experimental study on radar micro-doppler signatures of...

4
EXPERIMENTAL STUDY ON RADAR MICRO-DOPPLER SIGNATURES OF UNMANNED AERIAL VEHICLES Michael Jian, Zhenzhong Lu, and Victor C. Chen Ancortek Inc., Fairfax, VA, USA Abstract—In the paper, radar micro-Doppler signatures of rotating rotors are investigated for detection and identification of small UAVs. A 24 GHz dual-receiving channel interferometric radar is used to capture useful features of rotating rotors. Interferometric radar with two receiving channels can measure both radial velocity and angular velocity induced micro-Doppler modulations. The study found the angular micro-Doppler signature is a good complementary feature to the radial induced one for identifying small UAVs. Keywords—UAV detection, dual-receiver radar, interferometric radar, micro-Doppler effect, radial micro-Doppler, angular micro- Doppler I. INTRODUCTION Radar detection and identification of small unmanned aerial vehicles (UAVs) flying at low altitudes and moving at slow speeds become an emerging radar task [1–3]. Due to background clutter and other unwanted moving objects, the detection and identification of small and slow-moving UAVs are not easy. Advanced clutter suppression and target detection algorithms must be used. After target detection, an important issue is to classify whether the detected target is a UAV, recognize its category (such as very small group or small group, rotary or flipping wings), and even identify its ID (friends or foes). For the purpose of classification, recognition, and identification, we must first extract characteristics very unique to UAVs. Almost all of the UAVs have at least one or more fast rotating rotors. The fast rotating rotor is a special feature of the UAVs. The four-rotor quadcopter drone has two rotors rotating clockwise and the other two counter-clockwise about the vertical axis. It can take off and land vertically, hover, fly forwards, backward and laterally. Other multi-rotor UAVs, such as hexacopters and octocopters, all have matched sets of rotors rotating in opposite directions. These rotors all consist of blades rotating in a periodic manner as illustrated in Figure 1. Due to the Doppler effect, radar transmitted signal frequency can be shifted when reflected from a moving target. The Doppler shift depends on the signal wavelength, the speed of the target relative to the radar, and the moving direction of the target with respect to the radar line-of-sight (LOS). When the target is moving along the radar LOS, the reflected radar signal has the maximum Doppler shift. If the target is moving across the radar LOS, the reflected signal will have minimum Doppler shift. The radial velocity induced Doppler frequency shift has been widely used to classify targets with motions. For UAVs, the rotating blades of the rotor can periodically modulate the Doppler frequency induced by the bulk motion of the UAV. Thus, UAVs have a unique feature of periodic time- varying Doppler frequency modulations, i.e., micro-Doppler signature [4]. In the paper, we use an experimental radar to observe small UAVs and to extract micro-Doppler signatures. In Section 2, the experimental radar system configuration is introduced. Section 3 discusses the radar micro-Doppler signatures of UAVs. Experimental study of UAVs’ micro-Doppler signatures is given in Section 4. Fig. 1. Radar monitors small UAVs. II. EXPERIMENTAL RADAR SYSTEM CONFIGURATION A K-band (24.125 GHz) dual-receiving-channel radar is used to detect UAVs and extract micro-Doppler features of rotating rotors. The radar consists of an RF module and an FPGA-based processor module [5]. One transmitting antenna and two receiving antennas are connected to the RF module. Figure 2 shows the high-level block diagram of the K-band radar system. The RF module is based on a highly integrated Infineon transceiver BGT24MTR12 SiGe MMIC. The FPGA- based processor module has one DAC, four ADCs, and a USB interface. A graphical user interface (GUI) provides an interface for users to control the radar system via a USB 2.0 cable and to view the real-time range-velocity map and collect data. The Infineon RF transceiver has one transmit and two receive channels, VCO and mixer circuits on the MMIC. By incorporating a phased-lock loop (PLL) circuit, the RF module is improved in terms of linearity, reliability, and sensitivity. The phase noise measured at the transmit port is -96 dBc @ 1MHz offset. A power amplifier is used to get an output power of 16 dBm. The mixer circuit mixes the transmit signal with the receive signals in each receiving channel to generate beat signals, each of which has an in-phase (I) channel and a 978-1-4673-8823-8/17/$31.00 ©2017 IEEE 0854

Upload: lyminh

Post on 26-May-2019

215 views

Category:

Documents


0 download

TRANSCRIPT

EXPERIMENTAL STUDY ON RADAR MICRO-DOPPLER SIGNATURES OF UNMANNED AERIAL VEHICLES

Michael J ian, Zhenzhong Lu, and Victor C. Chen

Ancor tek Inc., Fair fax, VA, USA

Abstract—In the paper, radar micro-Doppler signatures of rotating rotors are investigated for detection and identification of small UAVs. A 24 GHz dual-receiving channel interferometric radar is used to capture useful features of rotating rotors. Interferometric radar with two receiving channels can measure both radial velocity and angular velocity induced micro-Doppler modulations. The study found the angular micro-Doppler signature is a good complementary feature to the radial induced one for identifying small UAVs.

Keywords—UAV detection, dual-receiver radar, interferometric radar, micro-Doppler effect, radial micro-Doppler, angular micro-Doppler

I. INTRODUCTION Radar detection and identification of small unmanned aerial

vehicles (UAVs) flying at low altitudes and moving at slow speeds become an emerging radar task [1–3]. Due to background clutter and other unwanted moving objects, the detection and identification of small and slow-moving UAVs are not easy. Advanced clutter suppression and target detection algorithms must be used. After target detection, an important issue is to classify whether the detected target is a UAV, recognize its category (such as very small group or small group, rotary or flipping wings), and even identify its ID (friends or foes). For the purpose of classification, recognition, and identification, we must first extract characteristics very unique to UAVs.

Almost all of the UAVs have at least one or more fast rotating rotors. The fast rotating rotor is a special feature of the UAVs. The four-rotor quadcopter drone has two rotors rotating clockwise and the other two counter-clockwise about the vertical axis. It can take off and land vertically, hover, fly forwards, backward and laterally. Other multi-rotor UAVs, such as hexacopters and octocopters, all have matched sets of rotors rotating in opposite directions. These rotors all consist of blades rotating in a periodic manner as illustrated in Figure 1.

Due to the Doppler effect, radar transmitted signal frequency can be shifted when reflected from a moving target. The Doppler shift depends on the signal wavelength, the speed of the target relative to the radar, and the moving direction of the target with respect to the radar line-of-sight (LOS). When the target is moving along the radar LOS, the reflected radar signal has the maximum Doppler shift. If the target is moving across the radar LOS, the reflected signal will have minimum Doppler shift. The radial velocity induced Doppler frequency shift has been widely used to classify targets with motions.

For UAVs, the rotating blades of the rotor can periodically modulate the Doppler frequency induced by the bulk motion of the UAV. Thus, UAVs have a unique feature of periodic time-varying Doppler frequency modulations, i.e., micro-Doppler signature [4].

In the paper, we use an experimental radar to observe small UAVs and to extract micro-Doppler signatures. In Section 2, the experimental radar system configuration is introduced. Section 3 discusses the radar micro-Doppler signatures of UAVs. Experimental study of UAVs’ micro-Doppler signatures is given in Section 4.

Fig. 1. Radar monitors small UAVs.

II. EXPERIMENTAL RADAR SYSTEM CONFIGURATION A K-band (24.125 GHz) dual-receiving-channel radar is

used to detect UAVs and extract micro-Doppler features of rotating rotors. The radar consists of an RF module and an FPGA-based processor module [5]. One transmitting antenna and two receiving antennas are connected to the RF module.

Figure 2 shows the high-level block diagram of the K-band radar system. The RF module is based on a highly integrated Infineon transceiver BGT24MTR12 SiGe MMIC. The FPGA-based processor module has one DAC, four ADCs, and a USB interface. A graphical user interface (GUI) provides an interface for users to control the radar system via a USB 2.0 cable and to view the real-time range-velocity map and collect data.

The Infineon RF transceiver has one transmit and two receive channels, VCO and mixer circuits on the MMIC. By incorporating a phased-lock loop (PLL) circuit, the RF module is improved in terms of linearity, reliability, and sensitivity. The phase noise measured at the transmit port is -96 dBc @ 1MHz offset. A power amplifier is used to get an output power of 16 dBm. The mixer circuit mixes the transmit signal with the receive signals in each receiving channel to generate beat signals, each of which has an in-phase (I) channel and a

978-1-4673-8823-8/17/$31.00 ©2017 IEEE 0854

quadrature-phase (Q) channel. The limiter showing in Figure 2 protects the receiver circuits from unwanted damages.

Fig. 2. Block diagram of the dual-receiver radar.

The core of the processor module is the Cyclone IV FPGA processor, which interfaces with the four ADCs, DAC, and USB. Because of FPGA’s programmability and re-configurability, without modifying the hardware, the radar could be adopted in different scenarios and change its operation modes, waveforms, frequency bandwidths, and processing modes. Digital samples of waveform control voltage are generated by the FPGA firmware, and then converted to analog control voltage through DAC.

The GUI communicates over the USB interface and receives a continuous stream of raw data from the radar. The data are then processed, and the resulting range-Doppler map and other maps are displayed on the GUI in real-time. The GUI provides a range of user control over the radar, such as waveform selection, bandwidth, the number of samples, PRF, filtering, clutter reduction, and so on.

III. EXTRACTING MICRO-DOPPLER SIGNATURES FROM UAVS Due to the background and other unwanted moving objects,

returned radar signals show strong clutter around zero-Doppler and in low-frequency region. For detecting small UAVs, clutter must be suppressed and subtracted from subsequent measurements.

A. Radial Velocity Induced Micro-Doppler Effect in UAVs Based on the Doppler effect, radar reflected signal from a

moving target will be shifted from the transmitting signal. The Doppler shift 𝑓" = 2𝑣&'"(')/𝜆 depends on the relative velocity 𝑣&'"('), the transmitting wavelength λ and moving direction of the target with respect to the radar. When the target is moving along the radar LOS, the reflected radar signal will have the maximum Doppler shift. The radial velocity induced Doppler frequency shift has been widely used to detect moving targets.

B. Angular Velocity Induced Micro-Doppler Effect in UAVs Besides the radial velocity, a moving target also has

tangential velocity if it is moving across the radar LOS. The tangential velocity is equivalent to the angular velocity.

A technique that measures the angular velocity of a moving target was proposed in [6,7]. It is based on the radar interferometry principle using two coherent receivers. When a moving target passes through the interferometric radar beam pattern, it will produce a periodic modulated signal, which is directly proportional to the angular velocity of the target. The angular velocity has frequency shifts from its transmitted frequency. The angular velocity induced frequency shift is determined by the equivalent rotation rate of the target, the spacing between the two receiving antennas, and the wavelength of the radar signal.

The angular velocity ω can be measured from the interferometric frequency 𝑓, , the baseline between the two receiving antennas 𝐷, and the wavelength λ by 𝜔 = 𝑓,𝜆/𝐷. By cross-correlation processing between two receiving channels, we can calculate the interferometric frequency shift 𝑓, when an angular velocity ω is present. For a given angular velocity, a longer baseline D will result in a higher frequency shift. Same as the radial velocity induced micro-Doppler, the angular velocity induced micro-Doppler signature can also be extracted by using the short-time Fourier transform.

As for UAVs, there can be little radial velocity, but the angular velocity remains the same. The high angular velocity makes angular micro-Doppler always evident. Thus, a combined radial velocity micro-Doppler and angular velocity micro-Doppler become a unique feature, especially for UAVs, which makes it possible to distinguish UAVs from other non-rotor objects.

IV. EXPERIMENTAL STUDY ON EXTRACTING MICRO-DOPPLER SIGNATURES FROM UAVS

The geometry of the radar and the quadcopter is illustrated in Figure 3, where the baseline spacing D between two receiving antennas and the depression angle β of the radar can be adjusted. The quadcopter drone has 4 rotors operated by 4 corresponding motors. Each motor has a number of metal poles. A propeller in a rotor is described by the diameter, pitch, and the number of blades. The diameter of the propeller used in our quadcopter drone is 15cm with two blades. The rotation rate of the propeller is more than 30 revolutions per second. Radar returns from the rotor can be modulated by the blade rotation, by rotating motor poles, and even by the propeller’s pitching.

Fig. 3. The geometry of the radar and the small UAV.

978-1-4673-8823-8/17/$31.00 ©2017 IEEE 0855

For our 15cm diameter’s propeller, the tip velocity of the blade (at a rotation rate of 30 revolutions per second) is v = 14m/s or the Doppler frequency of 2v/λ = 2.24kHz, where λ = 1.25 cm. Thus, the required sampling rate must be at least the Nyquist rate of 4.5 kHz. If we use FMCW waveform, the required FM sweep time must be shorter than 1/(4.5kHz) = 0.2ms, which is beyond the shortest sweep time of the experimental radar. However, if we use CW waveform, the Nyquist sampling rate of 4.5kHz is easy to achieve. Thus, in our experimental study, we use CW signal waveform to extract micro-Doppler signatures.

A. Scenario 1: Radar LOS is parallel to the axis of the rotating rotor In this scenario, only a single rotor on the quadcopter was

used. Radar angle of depression is 90º and the spacing between two receiving antennas is D = 5λ. Thus, the radar faces the top of the rotors, and the radial velocity of the rotating blades along the LOS is at its minimum as shown in Figure 4(a). However, the tangential velocity when the blades rotate across the radar LOS still exists. Thus, the angular velocity of the rotating blades and the motor are still at their maximum.

Figure 4 shows a radial micro-Doppler signature and an angular micro-Doppler signature of a single rotor on the quadcopter. Because the radar LOS is parallel to the axis of rotating with a 90º angle of depression, we can see the radial velocity induced micro-Doppler modulation is at its minimum. The micro-Doppler shifts of the blades’ tips cannot be seen. The strong returns are due to metal motor poles.

However, tips of rotating blades still have higher angular velocity and the angular micro-Doppler shifts are still high as seen in Figure 4(b).

Fig. 4. Micro-Doppler signatures of a single rotor.

B. Scenario 2: Radar LOS is 45º to the axis of the rotating rotor In this scenario, the spacing between two receiving antennas

is still D = 5λ, but the radar angle of depression becomes 45 º. In this case, radar can sense both the radial velocity and the angular velocity. Figure 5 shows that the radial micro-Doppler shifts of

the blades’ tips have been significantly increased compared to that in Figure 4(a).

Fig. 5. Significantly increased micro-Doppler signatures due to larger radial

velocity components.

C. Scenario 3: Radar LOS is perpendicular to the axis of the rotating rotor In this scenario, all four rotors on the quadcopter were used.

The spacing between two receiving antennas is still D = 5λ. When the radar angle of depression is 0º, the radar LOS is

perpendicular to the rotor rotation axis. Thus, the radar can sense the maximum radial velocity as seen in Figure 6.

During testing, all four rotors simultaneously rotated, but the initial rotation angle of each rotor was random. Figure 6(a) shows the radial micro-Doppler signature and 6(b) shows the angular micro-Doppler signatures of all four rotors. Because the radar LOS is perpendicular to the axis of rotation, the radial velocity induced micro-Doppler modulation is at its maximum. In Figure 6(a), we can clearly see the radial micro-Doppler signature of the four rotors. Because of random initial rotation phases, we can no longer see the periodic micro-Doppler pattern. The strong returns still come from metal motor parts.

Fig. 6. Micro-Doppler signatures of all four rotors of the quadcopter.

V. DISCUSSION For detection and identification of small UAVs, unique micro-Doppler features of rotating rotors on UAVs can be used and extracted by radar. We used an experimental Doppler radar to investigate these features. From our preliminary results, we found:

978-1-4673-8823-8/17/$31.00 ©2017 IEEE 0856

(1) Radial velocity induced micro-Doppler modulation is a most useful feature of the multi-rotor UAVs. However, for some geometry configurations between radar and UAVs, the radial velocity induced micro-Doppler signatures can be inconspicuous.

(2) Angular velocity induced micro-Doppler modulation is a good complementary feature to the radial induced one. Especially for small UAVs, the angular micro-Doppler signature is always noticeable.

(3) For small UAVs, due to fast rotation of the rotors (30 revolutions per second), to have a significant interferometric frequency shift (for example 1-2 kHz), relatively shorter baseline between two receiving antennas (for example, D = 5λ) is good enough.

(4) For multi-rotor UAVs, because each rotor has its own arbitrary initial rotating phase, the multi-rotor

combined micro-Doppler signatures cannot clearly show periodic Doppler modulations.

REFERENCES

[1] Drozdowicz, J. et al. “35 GHz FMCW drone detection system,” International Radar Symposium (IRS), May 10-12, 2016.

[2] Hoffmann, F., Richie, M., Fioranelli, F. et al., “Micro-Doppler based detection and tracking of UAVs with multistatic radar,” 2016 IEEE Radar Conference, May 2016.

[3] de Haag, M. U., Bartone, C. G., and Braasch, M. S., “Flight-test evaluation of small form-factor LiDAR and radar sensors for sUAV detect-and-avoid applications,” IEEE Digital Avionics Systems Conference, Sep. 26-29, 2016.

[4] Chen, V.C., The Micro-Doppler Effect in Radar, Artech House, 2011 [5] www.ancortek.com [6] Nanzer, J., “Micro-motion signatures in radar angular velocity

measurements,” 2016 IEEE Radar Conference, May 2016. [7] Nanzer, J. and Zilevu, K., “Dual interferometric-Doppler measurements

of the radial and angular velocity of human,” IEEE Trans. on Antennas and Propagation, vol.62, no.3, pp.1513-1517, 2014.

978-1-4673-8823-8/17/$31.00 ©2017 IEEE 0857