tends to zero.

[0047] The actual measured distance d of the real vehicle M and the actual distance d T The values ​​are relatively close, and the measured distance d Mand the actual distance d T The difference is due to measurement noise. Furthermore, it should be understood that the SNR constraint variable c for real vehicles is close to a constant value. In contrast, the measured distance d for ghost vehicles... M and the actual distance d T The mismatch is significant, and the difference is relatively large numerically. The SNR constraint variable c of the ghost vehicle changes over time, and increases as the ghost vehicle approaches the autonomous vehicle 10 ( Figure 1 And tends to zero. According to Equation 2, the SNR distance ratio 52 is determined, and the equation is as follows:

[0048]

[0049] Where t and t+1 represent consecutive time steps, and γ is the SNR distance ratio of 52. For real vehicles, d M (t+1)≈d T (t+1) and d M (t)≈d T (t), and the SNR distance ratio 52 is approximately 1.0 (γ≈1.0). For the ghost vehicle, assuming the spoofing vehicle maintains a constant distance from the autonomous vehicle 10 and the ghost vehicle is approaching the autonomous vehicle 10, the SNR distance ratio 52 is less than 1.0 and tends to zero because the real distance d T The distances are approximately equal across consecutive time steps; however, the measured distance d... M It decreases as the time step increases, or d T (t+1)≈d T (t) and d M (t+1)≤d M (t). Therefore, the SNR distance ratio of 52 indicates when the object is a ghost vehicle.

[0050] like Figure 1 As shown, the SNR distance ratio 52 is sent to the vehicle detection block 34 and the ghost vehicle tracking block 48. The ghost vehicle tracking block 48 uses a particle filter to determine the ghost position 54 of the object based on the SNR distance ratio 52, the object's position, and the object's velocity. Specifically, it should be understood that the SNR distance ratio 52 changes with time in a highly nonlinear manner. For moving object tracking, if the underlying system is linear in a Gaussian noise environment, a Kalman filter can be used. However, since the SNR distance ratio 52 is nonlinear, a particle filter is used to track ghost vehicles.

[0051] Figure 2 This is a block diagram showing the ghost vehicle tracking block 48, which includes a sampling block 70, a weighting block 72, a degradation check block 74, a resampling block 76, an estimation block 78, and a constraint ratio check block 80. See now. Figure 2 The sampling block 70 of the ghost vehicle tracking block 48 determines the state variable X based on the input probe point 50, previous values, and SNR distance ratio 52. t The importance of each variable in a subset is sampled. State variable X t This is represented in equation 3 as follows:

[0052] X t =[x(t), y(t), v x (t), v y (t), γ(t)] T Equation 3

[0053] Where x(t) and y(t) represent the x and y positions, respectively, and v x (t), v y (t) represents the x and y velocity coordinates, and γ(t) is the SNR distance ratio 52.

[0054] Observed variable Z t Represented as Z t =X t +u t , where u t It is an independent random noise process. For a given observed variable Z t X is the state variable at position x(t). t The importance sampling of each variable in part of the equation system 4 is represented as follows:

[0055] System of Equations 4

[0056] Where N(m, σ) represents a normally distributed sample with mean m and variance σ, and a x It is a constant used to control the variance of importance sampling and includes values ​​greater than 0. It is calculated by sampling block 70 as the state variable X. t (That is, x(t), y(t), v) x (t), v y Importance sampling of each variable in a part of (t), γ(t)) and expressed as follows in equations 5 to 8:

[0057] System of Equations 5

[0058] System of Equations 6

[0059] System of Equations 7

[0060] System of Equations 8

[0061] Among them, a y avx a vy a γ These are constants used to control the variance of importance sampling, and they are greater than zero. t is the time step, and t-1 is the time step immediately preceding t.

[0062] Weighted block 72 determines the value corresponding to X as a state variable based on equations 9 and 10. t The importance of each variable in a portion is sampled and weighted, as shown in Equations 9 and 10 below:

[0063]

[0064]

[0065] in, This represents the weight of the sampling point at time step t. It is a posterior distribution. It is a set of sampling points, and σ j This represents the variance that depends on the dynamic range of the corresponding component. Then, weighted block 72 normalizes the weights for each time step t based on Equation 11, which is as follows:

[0066]

[0067] Inspect degenerate block 74 to determine the effective particle number N eff The effective number of particles N eff Indicator as state variable X t The importance of each variable in the sample is determined by the degree of particle degradation. In one implementation, the effective number of particles N is determined based on Equation 12. eff Equation 12 is as follows:

[0068]

[0069] Where, N s This is the total number of particle points. Then, check if degenerate block 74 will have the effective particle count N. eff Compare with a predetermined threshold β. In one implementation, the predetermined threshold β is equal to half the total number of particle points, or N. s / 2, however, other values ​​can also be used. This is in response to determining the effective particle number N. eff If the values ​​are equal to or less than a predetermined threshold β, resampling block 76 re-executes the resampling operation to improve the particle distribution. The resampling operation distributes the concentrated weight values ​​across multiple samples. However, in response to determining the effective number of particles N... eff If the value is greater than the predetermined threshold β, resampling block 76 then combines the sampling points and the state variable X. t Send to estimation block 78.

[0070] Estimation block 78 is based on equation 13 and state variable X. t To estimate the ghost location 54 of the ghost vehicle, Equation 13 is as follows:

[0071] pos(x, y) = E[X t (1:2)] Equation 13

[0072] Where pos(x, y) represents the x and y coordinates of the ghost vehicle's ghost position 54. Once the ghost position 54 is determined, the constraint check block 80 checks the state variable X according to equation 14. t To estimate the SNR distance ratio 52, Equation 14 indicates the state variable X. t It includes five elements and determines the average or mean:

[0073]

[0074] in, This represents the estimated SNR distance ratio. In response to determining that the estimated SNR distance ratio is less than or equal to a predetermined value less than 1, estimation block 78 determines that ghost location 54 is a valid ghost point. In response to determining that the estimated SNR distance ratio is greater than or equal to a predetermined value less than 1, ghost location 54 is a non-ghost point. In one embodiment, the predetermined value less than 1 is 0.98. Once estimation block 78 has determined several consecutive non-ghost points, ghost vehicle tracking block 48 stops tracking non-ghost vehicles.

[0075] Figure 3 It shows through Figure 1 The diagram shows the process flow of the radar anti-spoofing system 32 for detecting and tracking ghost vehicles 200. See also: [Overall reference missing] Figures 1 to 3 Method 200 may begin at step 202. In step 202, the ghost vehicle detection block 46 determines the SNR range ratio 52 of the input detection point 50 based on multiple radar sensors 20, wherein the value of the SNR range ratio 52 indicates that the object is a ghost vehicle. As described above, the SNR range ratio 52 is determined based on Equation 2. Then, method 200 may proceed to step 204.

[0076] In step 204, the sampling block 70 of the ghost vehicle tracking block 48 determines the state variable X based on the input detection point 50 and the SNR distance ratio 52. t The importance of each variable in a subset is sampled and represented in equations 4 through 8. Method 200 can proceed to step 206.

[0077] In step 206, weighted block 72 is used as state variable X. tThe importance of each variable in a portion is sampled and weighted. Then, weighting block 72 normalizes the weights at each time step t based on equation 11. Then, method 200 can proceed to step 208.

[0078] In step 208, the degenerate block 74 is checked to determine the effective particle number N. eff The effective number of particles N eff Indicator as state variable X t The importance of each variable in a sample is determined by the degree of particle degradation. Then, method 200 can proceed to decision step 210.

[0079] In the determination step 210, in response to determining the effective number of particles N eff If the particle count is equal to or less than a predetermined threshold β, method 200 can then proceed to step 212. In step 212, resampling block 76 re-performs the resampling operation to improve the particle distribution. Then, method 200 can proceed to step 214. Returning to decision step 210, in response to determining the effective particle number N. eff If the value is greater than the predetermined threshold β, method 200 can then proceed to step 214.

[0080] In step 214, estimation block 78 uses Equation 13 as described above based on state variable X. t To estimate the ghost location of the ghost vehicle 54. Then, method 200 can proceed to step 216.

[0081] In step 216, the constraint ratio block 80 is checked using equation 14 based on the state variable X. t To estimate the SNR distance ratio 52, determine the estimated SNR distance ratio. Then, method 200 can proceed to decision step 218.

[0082] In determination step 218, the constraint ratio check block 80 compares the estimated SNR distance ratio with a predetermined value less than 1. In response to determining that the estimated SNR distance ratio is less than or equal to the predetermined value less than 1, the method proceeds to step 220, and estimation block 78 determines that ghost location 54 is a valid ghost point. In response to determining that the estimated SNR distance ratio is greater than or equal to the predetermined value less than 1, the method proceeds to step 222, and estimation block 78 determines that ghost location 54 is a non-ghost point. Method 200 can then terminate or return to step 202.

[0083] Referring generally to the accompanying drawings, the disclosed radar anti-spoofing system offers various technical effects and benefits for identifying and tracking ghost vehicles. Specifically, the radar anti-spoofing system provides an effective method for identifying and tracking ghost vehicles in nonlinear systems and non-Gaussian noise environments, enabling autonomous vehicles to react only to sensor data collected from the real vehicle. The radar anti-spoofing system identifies objects as ghost vehicles based on the SNR (Side-to-Noise Ratio) value. Furthermore, the disclosed radar anti-spoofing system includes particle filtering techniques to effectively track ghost vehicles.

[0084] Figure 4 This is an illustration of another embodiment of the radar anti-spoofing system 132 of the autonomous vehicle 110. In such... Figure 4 In the illustrated embodiment, the radar anti-spoofing system 132 includes one or more controllers 134 that communicate with multiple radar sensors 120. The one or more controllers 134 include an SNR offset module 140, a Doppler frequency offset module 142, a Kalman filter 144, a spoofing detection metric module 146, and a spoofed vehicle module 148. In such a manner... Figure 4 In the embodiment shown, the disclosed radar anti-spoofing system 132 identifies ghost vehicles based on Kalman filtering, which will be described in more detail below.

[0085] The SNR deviation module 140 determines the SNR deviation coefficient of an object in the environment based on input detection points 150 from multiple radar sensors 120. Specifically, the SNR deviation coefficient is calculated by determining the difference between the expected SNR and the measured SNR of the object. In one embodiment, the SNR deviation coefficient is determined based on Equation 15:

[0086] Dev SNR =1.0+|SNR Expected -SNR Measured Equation 15

[0087] Among them, Dev SNR It is the SNR deviation coefficient, SNR Expected It is the expected SNR, and SNR Measured This is the measured SNR. It should be understood that when the expected SNR and the measured SNR are equal, the additional constant 1.0 in Equation 15 is applied to preserve the original measurement noise covariance, which is what happens when the object is a real vehicle.

[0088] The Doppler frequency bias module 142 determines the Doppler bias coefficient of an object in the environment based on input detection points 150 from multiple radar sensors 120. Specifically, the Doppler bias coefficient is calculated by determining the difference between the expected Doppler frequency and the measured Doppler frequency of the object. In one embodiment, the Doppler bias coefficient is determined based on Equation 16:

[0089] Dev Doppler =1.0+|Doppler Expected -Doppler Measured Equation 16

[0090] Among them, Dev Doppler It is the Doppler bias coefficient. Expected It is the expected Doppler frequency, and Doppler Measured It is the measured Doppler frequency.

[0091] Both the SNR bias coefficient and the Doppler bias coefficient are sent to the Kalman filter 144. The Kalman filter 144 includes a prediction module 144A and an update module 144B, wherein the prediction module 144A receives the initial value and predicts the state before sending it to the update module 144B. The update module 144B determines four variables: the modified information covariance matrix S... k+1 Kalman gain K k+1 State estimate x k+1|k+1 and the updated state covariance matrix P k+1|k+1 When a standard Kalman filter is used, the new information covariance matrix S k+1 It is usually represented in Equation 17 as:

[0092]

[0093] Among them, H k+1 It is the observation matrix, R k+1 It is the measurement noise covariance matrix, and k represents the time step. For this disclosure, the innovation covariance matrix S k+1 It is achieved by measuring the noise covariance matrix R k+1 The modification is achieved by multiplying by the SNR deviation coefficient, and is expressed in Equation 18 as follows:

[0094]

[0095] It should be understood that when multiple radar sensors 120 are spoofed, the SNR bias coefficient increases over time. Consequently, the measurement noise covariance matrix R... k+1 The value of increases over time, and the Kalman gain K k+1 The value decreases over time. Therefore, the input detection points 150 generated by the ghost vehicle from multiple radar sensors 120 have a small impact on the state update over time. Furthermore, because the Kalman gain K... k+1 As it decreases over time, the updated state covariance matrix P k+1|k+1 The value also decreases to even smaller over time.

[0096] In one implementation, the new information covariance matrix S k+1 It is achieved by measuring the noise covariance matrix R k+1 Further modifications are made by multiplying by the SNR bias coefficient and the Doppler bias coefficient, and are expressed in Equation 19 as follows:

[0097]

[0098] The updated state covariance matrix P determined by Kalman filter 144 k+1|k+1 The deception detection metric is sent to the deception detection metric module 146. The deception detection metric module 146 determines the deception detection metric, which quantifies the updated state covariance matrix P determined by the Kalman filter 144. k+1|k+1 The relationship between the measured detection distance d between the autonomous vehicle 110 and the object. In one implementation, the deception detection metric is determined based on Equation 20, which is as follows:

[0099]

[0100] Where P represents the updated state covariance matrix, d represents the detection range of the object, and n represents the dimension of the state vector. It should be understood that, for a real vehicle, the updated state covariance matrix P determined by the Kalman filter 144... k+1|k+1 The magnitude of the deception detection metric is inversely proportional to the object's detection distance d. Therefore, a real vehicle will generate a nearly constant value for the deception detection metric over time. However, a ghost vehicle will exhibit relatively large fluctuations in the deception detection metric. Therefore, as detailed below, the deceitful vehicle module 148 monitors the value of the deception detection metric over time to determine whether the object is a real vehicle or a ghost vehicle.

[0101] See also Figure 4 The deceived vehicle module 148 determines the deception detection metric ratio based on the standard deviation of the deception detection metric determined within a defined time window divided by the mean of the deception detection metric determined within the defined time window. In one embodiment, the deceived vehicle module 148 compares the deception detection metric ratio with a threshold to determine whether the object is a ghost vehicle. Specifically, as shown in Equation 21 below, if the deception detection metric ratio is greater than the threshold th, the object is a ghost vehicle. The deception detection metric ratio is determined in Equation 21 as follows:

[0102]

[0103] in, It is the SDM vector between time m and m+d.

[0104] It should be understood that the threshold th is between 0 and 1 and is determined based on the relationship between detection probability and detection purity. Detection probability represents how many of the true probes are correctly classified as true, while purity represents the ratio of true probes classified as true to all probes classified as true. Typically, the ratio between detection probability and detection purity is a trade-off, where a lower threshold th results in higher detection purity, while a higher threshold th results in higher detection probability.

[0105] Figure 5 This is a flowchart illustrating a method 300 for identifying ghost vehicles using the disclosed radar anti-spoofing system 132. See also... Figure 4 and Figure 5 Method 300 may begin at step 302. In step 302, the SNR deviation module 140 determines the SNR deviation coefficient of an object in the environment based on input detection points 150 from multiple radar sensors 120. Specifically, the SNR deviation coefficient is calculated by determining the difference between the expected SNR and the measured SNR of the object, which is represented in Equation 15 above. Method 300 may then proceed to step 304.

[0106] In step 304, the Doppler frequency bias module 142 determines the Doppler bias coefficient of an object in the environment based on input detection points 150 from multiple radar sensors 120. Specifically, the Doppler bias coefficient is calculated by determining the difference between the expected Doppler frequency and the measured Doppler frequency of the object, as expressed in Equation 16 above. It should be understood that step 304 is optional and may be omitted in some embodiments. Method 300 can then proceed to step 306.

[0107] In step 306, the measurement noise covariance matrix R is... k+1 The innovation covariance matrix S of the update module 144B of the Kalman filter 144 is modified in combination with the SNR bias coefficient. k+1 In one implementation, the Doppler bias coefficient is also related to the new information covariance matrix S. k+1 Combined. Then, method 300 can proceed to step 308.

[0108] In step 308, the deceived vehicle module 148 determines a deception detection metric, which quantifies the updated state covariance matrix P determined by the Kalman filter 144. k+1|k+1 The relationship between the measured detection distance d between the autonomous vehicle 110 and the object. As mentioned above, when the object is a ghost vehicle, the value of the deception detection metric changes over time. Then, method 300 can proceed to step 310.

[0109] In step 310, the deceived vehicle module 148 determines a deception detection metric ratio based on the standard deviation of the deception detection metric determined within the defined time window divided by the mean of the deception detection metric determined within the defined time window, as expressed in equation 20 above. Then, method 300 can proceed to determination step 312.

[0110] In determination step 312, the deceived vehicle module 148 compares the deception detection metric ratio with a threshold. In response to determining that the deception detection metric ratio is greater than the threshold th, method 300 may proceed to step 314. In step 314, the deceived vehicle module 148 determines that the object is a ghost vehicle. In response to determining that the deception detection metric ratio is less than or equal to the threshold th, method 300 may proceed to step 316. In step 316, the deceived vehicle module 148 determines that the object is a real vehicle. Then, method 300 may terminate or return to step 302.

[0111] In a system-on-a-chip (SoC), a controller can refer to electronic circuitry, combinational logic circuitry, a field-programmable gate array (FPGA), a processor (shared processor, dedicated processor, or processor group) that executes code, or a combination of some or all of these, or a subset thereof. Furthermore, the controller can be based on a microprocessor, such as a computer having at least one processor, memory (RAM and / or ROM), and associated input and output buses. The processor can operate under the control of an operating system residing in memory. The operating system can manage computer resources so that computer program code embodied as one or more computer software applications (such as applications residing in memory) can have instructions that are executed by the processor. In an alternative implementation, the processor can directly execute the application, in which case the operating system can be omitted.

[0112] The descriptions in this disclosure are merely exemplary in nature, and any changes that do not depart from the spirit and scope of this disclosure are included within its scope. Such changes should not be considered as departing from the spirit and scope of this disclosure.

Claims

1. A radar anti-deception system for an autonomous vehicle, the radar anti-deception system comprising: a plurality of radar sensors producing a plurality of input detections representing radio frequency (RF) signals reflected from an object; and one or more controllers in electronic communication with the plurality of radar sensors, wherein the one or more controllers execute instructions to: determine a signal-to-noise ratio (SNR) distance ratio for the plurality of input detections produced by the plurality of radar sensors, wherein a value of the SNR distance ratio indicates that the object is a ghost vehicle; determine an importance sample for each variable that is part of a state variable based on the plurality of input detections and the SNR distance ratio; weight the importance sample for each variable that is part of the state variable; determine an effective number of particles that indicates a degree of particle degeneracy for the importance sample for each variable that is part of the state variable; and in response to determining that the effective number of particles is equal to or less than a predetermined threshold, estimate a ghost location of the ghost vehicle based on the state variable; wherein the SNR distance ratio is determined by: wherein, is an SNR constraint variable, is a constant derived from parameters of a particular radar sensor and is independent of range, denotes a true range, denotes a measured range, and is a difference between the true range and the measured range, and denotes a consecutive time step, and is an SNR range ratio.

2. The radar anti-deception system of claim 1, wherein the SNR distance ratio represents a difference in SNR constraint variables for the object over time, wherein the SNR constraint variables represent a physical constraint between a measured SNR and a distance of the object.

3. The radar anti-deception system of claim 1, wherein the state variable is expressed as: wherein is a state variable, denotes the x and y position, denotes the x and y velocity coordinates, and is the SNR distance ratio.

4. The radar anti-deception system of claim 1, wherein the importance sample for each variable that is part of the state variable is weighted by determining a value for: wherein denotes a time step a weight of a sample point at, is a posterior distribution, is a set of sample points, and denotes a variance depending on a dynamic range of the respective component.

5. The radar anti-deception system of claim 1, wherein the effective number of particles is determined by: wherein is the effective number of particles, is the total number of particle points, and denotes the time step is the weight of the sampling point at 6. The radar anti-deception system of claim 5, wherein the predetermined threshold is equal to half of the total number of particle points.

7. The radar anti-deception system of claim 1, wherein the one or more controllers execute instructions to: re-perform a resampling operation to improve a particle distribution in response to determining that the effective number of particles is equal to or less than the predetermined threshold.

8. The radar anti-deception system of claim 1, wherein the one or more controllers execute instructions to: estimate the SNR distance ratio based on the state variable, thereby determining an estimated SNR distance ratio.

9. The radar anti-deception system of claim 8, wherein the one or more controllers execute instructions to: compare the estimated SNR distance ratio to a predetermined value that is less than 1; and determine that the ghost location is a valid ghost point in response to determining that the estimated SNR distance ratio is less than or equal to the predetermined value that is less than 1.

Images