An adaptive detection method for GPS spoofing based on multi-UAV collaborative positioning

Abstract

The present invention discloses a GPS spoofing adaptive detection method based on multi-UAV collaborative positioning. The present invention uses a UAV cluster to broadcast the location information received by its own GPS receiver at the current moment to the local area of ​​the cluster at a specific frequency by each sub-machine. The cluster uses an AdHoc network to achieve communication between sub-machines through a flooding broadcast mode, so that each sub-machine in the cluster shares the broadcast information of all other sub-machines within the effective broadcast distance; when at least two aircraft broadcast the same location, it can be determined that there is at least one GPS spoofing source in the mission environment; trigger the optimal calibration mechanism of the relatively safe machine, lock the relatively safe sub-machine in the cluster, and then locate the sub-machine attacked by deception. The present invention has a light computational level, strong real-time performance, and breaks the conditional constraints of predecessors. It also has a strong application value under the premise of innovation.

Description

Technical Field

[0001] The present invention belongs to the field of unmanned aerial vehicle navigation information security, and in particular relates to a GPS spoofing adaptive detection method based on multi-unmanned aerial vehicle collaborative positioning. Background Art

[0002] Unmanned aerial vehicle (UAV) refers to an unmanned aircraft controlled by radio remote control equipment and self-contained program control devices, which can provide services in places that are difficult for humans to reach. In recent years, with the rapid improvement of related technologies such as sensing, remote sensing, flight control, gimbal, computational vision, and image transmission, the development of drones has also entered the fast lane. Especially since 2015, with the continuous improvement of civil drone technology, its application in agricultural and forestry plant protection, power inspection, geographic surveying and mapping, aerial photography, etc. has become more and more normalized, and the share of the drone industry in the civil drone market is increasing year by year. After 2019, drone autonomous control and application technology has made great progress and presented some new development trends. Since the mission payload that a single drone can carry is relatively single and its mission execution capability is limited, the complementary capabilities and coordinated actions of multiple drones can improve the efficiency of the entire system. Therefore, the application of drones is gradually developing from a single platform to a multi-platform direction.

[0003] Drone swarm refers to the use of the self-organizing mechanism of nature to enable multiple drones with limited autonomous capabilities to produce an overall effect through information communication without centralized command and control, and achieve a high degree of autonomous collaboration, so as to complete the expected mission objectives with minimal human intervention. As a disruptive technology and combat concept, drone swarm intelligence has become an important direction for future drone research and development.

[0004] The widespread use of drones in different fields, both individually and in groups, exposes them to too many security threats, such as network attacks, channel attacks, and signal attacks. Among these attacks, the most pressing threat is GPS spoofing, which is a signal attack. The control systems of modern drones are highly dependent on satellite navigation signals, from major telecommunications services to positioning and navigation systems, all rely on GPS signals. If the positioning and navigation information used by drones is spoofed, it is possible to interfere with the normal flight path of the drone, forcing it to deviate from the normal flight path without prior knowledge and appear in areas where it should not appear, thereby taking over the drone or directly causing the drone to crash. Therefore, with the rapid development of drones today, it is urgent to solve the security issues caused by GPS signals.

[0005] For the unit of detecting deceptive signals and ensuring mission safety of highly autonomous UAVs in GPS deceptive environments, most of the existing work is carried out in single-machine mission scenarios. AbdelRahman Eldosouky et al. proposed a model based on game countermeasures, and used the idea of ​​coordination in the first cluster mission, but the design constraints of the model were too ideal. Although it has certain guiding significance, it is difficult to implement and has little application value. Summary of the invention

[0006] The purpose of the present invention is to provide a GPS spoofing adaptive detection method based on multi-UAV collaborative positioning.

[0007] The purpose of the present invention is achieved through the following technical solutions:

[0008] Step 1: Each drone in the UAV cluster broadcasts the location information received by its own GPS receiver at the current moment to the local area of ​​the cluster at a specific frequency. The cluster uses the AdHoc network to achieve communication between drones through flooding broadcast mode, so that each drone in the cluster can share the broadcast information of all other drones within the effective broadcast distance;

[0009] Step 2: According to step 1, when the UAV cluster enters a stable mission state, the GPS spoofing signal detection mechanism based on cluster collaborative positioning is started;

[0010] Step 3: When all sub-machines attacked by GPS spoofing cannot be accurately located in step 2, the optimal calibration mechanism of relatively secure machines is triggered.

[0011] Furthermore, the specific steps of step 2 are as follows:

[0012] Step 2.1: When at least two aircraft broadcast the same position at the moment when each slave aircraft broadcasts to the local area of ​​the cluster, it can be determined that there is at least one GPS spoofing source in the mission environment;

[0013] Step 2.2: When each sub-machine broadcasts to the local area of ​​the cluster, there is no initial abnormality. A sub-machine X is randomly selected, and the remaining sub-machines calculate X′ based on their own position estimates according to the established geometric relationship between the machines in the formation design, and then broadcast the calculation results. By comparing the results, only one sub-machine in the cluster has been deceptively attacked, which can be accurately identified.

[0014] Further, according to the existence of at least one GPS spoofing source in the mission environment in step 2.1, the specific steps of step 3 are as follows:

[0015] Step 3.1: Submachine yaw independent model ψ

[0016] The yaw angle of each aircraft is provided by the GPS receiver and the magnetometer, which are expressed as and The yaw independent model is obtained by weighting the two, and the formula is:

[0017]

[0018] in is the weighted factor;

[0019] Get and

[0020] The position information (lonA, latA) and (lonB, atB) of the known broadcast GPS receiver at two times A and B; where the east and north directions are positive and the west and south directions are negative; The formula is:

[0021]

[0022] Known magnetometer measurements Get the projection of the magnetometer measurements on the horizontal plane: limited The formula is:

[0023]

[0024] Step 3.2: Virtual center machine yaw independent model Ψ

[0025] The measurement value of the magnetometer is easily affected by the physical environment, so the error coefficient (e1, e2, e3, e4, e5) of each aircraft at time k is obtained according to the output of the flight control board filter, and it is exponentially normalized by the softmax function to obtain ρ1+ρ2+ρ3+ρ4+ρ5=1, the formula is:

[0026]

[0027] Where ω is the weight coefficient;

[0028] Step 3.3: Relative safety machine optimal calibration model

[0029] Difference between the yaw result of each slave and the yaw result of the virtual center is D = |Ψ-ψ n |, we take D min The corresponding sub-machine is taken as the optimal estimate of the relatively secure machine.

[0030] The beneficial effects of the present invention are:

[0031] The present invention focuses on cluster collaboration as a whole. Facing a complex task confrontation environment, a GPS spoofing adaptive detection method based on multi-UAV collaborative positioning is designed. The present invention has a light computational level, strong real-time performance, and breaks the constraints of predecessors. It also has strong application value under the premise of innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of the GPS spoofing adaptive detection method based on multi-UAV collaborative positioning of the present invention;

[0033] Figure 2 This is a working diagram of the GPS spoofing signal detection mechanism based on cluster collaborative positioning of the present invention;

[0034] Figure 3 It is a flow chart of the optimal calibration mechanism of the relative safety machine of the present invention;

[0035] Figure 4 Schematic diagram of effective flooding communication between drones in a swarm of drones. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings.

[0037] The present invention is two sets of detection mechanisms with serial relationship: GPS spoofing signal detection mechanism based on cluster collaborative positioning and relative security machine optimal calibration mechanism. The specific technical solution includes the following steps:

[0038] Step 1: Each drone in the UAV cluster broadcasts the location information received by its own GPS receiver at the current moment to the local area of ​​the cluster at a specific frequency. The cluster uses an Ad Hoc network to achieve communication between drones through a flooding broadcast mode, so that each drone in the cluster can share the broadcast information of all other drones within the effective broadcast distance.

[0039] Step 2: As a priority, when the drone cluster enters a stable mission state, the GPS spoofing signal detection mechanism based on cluster collaborative positioning is enabled. The specific steps are as follows:

[0040] Step 2.1: When at least two aircraft broadcast the same position at the moment when each slave aircraft broadcasts to the local area of ​​the cluster, it can be determined that there is at least one GPS spoofing source in the mission environment;

[0041] Step 2.2: When each sub-machine broadcasts to the local area of ​​the cluster, if there is no abnormality initially, a sub-machine X is randomly selected, and then the remaining sub-machines calculate X′ based on their own position estimates according to the established geometric relationship between the machines in the formation design, and then broadcast the calculation results. By comparing the results, if only one sub-machine in the cluster is subjected to a deception attack, it can be accurately identified.

[0042] Step 3: When the GPS spoofing signal detection mechanism based on cluster collaborative positioning cannot accurately locate all sub-machines attacked by GPS spoofing, the optimal calibration mechanism of the relatively safe machine is triggered.

[0043] The core of this mechanism is: this mechanism does not directly realize the calibration of the sub-machines that are under deception attack, because there are more than one unsafe sub-machines at present, so it locks the relatively safe sub-machines in the cluster and then locates the sub-machines that are under deception attack.

[0044] The working principle of this mechanism is: according to the principle of GPS signal deception, when the drone is attacked by deception, the position and attitude of the drone itself have not changed. The deception signal only changes the drone's belief in its own position, and the impact on the drone's motion attitude occurs at the next moment of being deceived. At the moment of GPS failure, the relative controllability of attitude dynamics can be used to maximize the asymptotic stability of the closed-loop system in the application of the optimal control signal. Therefore, the Euler angle is used in this mechanism to estimate the flight attitude of the entire cluster, which can provide us with relatively reliable information when the GPS signal is deceived. Euler angle is an intuitive attitude representation method with clear physical meaning. In this mechanism, the deviation of the attitude of the sub-machine in the drone cluster relative to the flight attitude of the entire cluster is used to achieve the optimal estimation of the trusted machine. Because the main element involved in the final judgment model is the yaw angle, we simplified and divided the attitude model of the drone, and finally obtained an independent yaw linear model for each attitude.

[0045] The yaw representation of the flight attitude of the entire cluster is achieved by fusing the yaw attitudes of each member aircraft into a new yaw attitude model using the weighted average method. This can be regarded as a virtual central aircraft selected for the cluster. The new yaw attitude model is the yaw representation of the virtual central aircraft. Its physical meaning is that it can represent the flight attitude of the cluster to the greatest extent.

[0046] The specific steps of step 3 are as follows:

[0047] Step 3.1: Submachine yaw independent model ψ

[0048] For each aircraft, the yaw angle is provided by the GPS receiver and the magnetometer, which are expressed as and The yaw-independent model is obtained by weighting the two:

[0049]

[0050] in Time weighting factor. and Obtained respectively:

[0051] The position information (lonA, latA) and (lonB, latB) of the GPS receiver at two times A and B are known. The east and north directions are positive, and the west and south directions are negative. So:

[0052]

[0053] Known magnetometer measurements Its physical meaning is the measurement value of the magnetometer in the collective coordinate system. The research of this invention is based on the two-degree-of-freedom level of the drone (which can be expanded to four-degree-of-freedom and six-degree-of-freedom), so the pitch angle and roll angle are not considered. The projection of the magnetometer measurement value on the horizontal plane is obtained: limited So

[0054]

[0055] Step 3.2: Virtual center machine yaw independent model Ψ

[0056] (The UAV flight control board has built-in filters. Since it is not the focus of this invention, the design principle of the filter will not be described in detail here.) The measured value of the magnetometer is easily affected by the physical environment, so the error coefficient (e1, e2, e3, e4, e5) of each aircraft at time k is obtained according to the output of the flight control board filter, and it is exponentially normalized by the softmax function to obtain ρ1+ρ2+ρ3+ρ4+ρ5=1. Then:

[0057]

[0058] Where ω is the weight coefficient.

[0059] Step 3.3: Relative safety machine optimal calibration model

[0060] Difference between the yaw result of each slave and the yaw result of the virtual center is D = |Ψ-ψ n |, we take D min The corresponding sub-machine is taken as the optimal estimate of the relatively secure machine.

[0061] The present invention is to ensure that at least one aircraft in the cluster is absolutely safe in a harsh GPS signal confrontation environment. This setting is to ensure the recovery of the formation in the later stage and the final completion of the task. If it only stays at the detection level, this setting can be broken. min >d (d is the normal error range, which can be obtained through engineering measurement), the optimal calibration cannot be achieved, and manual takeover is performed.

[0062] according to Figure 4, consider a UAV swarm \(N\) performing a common task. Each UAV is equipped with a GPS receiver, a wireless communication module, and other sensors specific to the application. The UAV Ad Hoc network is a special scenario application of the mobile Ad Hoc network, which can enable the interconnection between terminals of small and medium-sized UAV swarms. In the UAV swarm model, this study adopts the flooding broadcast mode most commonly used in the Ad Hoc network to achieve communication between UAVs, that is, each UAV in the swarm shares its position information with each other within the effective distance of the broadcast.

[0063] Among them, \(e_{max}\) is the maximum effective range of UAV broadcast, and \(d_{max}\) is the farthest distance between UAVs in the swarm. When designing the swarm formation, let \(d_{max}<e_{max}\) to ensure that each UAV in the swarm can receive the position information broadcast by other UAVs.

[0064] The relative position between UAVs is not only the key basis for formation, but also when a UAV's navigation information fails, the position where the signal should be can be obtained through the real-time positions of other UAVs and the relative positions between UAVs. Therefore, when designing the model, the position of each UAV in the swarm relative to all other UAVs is known. When designing the formation, the object of this invention is mainly small UAV swarms, so the full-connection mode is adopted for information interaction between UAVs. For the model \(N = \{n\in N^+|u_1, u_2,\cdots, u_n\}\), the expression of the position relationship between any two UAVs is \(\theta\) and \(l\) respectively represent the angular and distance relationships between any two sub-UAVs in two degrees of freedom.

[0065] Relying on the Gagebo and Matlab simulation environments, we built a UAV swarm system model and the method model proposed by this invention, and finally verified that the method of this invention can achieve the expected effect.

[0066] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A GPS spoofing adaptive detection method based on multi-UAV collaborative positioning, characterized by: Here are the steps: Step 1: Each drone in the UAV cluster broadcasts the location information received by its own GPS receiver at the current moment to the local area of ​​the cluster at a specific frequency. The cluster uses the AdHoc network to achieve communication between drones through flooding broadcast mode, so that each drone in the cluster can share the broadcast information of all other drones within the effective broadcast distance; Step 2: According to step 1, when the UAV cluster enters a stable mission state, the GPS spoofing signal detection mechanism based on cluster collaborative positioning is started; Step 2.1: When at least two aircraft broadcast the same position at the moment when each slave aircraft broadcasts to the local area of ​​the cluster, it is determined that there is at least one GPS spoofing source in the mission environment; Step 2.2: When each slave machine broadcasts to the local area of ​​the cluster, there is no anomaly initially. A slave machine X is randomly selected. The remaining slave machines calculate X′ based on their own position estimation according to the established geometric relationship between the machines in the formation design, and then broadcast the calculation results. By comparing the results, only one slave machine in the cluster has been deceptively attacked, which can be accurately identified; Step 3: When all slaves attacked by GPS spoofing cannot be accurately located in step 2, the optimal calibration mechanism of relatively secure machines is triggered; Step 3.1: Submachine yaw independent model ψ The yaw angle of each aircraft is provided by the GPS receiver and the magnetometer, which are expressed as and The yaw independent model is obtained by weighting the two, and the formula is: in is the weighted factor; Get and The position information (lonA, latA) and (lonB, atB) of the known broadcast GPS receiver at two times A and B; where the east and north directions are positive and the west and south directions are negative; The formula is: Known magnetometer measurements Get the projection of the magnetometer measurements on the horizontal plane: limited The formula is: Step 3.2: Virtual center machine yaw independent model Ψ The measurement value of the magnetometer is easily affected by the physical environment, so the error coefficient (e1, e2, e3, e4, e5) of each aircraft at time k is obtained according to the output of the flight control board filter, and it is exponentially normalized by the softmax function to obtain ρ1+ρ2+ρ3+ρ4+ρ5=1, the formula is: Where ω is the weight coefficient; Step 3.3: Relative safety machine optimal calibration model Difference between the yaw result of each slave and the yaw result of the virtual center is D = |Ψ-ψ n |, we take D min The corresponding sub-machine is taken as the optimal estimate of the relatively secure machine.

Images