J. Electromagn. Eng. Sci Search

CLOSE


J. Electromagn. Eng. Sci > Volume 25(2); 2025 > Article
Chae, Park, Choi, and Choi: Multi-Rotor Redirection Algorithm Using GNSS Spoofer and Radar

Abstract

The rise in drone attacks on key state security infrastructures has led to a high demand for anti-drone technologies. In this regard, global navigation satellite system (GNSS) deception can mislead autopilot drones and redirect them to specific targets. However, existing redirection algorithms for multi-rotors in the Auto and return-to-launch (RTL) modes do not safeguard core facilities effectively. Therefore, this study proposes a practical multi-rotor redirection algorithm that can be applied to security systems. To redirect a multi-rotor to the designated target position in the Loiter, Auto, and RTL modes, a redirection algorithm using a GNSS spoofer and radar is designed that requires no estimation, unlike existing methods that estimate the target position or waypoint of a multi-rotor in the Auto and RTL modes. Redirection to the designated target position from all directions was confirmed through simulations. Furthermore, the multi-rotor redirection algorithm was verified by redirecting a DJI Phantom 4 to a designated target position.

I. Introduction

While drones are widely used in various applications in the private sector, they are also employed to carry out reconnaissance and attacks during wars and other terrorist activities [1, 2]. As a result, extensive research has been conducted on anti-drone technologies to address the illegal invasion of drones into facilities critical for state security [36]. Among these, global navigation satellite system (GNSS) signal deception is a commonly used method for addressing such vulnerabilities. Notably, autopilot drones rely heavily on GNSS information [79]. GNSS deception refers to spoofing the GNSS position and velocity of a drone. It can change a navigation solution by altering the navigation message [10] or code delay and Doppler shift [11]. Several studies have attempted to change navigation solutions using GNSS deception and have analyzed its effects [1215]. Additionally, studies have also been conducted to deter illegal multirotor intrusions using GNSS deception and then fly the drone to a safe area or a designated target position [1624]. In particular, several redirection experiments have been conducted in the Loiter mode, since the current position of a hovering multirotor is considered the target position in this mode [16, 1922]. In the case of Auto and RTL modes, redirection algorithms that estimate the multi-rotor’s target position or waypoint have been proposed [19, 2224]. However, the strategies proposed for these modes have largely been premised on the waypoint estimation of a drone, with no description provided for the estimation method and no verification conducted through field tests.
Additionally, as shown in Fig. 1, the redirection algorithms for the Loiter or Auto/RTL modes were separately suggested in [1924], and they were verified using open-loop or human-in-the-loop methods in a field test. In this context, multi-rotor redirection using a drone detection device can improve GNSS deception synchronization [16]. Moreover, since it can operate in a closed-loop structure, it results in relatively less redirection errors and enables a rapid response to drone invasion. Notably, a multi-rotor redirection algorithm that uses a drone detection device must be able to redirect the multi-rotor in the Auto and RTL modes without estimating the multi-rotor’s target position. Furthermore, an algorithm that enables redirection, regardless of the multi-rotor being in the Auto, RTL, and Loiter modes, can be depended upon to effectively respond to illegal intrusions.
The proposed multi-rotor redirection algorithm uses the radar and spoofer presented in [25]. However, while the redirection algorithm employed in [25] was developed for a commercial fixed-wing drone, this study proposes a new design for multi-rotor redirection. An algorithm capable of redirecting a multi-rotor in the Auto, RTL, and Loiter modes is simulated using a radar noise model, and its performance is verified. The developed redirection algorithm is used along with a real-time controllable GNSS spoofer and radar, with flight tests conducted to verify the results.
The contributions of this study are as follows:
  • • We propose a multi-rotor redirection algorithm that does not involve target position estimation of the multi-rotor in the Auto and RTL modes, which can be used along with a GNSS spoofer and radar.

  • • Simulations and flight tests confirm that the proposed redirection algorithm can redirect multi-rotors to a designated target position in any direction in the Auto, RTL, and Loiter modes.

  • • Simulation tests also confirm that the proposed algorithm exhibits similar redirection performance when the waypoint of the multi-rotor changes in the Auto mode, and it can also redirect multi-rotors that do not use a path-following algorithm.

The remainder of this paper is organized as follows: Section II describes the multi-rotor modelling of an Arducopter; Section III describes radar modelling, the GNSS spoofer, the design of the multi-rotor redirection algorithm, and the simulation results of the algorithm applied to the radar model; Section IV presents the redirection results obtained by incorporating a GNSS spoofer and a radar to the algorithm based on flight tests; and Section V presents a discussion of results. Finally, Section VI concludes the study.

II. Multi-Rotor Modelling based on Arducopter

For GNSS deception, the parts that require multi-rotor modelling based on each flight mode include the GNSS-IMU fusion and controller. Notably, the Auto/RTL and Loiter modes of the multi-rotor proposed in this study were modeled based on Arducopter [26, 27].
Fig. 2 illustrates the multi-rotor modeled for this study, comprising an extended Kalman filter (EKF), a position controller, an attitude controller, a multi-rotor state calculator, a multirotor state, a delay block, and a switch. The GNSS-IMU fusion was designed using the EKF [25, 28]. Notably, the EKF receives the multi-rotor acceleration uk and the switch outputs the current position Dks and current velocity Vks from the EKF to the position controller. Here, the subscript k indicates the k-th time step. When modelling the multi-rotor, the position and attitude controllers were simplified, while only the horizontal axis was accounted for in the north-east-down (NED) coordinate system. Meanwhile, the position controller computes the target acceleration uk+1t so that the multi-rotor can be positioned at the target position Dkt. The target velocity Vkt and the position error ekpos are the output and input of the square root (SQRT) controller, respectively. Furthermore, the value obtained by the velocity error ekvel, which refers to the difference between Vkt and Vks, when passing through the proportional-integral-derivative (PID) controller and filter is uk+1t. The attitude controller calculates the target pitch angle rate and target roll angle rate for multi-rotor acceleration uk+1 to be uk+1t. The multi-rotor state calculator, used for simulation, computes the next state of the multi-rotor ξk+1. The state of a multi-rotor ξk is determined by its actual current position Dk and actual current velocity Vk. Moreover, uk+1 of the multirotor can be calculated using the target pitch angle rate, the target roll angle rate in the attitude controller, Vk, and the linear drag coefficients—Ax and Ay [29]. Furthermore, ξk+1 can be obtained using ξk, uk+1, system matrix A, and input matrix B, as described in Eq. (1) below:
(1)
ξk+1=Aξk+Buk+1=[Dk+1,Vk+1]T.
Since the multi-rotor state stores the actual multi-rotor state, the actual position, velocity, acceleration, and attitude of the multi-rotor can be monitored when GNSS deception is applied. Furthermore, a switch is employed to interface with the GNSS deception during the simulation. The multi-rotor state produces acceleration as an output (via the delay block) to estimate the next state to the EKF and the position and velocity to the switch. Meanwhile, the switch selects the position and velocity, which are the inputs to the EKF. Therefore, if the GNSS deception position Dksp and velocity Vksp are the inputs, the switch selects them; otherwise, it selects the ξk+1 from the multi-rotor state.

1. Loiter Mode Modelling

In Loiter mode, a multi-rotor automatically holds its position and altitude at its current location using GNSS. The multirotor hovers at its current position, which signifies the next target position Dk+1t. Therefore, to calculate Dk+1t, the maximum absolute value of ekpos should be limited by the track leash length Lt, which was set to 20 m in this study. Accordingly, Dk+1t was determined by implementing Eq. (2), using Dks and ekpos as follows:
(2)
Dk+1t=Dks+ekpos.

2. Auto & RTL Mode Modelling

The Auto and RTL modes employ the same path-following algorithm to fly, with the multi-rotor covering the path line connecting WP1 and WP2, or the current and take-off positions, as shown in Fig. 3. Since these two modes are almost identical, they were grouped together to simplify the modelling process. The cross-track error (CTX) in Fig. 3 indicates the vertical distance between Dks and the path line. The path-following algorithm determines Dk+1t by calculating the desired track based on the CTX. Notably, Dk+1t is the same as Virtual target point (VTP), which is calculated to advance or stop at a specific speed according to Dks and Vks. The path-following algorithm sets the VTP at the intersection of the path line and the circle with a radius given by Lt at Dks. This intersection represents the maximum acceptable VTP from WP1 on the path line, where the VTP can be set. If there is no intersection due to the CTX being longer than Lt, the maximum acceptable VTP is set as the orthogonal point between the path line and Dks. Notably, VTP change conditions have been analyzed in [19].

III. Proposed Multi-Rotor Redirection Algorithm using GNSS Spoofer and Radar

In this study, a multi-rotor redirection algorithm is designed using a GNSS spoofer and radar, as shown in Fig. 4. A radar was employed to detect the position and speed of the multirotor, and a GNSS spoofer was utilized to generate a GNSS deception signal by automatically calculating the GNSS deception position and velocity based on the radar measurements.
Aperiodic radar data were changed into periodic data using a Kalman filter, enabling the redirection algorithm to calculate the GNSS deception position and velocity, even in the case of radar tracking loss. On calculating the GNSS deception position and velocity, the multi-rotor redirection algorithm redirects the multi-rotor to a designated target position based on the Kalman filtered data. Subsequently, a deception signal is generated based on the calculated deception position and velocity. Notably, deception signal generation, the Kalman filter for the GNSS spoofer, and the radar were used with regard to [25].

1. Radar Modelling

A radar was included in the proposed algorithm owing to its relatively long detection range. For the radar, we employed FIELDctrl by Advanced Protection Systems (APS) [30], which can measure both the three-dimensional position and the speed of multi-rotors. Radar modeling was performed to simulate the redirection algorithm based on the radar error measurement, which refers to the difference between the radar measurement data and the GPS log of DJI Phantom 4. Although APS’s radar does not provide internal signal processing information, it offers the predicted detection quality (PDQ) feature to check the ground clutter environment before operation. In this study, radar measurements were performed within 500 m of ground distance, representing an area that offers high PDQ. Fig. 5 shows the radar measurement errors in range, azimuth, elevation, and speed relative to the ground distance.
Unlike the other errors, the elevation error was observed to be inversely proportional to the ground distance. Since the radar used in this study was installed at a height of approximately 1.5 m from the ground, it was affected by multipath ground reflections. Therefore, the elevation error calculated the bias and bias weight, which is the slope of the error relative to the ground distance, through linear regression. Additionally, when the flight direction of the drone changed rapidly, all errors exhibited a random slope change owing to increased radar prediction error. However, it was confirmed that the variance between the specific sample windows was not too large. Therefore, after calculating the standard deviation of the errors, the random slope bias error was modeled to be in the range of approximately 1.5 sigma of the standard deviation, based on the average. The slope was set to change randomly every 10 seconds within the range resolution. Furthermore, the average of the errors was considered the mean of the Gaussian distribution. Additionally, after obtaining the standard deviation of the measurement error for a specific sample window, it was considered the standard deviation of the Gaussian distribution. As listed in Table 1, the range, azimuth, elevation, and speed errors of the radar were modeled by adding noise using Gaussian distribution, random slope bias, and linear regression. Fig. 6 depicts the modeled radar noise obtained after 10 simulations. Evidently, the error generated by the radar model partially included the range of the measured error.

2. Real-Time Controllable GNSS Spoofer

A GNSS spoofer is responsible for generating a deceptive position and velocity using time delay, which changes the pseudo range based on the position and velocity of the target to obtain a deceptive navigation solution. The GNSS spoofer used in this study was fabricated using a GNSS receiving antenna, a navigation message extractor, a time synchronization receiver, an RF front end, an field programmable gate array (FPGA) and digital signal processor (DSP), and a GNSS transmitting antenna. The DSP and FPGA were synchronized using the 1 pulse per second and reference clock generated by the synchronization receiver. A DSP calculates the time delay to be generated by the spoofer, as well as the Doppler frequency, using the deception information obtained from the redirection algorithm and the extracted navigation message. Meanwhile, the FPGA generates a deception signal by applying the time delay and Doppler frequency to each satellite. Upon combining the deception signals of each satellite, the signal can be normalized to prevent saturation in the DAC. Subsequently, the frequency of the deception signal can be upconverted in the RF front end, after which the signal is radiated through the antenna.

3. Design of the Multi-Rotor Redirection Algorithm

This study aimed to design a redirection algorithm that can redirect multi-rotors in both Auto/RTL and Loiter modes without having to estimate their target position in the different modes. The redirection concepts adopted in this study achieved this objective by initiating multi-rotor redirection in the Loiter mode and subsequently extending it to the Auto/RTL modes. Notably, the design concepts explained below are based on the multi-rotor model discussed in Section II.

3.1 Concept for Loiter mode multi-rotor redirection

In Loiter mode, the hovering position is considered the next target position. As shown in Fig. 7, a deceptive position Dksp and velocity Vksp are generated in the direction opposite to the initial target direction ∠Tinit, indicated by the blue dotted line connecting the positions of the multi-rotor at Dk and DDT, as described in Eq. (3), to redirect it toward the designated target position DDT.
(3)
Tinit=(DDT-Dk).
Here, Dk+1 and Vk+1 indicate the actual position and velocity of the multi-rotor in the next step, respectively.
For clearer understanding, it was assumed that when Dksp and Vksp are fed as inputs, Dks and Vks in Fig. 2 becomes Dksp and Vksp, respectively. Hence, the calculated direction of uk+1 obtained using Dksp and Vksp is the same as ∠Tinit. The conceptual diagram in Fig. 7 depicts the direction of uk+1.
Additionally, assuming that the magnitude of acceleration is too large to enable easy understanding, it was established that the direction of the velocity of the multi-rotor in the next step is the same as that of the acceleration in the next step, as shown in the conceptual diagram. However, in reality, radar errors exist in Dksp and Vksp because the deception position and velocity are calculated based on radar measurements. Additionally, based on the characteristics of the GNSS receiver and the sensor fusion of the multi-rotor, Dks and Vks of the multi-rotor may not be the same as Dksp and Vksp when GNSS deception is induced. Therefore, even when the multi-rotor flies approximately along the initial target direction, it is still susceptible to errors.

3.2 Concept for Auto/RTL mode multi-rotor redirection

In the Auto/RTL mode, a redirection concept similar to the Loiter mode was initially considered using the initial target direction ∠Tinit. The direction opposite to ∠Tinit was used as a reference for calculating the deceptive position and velocity. To establish a redirection concept for the Auto/RTL mode, the designated target position settings at the front and back of the multi-rotor position were divided based on a path line. Fig. 8 depicts this control concept using the Arducopter discussed in Section II, which controls the attitude of uk+1 based on Dks and Vks to fly from Dks to Dkt. When setting the designated target position DDT at the front of the multi-rotor based on the path line, as shown in Fig. 8(a), ∠Tinit was used to calculate the initial direction of Vksp. In this context, the target direction ∠Tk+1, which was calculated for each sample step, represents the direction connecting the current position of the multi-rotor measured by the radar and DDT. It is also used to calculate the direction error ∠Ek+1 between ∠Tk+1 and the flight direction ∠Vk+1 of the multi-rotor, measured using radar. Notably, ∠Tk+1 and ∠Ek+1 can be expressed as Eqs. (4) and (5), respectively, where Dk+1 represents the position of the multirotor measured using radar:
(4)
Tk+1=(DDT-Dk+1),
(5)
Ek+1=Tk+1-Vk+1.
The top left image in Fig. 8(a) indicates that the speed along the path line becomes negative because of Vksp. Therefore, the VTP is kept fixed, as demonstrated in [19]. In contrast to the Loiter mode, the VTP is located ahead of Dks in the Auto/RTL mode, as a result of which the multi-rotor flies with a slightly larger error from ∠Tk+1 in the next step, owing to Dksp and Vksp.
For clearer understanding, the direction of uk+1 is shown to be the same as the direction of ekvel, which signifies the difference between Vkt and Vksp of the position controller. Furthermore, the direction of Vkt is considered the same as the direction of the vector connecting Dksp and the VTP, while the magnitude of Vkt is approximately equal to the square root of twice the magnitude of the vector, as in [27]. The image at the bottom of Fig. 8(a) highlights that the direction of uk+1 can be obtained intuitively. It is observed that if the deception velocity is changed in the counterclockwise direction, based on its direction in the previous step, the acceleration of the multi-rotor will also change to move in the counterclockwise direction, thus reducing ∠Ek+1. Moreover, it is evident that the same case can be realized in the clockwise direction according to the same principle.
Furthermore, if DDT is set to be located behind the multirotor based on the path line, and the initial deception direction is calculated based on the direction opposite to ∠Tinit, the flight direction of the multi-rotor will change based on the deception speed. If the deception speed is lower than or equal to the VTP speed of the multi-rotor, the VTP will always be located ahead of the deception position along the path line or perpendicular to the path line, as in [19, 27]. Moreover, the deception position and acceleration direction of the multi-rotor resulting from the deception will be in the front direction or in the limited backward direction of the multi-rotor position based on the path line. As a result, it cannot be redirected to a designated target position located in the direction beyond a certain backward angle of the current position of the multi-rotor based on a path line. Furthermore, if the deception speed is faster than the VTP speed, the deception position will be located ahead of the VTP based on the path line after a few time steps. Therefore, as shown in Fig. 8(b), the multi-rotor can be redirected toward a designated target position located behind the initial position of the multi-rotor Dk based on the path line. For a clearer understanding, the direction of uk+1 in Fig. 8(b) follows the trajectory depicted in Fig. 8(a). To reduce ∠Ek+1, if the deception velocity direction is changed to the counterclockwise direction based on its direction in the previous step, the acceleration direction of the multi-rotor also changes to move in the counterclockwise direction. In this way, the multi-rotor can be redirected to the designated target position. Moreover, the clockwise case follows the same principle.
The maximum VTP speed was set to 1.25 times faster than the normal multi-rotor flight speed. Furthermore, since radar can measure multi-rotor speed, the deception speed should be set based on it. Overall, if the deception speed is set appropriately, regardless of whether the designated target position is at the front or the back of the multi-rotor position based on the path line, the multi-rotor acceleration direction changes based on the directional change of the deception velocity.

3.3 Proposed multi-rotor redirection algorithm

The multi-rotor redirection algorithm for the Auto/RTL and Loiter modes was designed based on the abovementioned redirection concepts. Fig. 9 depicts the flowchart of the multi-rotor redirection algorithm. The procedure followed by the algorithm for redirection in the Loiter and Auto/RTL modes is explained below:
  • 1. The target direction ∠Tk for each sample step is calculated using the multi-rotor’s position, measured using the radar and a user-defined designated target position.

  • 2. Before the onset of GNSS deception, the initial target direction ∠Tinit, which refers to the calculated target direction in the current time step, is stored in memory. Furthermore, the maximum deception speed is set to be greater than the measured multi-rotor speed.

  • 3. Once GNSS deception begins, the deception position and velocity are calculated based on the reference direction using Eq. (6) for each sample step to ensure that the direction error ∠Ek+1 reaches zero.

(6)
Vk+1sp=PID(Ek+1)+(Tinit-180°).
PID refers to the PID controller. A direction error ∠Ek+1 usually occurs when the flight direction of a multi-rotor is controlled by GNSS deception. Accordingly, a redirection algorithm was designed to calculate the position and velocity of GNSS deception using only the position and speed of the multi-rotor, as measured by radar, and the designated target position set by the user.
Notably, since the modeled multi-rotor was designed based on Arducopter, the VTP setting method might differ for other types of multi-rotors. Therefore, after designing the redirection algorithm, the maximum deception speed and PID gain must be adjusted to calculate the velocity of the deception.

4. Multi-Rotor Redirection Algorithm Simulation

For simulation, the multi-rotor was modeled according to the configuration presented in Fig. 2, while the designed multi-rotor redirection algorithm was modeled based on the configuration shown in Fig. 4, including the multi-rotor model and the radar model but excluding the GNSS deception signal generation. All simulations were performed using MATLAB Simulink.

4.1 Redirection of multi-rotors using a path-following algorithm

Fig. 10 shows the possible redirection range and average minimum target distance obtained by simulating the multi-rotor redirection algorithm using the radar model, with the multirotor model set to use a path-following algorithm in the RTL mode. Here, the target distance refers to the distance between the designated target position and the multi-rotor position measured by the radar after Kalman filtering. The designated target position was set between −500 to 500 m at intervals of 50 m along the north and east axes. The average minimum target distance was obtained by performing 10 iterations of the simulation for each designated target position. The squares in Fig. 10 signify the designated target positions, with the average minimum target distance being 50 m or less, while the distribution of the squares denote possible redirection range.
Fig. 10(a) depicts the possible redirection range and average minimum target distance of the multi-rotor in Loiter mode. The multi-rotor flight speed, maximum flight speed, and maximum deception speed were set to 4 m/s, 5 m/s, and 1.5 times the measured flight speed, respectively. The simulation conducted for the Loiter mode confirmed successful redirection for all designated target positions while maintaining an average minimum target distance of 24 m or less. Fig. 10(b) depicts the possible redirection range and average minimum target distance of the multi-rotor in the RTL mode. Here, WP1 was set at the origin, and WP2 was set 1,000 m east of the origin. The simulation results confirm that redirection was achieved for all designated target positions, with the average minimum target distance being 41 m or less.
In an actual scenario, the waypoint or take-off position of a target multi-rotor in Auto/RTL mode is usually unknown. Accordingly, the traveling direction of VTP will change based on the difference in the path line directions of WP2 and WP3 from those of WP1 and WP2. In such a case, if the VTP advances owing to the GNSS deception position and velocity, the acceleration direction of the multi-rotor will change accordingly. Therefore, as shown in Fig. 11, a simulation was performed to examine whether the possible redirection range would be the same as the range depicted in Fig. 10(b) when WP2 is set close to the multi-rotor.
For this analysis, WP2 was set 250 m east of the origin, while WP3 was set 1,000 m north of WP2, south of WP2, and west of the origin, indicating three different cases. The simulation results in Fig. 10(b) show that the possible redirection ranges of all three cases are the same. In other words, when setting WP3 as the next waypoint for the multi-rotor, the VTP either remained fixed or showed little advance due to the GNSS deception position and velocity. In short, the waypoints of the multirotor have a minimal effect on the possible redirection range of the proposed redirection algorithm.
Therefore, we concluded that both modes can be redirected in all directions using the proposed algorithm. In practice, the designated target position must be set based on the detection range of the radar and the effective range of the GNSS spoofer.

4.2 Redirection of multi-rotors without a path-following algorithm

Additional simulations were conducted to verify the performance of the redirection algorithm when the multi-rotor model did not employ a path-following algorithm. As shown in Fig. 12, the redirection algorithm was able to redirect the multirotor when the designated target position was set in front of its flight direction. Notably, with the target position located farther away, the direction of the acceleration caused by the deception was limited. This implies that the farther the location of WP2, the smaller the redirection range.

IV. Experiments

Flight tests were performed using DJI Phantom 4 Pro V2.0 [31] to confirm the performance of the multi-rotor redirection algorithm using a GNSS spoofer and radar in both Hover and RTH modes.
Fig. 13 shows the test configuration for the proposed multirotor redirection algorithm. The blue, orange, and red solid lines represent the radar measurements, authentic GNSS signals, and deceptive GNSS signals, respectively. The radar measurements were transmitted to the GNSS spoofer using a data link. Additionally, the direction of the transmit antenna of the GNSS spoofer was set based on radar measurements to continuously track the direction of the multi-rotor.
Fig. 14 presents the redirection test results for the multi-rotor redirection algorithm in the Hover and RTH modes. As shown in Fig. 14(a) and 14(b), the system was located at the origin in the NED coordinate system, with the units of the coordinates expressed in meters. During the test, the flight altitude was set to 50 m absolute ground level (AGL) or more, while the maximum deception speed was set to 1.5 times the measured flight speed before deception was initiated. The flight tests in the Hover and RTH modes were performed by setting the designated target position at approximately 400 m and 320 m from the position of the Phantom 4, respectively. The red and blue lines trace the deception trajectory and flight trajectory of Phantom 4, respectively. In the Hover mode, Phantom 4 briefly flew south to then change its direction toward the north after the initiation of deception, as shown in Fig. 14(a), since the deception position was set north of the drone’s position owing to the radar’s position measurement error. The redirection algorithm continuously changed the deception direction based on the direction error, compelling the drone to fly toward the designated target, and accordingly, the drone flew near the designated target position. In the RTH mode, on initiating deception, Phantom 4 flew in the direction of the designated target position but with a small direction error, as shown in Fig. 14(b). Thereafter, the redirection algorithm maintained the deception direction to change the flight direction toward the northeast. Once the drone passed approximately 220 m toward the north, its flight direction changed slightly to the northeast. Accordingly, the deception direction was changed to slightly alter the flight direction toward the northwest. Consequently, the drone flew near the designated target position.
Fig. 15 illustrates the results obtained on changing the multirotor or the designated target positions for the tests in the Hover (4 times) and RTH modes (5 times). Five tests were conducted, the results of which were obtained within 10 m. The results attained for the Hover and RTH modes were approximately 20 m and 35 m, respectively. This value could be improved by optimizing the redirection algorithm through more field tests. However, the number of tests that could be conducted was limited for safety reasons. Nonetheless, it was confirmed that redirection can be achieved in both the Auto/RTL and Loiter modes using the proposed method.

V. Discussion

The proposed multi-rotor redirection algorithm using a GNSS spoofer and radar was verified by conducting simulations and field tests. Additionally, the multi-rotor redirection performance was confirmed through simulation using Cases 1 and 3 of the redirection algorithm proposed in [25], as depicted in Fig. 16, which is similar in concept to the algorithm proposed in this study. The maximum deception speed was set to 1.5 times the measured flight speed before deception was initiated, thus applying the same concept as the proposed algorithm. The squares in Fig. 16 represent the designated target positions, with the average minimum target distance being 50 m or less, while the distribution of the squares denote a possible redirection range. The results show a limited redirection range compared to that presented in Fig. 10. Notably, since the redirection algorithm in [25] was designed for fixed-wing drones, the direction of the path line had to be estimated based on the direction of the drone to achieve redirection. Therefore, Cases 1 and 3 were performed based on the estimated direction of the path line. Accordingly, the baselines for Cases 1 and 3 were set without accounting for the designated target position. However, estimation of the path line in Loiter mode exhibited certain limitations in terms of accuracy.
In contrast, the proposed redirection algorithm was designed by accounting for the flight characteristics of a multi-rotor. It sets the initial target direction based on the drone’s position and the designated target position before redirection, and later performs deception based on this data. As a result, the baseline of deception depends on the designated target position in this case. The proposed redirection algorithm was designed by modifying the deception baseline and setting the maximum deception speed of the redirection algorithm proposed in [25], which resulted in a difference in multi-rotor redirection performance between the two.

Conclusion

This study proposes a multi-rotor redirection algorithm that uses a GNSS spoofer and radar for redirecting a multi-rotor toward a designated target position in the Auto/RTL and Loiter modes. Through simulation tests, a redirection performance similar to the designated target was confirmed, regardless of the waypoint location set by the multi-rotor pilot. Furthermore, the proposed algorithm performed multi-rotor redirection without using a path-following algorithm. In other words, the proposed redirection algorithm demonstrated that multi-rotor redirection can be achieved without using automatic flight mode and path-following algorithms.
However, since the number of tests that could be conducted was limited due to a safety-conscious test schedule, only the DJI Phantom 4 was used for testing. Further research is required to determine the applicability of the proposed redirection algorithm for other drones that employ multiple GNSS constellations and other EKF failsafe methods. Furthermore, a GNSS spoofer that can deceive all GNSS constellations used by the latest drones should be developed.
In the proposed algorithm, redirection was achieved even when the multi-rotor pilot changed from the Auto/RTL mode to the Loiter mode, or when it changed waypoints during GNSS deception. Moreover, through redirection in a closed-loop structure using radar, the redirection was performed quickly and accurately compared to open-loop or human-in-the-loop methods. As a result, the proposed algorithm is expected to be applied and utilized in security systems to prevent illegal drone intrusions.

Fig. 1
Comparison of previous and proposed redirection algorithms.
jees-2025-2-r-284f1.jpg
Fig. 2
Multi-rotor model based on Arducopter.
jees-2025-2-r-284f2.jpg
Fig. 3
Path-following algorithm in the Auto/RTL mode.
jees-2025-2-r-284f3.jpg
Fig. 4
Multi-rotor redirection algorithm using the GNSS spoofer and radar configuration.
jees-2025-2-r-284f4.jpg
Fig. 5
Radar measurement error using DJI Phantom 4: (a) range error, (b) azimuth error, (c) elevation error, and (d) speed error.
jees-2025-2-r-284f5.jpg
Fig. 6
Simulation results of the radar model: (a) range error, (b) azimuth error, (c) elevation error, and (d) speed error.
jees-2025-2-r-284f6.jpg
Fig. 7
Conceptual diagram of multi-rotor redirection in the Loiter mode.
jees-2025-2-r-284f7.jpg
Fig. 8
Conceptual diagram for multi-rotor redirection in the Auto/RTL mode: (a) redirection to the forward designated target position and (b) redirection to the backward designated target position.
jees-2025-2-r-284f8.jpg
Fig. 9
Flowchart of the multi-rotor redirection algorithm.
jees-2025-2-r-284f9.jpg
Fig. 10
Simulation results of the multi-rotor redirection algorithm: possible redirection range and average minimum target distance in Loiter mode (a) and RTL mode (b).
jees-2025-2-r-284f10.jpg
Fig. 11
Simulation results of the multi-rotor redirection algorithm: redirection performance is achieved when WP2 is located close to WP1 and the position of WP3 changes. Three cases of WP3 are considered: 1,000 m north of WP2 (small circle), south of WP2 (medium circle), and west of origin (large square).
jees-2025-2-r-284f11.jpg
Fig. 12
Simulation results of the redirection algorithm for the multirotor without using the path-following algorithm. Redirection performance achieved when WP2 is located close to or far from WP1: (a) WP2: [0,1000] and (b) WP2: [0,3000].
jees-2025-2-r-284f12.jpg
Fig. 13
Test configuration of the multi-rotor redirection algorithm using a GNSS spoofer and radar.
jees-2025-2-r-284f13.jpg
Fig. 14
Redirection test results of the multi-rotor redirection algorithm using a GNSS spoofer and radar for DJI Phantom 4 in Hover mode (a) and RTH mode (b).
jees-2025-2-r-284f14.jpg
Fig. 15
Redirection test results of DJI Phantom 4 in the Hover and RTH modes.
jees-2025-2-r-284f15.jpg
Fig. 16
Simulation results for Cases 1 and 3 of the redirection algorithm in the study of Chae et al. [25]: possible redirection range and average minimum target distance in Loiter mode (a) and RTL mode (b).
jees-2025-2-r-284f16.jpg
Table 1
Parameters of the radar noise model
Noise type Range error (m) Azimuth error (°) Elevation error (°) Speed error (m/s)
Gaussian distribution
 Mean −0.6 0.57 2.13 0
 Standard deviation 0.5 0.1 0.1 0.1
Linear regression
 Bias weight/bias −0.014/6.5
Random slope bias
 Range −10 to 10 −3 to 3 −3.1 to 3.1 −2 to 2
 Range resolution 5 1.5 1.55 1
 Changing time (s) 10 10 10 10

References

1. S. O. Ajakwe, D. S. Kim, and J. M. Lee, "Drone transportation system: systematic review of security dynamics for smart mobility," IEEE Internet of Things Journal, vol. 10, no. 16, pp. 14462–14482, 2023. https://doi.org/10.1109/JIOT.2023.3266843
crossref
2. M. R. DeVore, "“No end of a lesson:” observations from the first high-intensity drone war," Defense & Security Analysis, vol. 39, no. 2, pp. 263–266, 2023. https://doi.org/10.1080/14751798.2023.2178571
crossref
3. C. Lyu and R. Zhan, "Global analysis of active defense technologies for unmanned aerial vehicle," IEEE Aerospace and Electronic Systems Magazine, vol. 37, no. 1, pp. 6–31, 2022. https://doi.org/10.1109/MAES.2021.3115205
crossref
4. H. Kang, J. Joung, J. Kim, J. Kang, and Y. S. Cho, "Protect your sky: a survey of counter unmanned aerial vehicle systems," IEEE Access, vol. 8, pp. 168671–168710, 2020. https://doi.org/10.1109/ACCESS.2020.3023473
crossref
5. D. Schmidt, K. Radke, S. Camtepe, E. Foo, and M. Ren, "A survey and analysis of the GNSS spoofing threat and countermeasures," ACM Computing Surveys (CSUR), vol. 48, no. 4, article no. 64, 2016. https://doi.org/10.1145/2897166
crossref
6. D. He, G. Yang, H. Li, S. Chan, Y. Cheng, and N. Guizani, "An effective countermeasure against UAV swarm attack," IEEE Network, vol. 35, no. 1, pp. 380–385, 2021. https://doi.org/10.1109/MNET.011.2000380
crossref
7. Z. Renyu, S. C. Kiat, W. Kai, and Z. Heng, "Spoofing attack of drone," In: Proceedings of 2018 IEEE 4th International Conference on Computer and Communications (ICCC); Chengdu, China. 2018, pp 1239–1246. https://doi.org/10.1109/CompComm.2018.8780865
crossref
8. S. Daniel, B. A. Jashan, and H. E. Todd, Drone hack: deception attack demonstration on a civilian unmanned aerial vehicle, 2012. [Online]. Available: https://www.gpsworld.com/drone-hack/

9. S. P. Arteaga, L. A. M. Hernandez, G. S. Perez, A. L. S. Orozco, and L. J. G. Villalba, "Analysis of the GPS spoofing vulnerability in the drone 3DR solo," IEEE Access, vol. 7, pp. 51782–51789, 2019. https://doi.org/10.1109/ACCESS.2019.2911526
crossref
10. Q. Yang, Y. Zhang, and C. Tang, "A novel GPS spoofing algorithm based on modifying navigation message," In: Communications and Networking; . Cham, Switzerland: Springer. 2018, pp 46–53. https://doi.org/10.1007/978-3-319-78130-3_6
crossref pmid
11. T. E. Humphreys, B. M. Ledvina, M. L. Psiaki, B. W. O’Hanlon, and P. M. Kintner, "Assessing the spoofing threat: Development of a portable GPS civilian spoofer," In: Proceedings of the 21st International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS); Savannah, GA, USA. 2008, pp 2314–2325.

12. J. Bhatti and T. E. Humphreys, "Hostile control of ships via false GPS signals: demonstration and detection," NAVIGATION: Journal of the Institute of Navigation, vol. 64, no. 1, pp. 51–66, 2017. https://doi.org/10.1002/navi.183
crossref
13. M. L. Psiaki and T. E. Humphreys, "GNSS spoofing and detection," Proceedings of the IEEE, vol. 104, no. 6, pp. 1258–1270, 2016. https://doi.org/10.1109/JPROC.2016.2526658
crossref
14. D. Mendes, N. Ivaki, and H. Madeira, "Effects of GPS spoofing on unmanned aerial vehicles," In: Proceedings of 2018 IEEE 23rd Pacific Rim International Symposium on Dependable Computing (PRDC); Taipei, Taiwan. 2018, pp 155–160. https://doi.org/10.1109/PRDC.2018.00026
crossref
15. A. Jafarnia-Jahromi, A. Broumandan, J. Nielsen, and G. Lachapelle, "GPS vulnerability to spoofing threats and a review of antispoofing techniques," International Journal of Navigation and Observation, vol. 2012, article no. 127072, 2012. https://doi.org/10.1155/2012/127072
crossref pdf
16. A. J. Kerns, D. P. Shepard, J. A. Bhatti, and T. E. Humphreys, "Unmanned aircraft capture and control via GPS spoofing," Journal of Field Robotics, vol. 31, no. 4, pp. 617–636, 2014. https://doi.org/10.1002/rob.21513
crossref
17. Y. Guo, M. Wu, K. Tang, J. Tie, and X. Li, "Covert spoofing algorithm of UAV based on GPS/INS-integrated navigation," IEEE Transactions on Vehicular Technology, vol. 68, no. 7, pp. 6557–6564, 2019. https://doi.org/10.1109/TVT.2019.2914477
crossref
18. J. Gaspar, R. Ferreira, P. Sebastiao, and N. Souto, "Capture of UAVs through GPS spoofing," In: Proceedings of 2018 Global Wireless Summit (GWS); Chiang Rai, Thailand. 2018, pp 21–26. https://doi.org/10.1109/GWS.2018.8686727
crossref
19. J. Noh, Y. Kwon, Y. Son, H. Shin, D. Kim, J. Choi, and Y. Kim, "Tractor beam: safe-hijacking of consumer drones with adaptive GPS spoofing," ACM Transactions on Privacy and Security (TOPS), vol. 22, no. 2, article no. 12, 2019. https://doi.org/10.1145/3309735

20. R. Ferreira, J. Gaspar, P. Sebastiao, and N. Souto, "A software defined radio based anti-UAV mobile system with jamming and spoofing capabilities," Sensors, vol. 22, no. 4, article no. 1487, 2022. https://doi.org/10.3390/s22041487
crossref pmid pmc
21. H. Sathaye, M. Strohmeier, V. Lenders, and A. Ranganathan, "An experimental study of GPS spoofing and takeover attacks on UAVs," In: Proceedings of the 31st USENIX Security Symposium (USENIX Security 22); Boston, MA, USA. 2022, pp 3503–3520. https://www.usenix.org/conference/usenixsecurity22/presentation/sathaye

22. D. He, Y. Qiao, S. Chen, X. Du, W. Chen, S. Zhu, and M. Guizani, "A friendly and low-cost technique for capturing non-cooperative civilian unmanned aerial vehicles," IEEE Network, vol. 33, no. 2, pp. 146–151, 2019. https://doi.org/10.1109/MNET.2018.1800065
crossref
23. W. Chen, Y. Dong, and Z. Duan, "Accurately redirecting a malicious drone," In: Proceedings of 2022 IEEE 19th Annual Consumer Communications & Networking Conference (CCNC); Las Vegas, NV, USA. 2022, pp 827–834. https://doi.org/10.1109/CCNC49033.2022.9700664
crossref
24. M. Li, Y. Kou, Y. Xu, and Y. Liu, "Design and field test of a GPS spoofer for UAV trajectory manipulation," In: China Satellite Navigation Conference (CSNC) 2018 Proceedings; Singapore, Springer. 2018, pp 161–173. https://doi.org/10.1007/978-981-13-0014-1_15
crossref pmid
25. M. H. Chae, S. O. Park, S. H. Choi, and C. T. Choi, "Commercial fixed-wing drone redirection system using GNSS deception," IEEE Transactions on Aerospace and Electronic Systems, vol. 59, no. 5, pp. 5699–5713, 2023. https://doi.org/10.1109/TAES.2023.3264193
crossref
26. E. Ebeid, M. Skriver, K. H. Terkildsen, K. Jensen, and U. P. Schultz, "A survey of open-source UAV flight controllers and flight simulators," Microprocessors and Microsystems, vol. 61, pp. 11–20, 2018. https://doi.org/10.1016/j.micpro.2018.05.002
crossref
27. Ardupilot. Ardupilot Copter Project, c2024. [Online]. Available: https://Ardupilot.org/plane/

28. Y. Yang, J. Zhou, and O. Loffeld, "Quaternion-based Kalman filtering on INS/GPS," In: Proceedings of 2012 15th International Conference on Information Fusion; Singapore. 2012, pp 511–518.

29. T. Luukkonen, Modelling and control of quadcopter. School of Science, Aalto University, Espoo, Finland: 2011.

30. Advanced Protection Systems. FIELDctrl Range 3D RADAR, c2021. [Online]. Available: https://apsystems.tech/en/products/ultra-precise-3d-mimo-radars/

31. DJI. Phantom 4 Pro V2.0 user manual, c2024. [Online]. Available: https://www.dji.com/downloads/products/phantom-4-pro-v2

Biography

jees-2025-2-r-284i1.jpg
Myoung-Ho Chae, https://orcid.org/0000-0001-7741-1818 received his B.S. and M.S. degrees in radiowave engineering from Chungnam National University, Korea, in 2012 and 2014, respectively. In 2024, he received his Ph.D. degree in electrical engineering from the Korea Advanced Institute of Science and Technology, Daejeon, Korea. He is currently a senior researcher at the Agency for Defense Development in Daejeon, Korea. His research interests include wideband frequency synthesizers, wideband receivers, and electronic warfare systems.

Biography

jees-2025-2-r-284i2.jpg
Seong-Ook Park, https://orcid.org/0000-0002-2850-8063 received his B.S. degree from Kyungpook National University, Korea, in 1987; his M.S. degree from the Korea Advanced Institute of Science and Technology, Daejeon, Korea, in 1989; and his Ph.D. degree from Arizona State University, Tempe, in 1997, all in electrical engineering. From March 1989 to August 1993, he was a research engineer at Korea Telecom, Daejeon, working with microwave systems and networks. He later joined the Telecommunication Research Center at Arizona State University, where he worked until September 1997. In October 1997, he joined the Information and Communications University in Daejeon. He is currently a professor at the Korea Advanced Institute of Science and Technology. His research interests include mobile handset antennas, and analytical and numerical techniques in electromagnetics. Dr. Park is a member of the Phi Kappa Phi.

Biography

jees-2025-2-r-284i3.jpg
Seung-Ho Choi, https://orcid.org/0000-0002-5038-1871 received his B.S. degree from Yeungnam University, South Korea, in 1992; his M.S. degree from Pohang University of Science and Technology in 1998; and his Ph.D. degree from the Korea Advanced Institute of Science and Technology, Daejeon, Korea, in 2008, all in electrical engineering. He later joined the Agency for Defense Development in Daejeon, Korea, where he is currently a principal researcher. His research interests include electronic attack technologies in the EW area.

Biography

jees-2025-2-r-284i4.jpg
Chae-Taek Choi, https://orcid.org/0009-0006-9796-8035 received his B.S. and M.S. degrees in computer science and statistics from Chungnam National University, Korea, in 1989 and 1991, respectively. In 1991, he joined the Agency for Defense Development, where he is currently a principal researcher. His research interests include neural network optimization, RF-counter unmanned aerial system technology, and electronic-warfare systems.

ABOUT
ARTICLE CATEGORY

Browse all articles >

BROWSE ARTICLES
AUTHOR INFORMATION
Editorial Office
#706 Totoo Valley, 217 Saechang-ro, Yongsan-gu, Seoul 04376, Korea
Tel: +82-2-337-9666    Fax: +82-2-6390-7550    E-mail: admin-jees@kiees.or.kr                

Copyright © 2025 by The Korean Institute of Electromagnetic Engineering and Science.

Developed in M2PI

Close layer
prev next