09ATC-0385

Acoustic vector sensors increasing UAV’s situational awareness

Hans-Elias de Bree Microflown Technologies / HAN University dept. vehicle acoustics, Arnhem, the Netherlands

Copyright © 2009 SAE International

ABSTRACT

Acoustic vector sensors (AVS), as they were recently developed based upon acoustic particle velocity sensors may improve the performance of Unmanned Aerial Vehicles (UAV) in terms of automatic take off and landing (ATOL), ground surveillance and mid air anti collision.

Any sound field can be described by both the scalar value sound pressure and the vector value acoustic particle velocity.

Only with the recent invention of the Microflown sensor, acoustic particle velocity in air has become a measurable quantity.

The combination of a sound pressure and three acoustic particle velocity sensors in one single node creates a broad banded acoustic vector sensor that extends the audible range.

Whereas the concept of acoustic vector sensors is known in the radar community and for underwater acoustics, it was never applied in the air due to a lack of appropriate particle velocity sensors.

AVS offer a set of features that are complementary to other categories of sensors, such as radars, lasers and electro optical cameras.

AVS are passive, and their performance is not affected by adverse weather conditions like rain, clouds, fog, dust or a lack of daylight

AVS cover the entire audible range and can detect sound sources without a line of sight, e.g. behind a mountain. The detection range may exceed ten kilometers.

For airborne applications in general, the low power consumption, the low weight and small footprint are of relevance.

Acoustic vector information can be obtained even in the presence of wind and background noise as generated by the airborne vehicle itself.

As AVS have a full and complete spherical field of detection, they can be used to cue narrow field of view sensors.

For automatic take off and landing purposes, completely sound pressure based acoustic detection systems on the ground were reported to be able to determine the geometric position of a helicopter with a high level accuracy.

The deployment of AVS allows the integration of such a system on a lesser number of measurement nodes, whilst increasing the frequency range of the system and taking into account novel ground impedance modeling.

Successful tests were done on helicopters in 2008 and 2009.

For airborne based ground reconnaissance, arrays of sound pressure transducers are traditionally used. Their frequency range is limited by their spacing, the outcome of all their signals combined one produces a single vector.

AVS are broad banded, and each node itself already provides the full vector information. A collaborative mode only improves the accuracy of the information.

Civil aviation agencies put stringent requirements on the use of UAV in nonsegregated airspace.

As AVS provide in each node both amplitude and phase information, advanced signal processing techniques can be used to locate and track a large number of intruders in the “UAV bubble” simultaneously.

With four AVS, up to thirty broad banded sound sources can be located and monitored simultaneously.

In this paper, both the performance of the AVS and novel related signal processing routines will be discussed.

Results of simulations, laboratory and outdoor test results of this ongoing research will be presented.

THE MICROFLOWN SENSOR

Any sound field can only be described completely by knowing both dimensions: the scalar sound pressure and the vector acoustic particle velocity.

If sound pressure were the acoustic equivalent of voltage, acoustic particle velocity is the acoustic equivalent of current. Only with the recent invention of the Microflown sensor, acoustic particle velocity has become a directly measurable quantity [1].

Based upon MEMS technology, the Microflown sensor working principle uses two extremely sensitive heated platinum wires that have hardly any heat capacity. If airflow occurs around these wires, heat transfer will take place, causing the upstream wire to be cooled down and the air to be heated. As a consequence, the downstream wire is cooled down just a little bit less. The temperature difference that will arise in the cross section happens to be a direct measure for the acoustic particle velocity. The sensor is linear and provides output in voltage.

Fig. 1: the Microflown sensor

ACOUSTIC VECTOR SENSORS

An acoustic vector sensor is created by three orthogonally placed particle velocity sensors and a sound pressure sensor at the same location. Three different types of acoustic vector sensors are available.

The first realization is an assembled 4 channel probe, combining three orthogonally placed acoustic particle velocity sensors and a 1/10” sound pressure transducer, the so called USP.

If the Microflown is packaged in a half inch housing the sensitivity increases. A set of three one dimensional PU probes, a six channel so called triple PU. This setup has a 10dB higher sensitivity as the USP and is normally used for lower frequency higher range applications.

Furthermore, a flat, but 3D, unintrusive monolithic sound chip. The chip consists of a fully integrated 3D Microflown and an assembly that is sound pressure sensitive [2].

Fig. 2: the USP

Fig. 3: the triple PU

Fig. 4: the 3D monolithic sound chip

The Microflown sensors have a very good signal to noise ratio at lower frequencies. As a comparison: for frequencies below 1kHz a half inch packaged Microflown has a higher signal to noise level as a half inch sound pressure microphone.

SOURCE LOCALIZATION WITH AVS

The Microflown measures the acoustic particle velocity instead of the acoustic pressure which is measured by conventional sound pressure microphones. With three perpendicular Microflowns and a microphone at the same place an acoustic vector sensor (AVS) is constructed.

Sound pressure is a scalar value and therefore sound pressure microphones do not have any directionality. Directional systems that are based on microphones make use of a spatial distribution and the directivity is based on phase differences between the sound pressure at the different locations. There is no directional information found in the amplitude responses. Because the phase shifts are caused by spatial distribution, the method is depending on the wavelength and is thus frequency dependent.

Acoustic vector sensors (AVS) are directional, so making a directional system is relatively straightforward. Because a single AVS measures the sound field at one point, there is limited phase information and the directional information is found in the amplitude responses of the individual particle velocity probes.

Benefits of AVS versus arrays of microphones are fast setup times, low data acquisition channel count and no (lower and higher) frequency limit. The frequency is limited by the sensors (in the order of 0Hz-120kHz).

UAV’S SITUATIONAL AWARENESS

The UAV’s situational awareness can be divided in three themes:

• anti collision systems during the flight,

• ground surveillance for operations

• automatic take off and landing (ATOL) systems to start and end the mission.

Each of these topics will be dealt with below. First some background information and the possible set up will be explained then some experiments will be presented.

The are several possible realizations thinkable for an ATOL system.

ATOL I: SENSORS ON THE GROUND

The acoustic localization of the UAV can be on the ground. This has the advantage that less equipment has to be carried with the UAV and that the system can be used for multiple UAV’s. One disadvantage is the need for a datalink from the ground to the UAV and the acoustic system is required at the landing site (so if a mission is a flight from A to B, a system is required at both locations).

There are several techniques to find the 3D location (bearing angle, elevation angle and distance) of a single

dominant source. These are summarized in [12]. For an automatic take off and landing system (ATOL) it is possible to locate the acoustic sensors on the ground, locate the noise of the UAV and transmit the location to the UAV.

The simplest method for localization is using two spaced AVS, see Fig. 5. The influence of the ground reflection is neglected or masked with acoustic damping material and with this a reasonable estimation of the source position is computed.

Fig. 5: Triangulation with 2 AVS probes

With two of these probes placed at different locations not only the direction but also the position can be determined, using triangulation. Such a measurement has been performed outdoors in order to localize a helicopter [9], [10]. Some results for the reconstructed trajectory during the landing procedure are given in Fig. 11. The trajectories are close to the real trajectories of the source, but improvements can still be made.

Fig. 6: Trajectory measurement of a landing helicopter

The alignment problems are solved by an orientation calibration [11].

If the AVS are positioned directly on the ground the bearing and elevation can be computed in a slightly different method [12]. With two spaced AVS placed directly on the ground the location of the acoustic source can be computed by triangulation [13].

Sound source

A B

x

y z

A measurement showing this method is explained below.

The probes are spaced 16cm from each other in a gym with rather reflective walls. Therefore the results are smoothed by a moving average, which assumes that the gym is fairly diffuse and that the result in a certain frequency band is not influenced by reflections [14].

500 1000 1500 2000 2500 3000 3500 40000

10

20

30

40

50

60

70

80

90

38DEG

53DEG

83DEG

85DEG

500 1000 1500 2000 2500 3000 3500 40000

10

20

30

40

50

60

70

80

90

20DEG

41DEG

72DEG

85DEG

Fig. 7: Angle estimation of both AVS as function of frequency

The position of the source is computed out of the two angles by triangulation. The result is shown in Fig. 8.

-1 -0.8 -0.6 -0.4 -0.2 0 0.20

0.2

0.4

0.6

0.8

1

Fig. 8: True source locations (dots) and estimated source angles of the two AVS. Where the lines cross, a source is expected.

With some more mathematics applied, it is also possible to compute the location 3D location (bearing, elevation and distance) with a single AVS [13]. A schematic drawing is shown in Fig. 9.

x ’L

z

RQ

AVS

Q

ground, Rh

β1

β2

r1

r2

Fig. 9: An AVS elevated from the ground. With this set up it is possible to find the location of a UAV.

The idea of the elevated AVS is that the bearing and elevation angle of the source and the mirror source are determined. Out of the difference in these angles, the distance can be derived [13].

ATOL II: SENSOR IN THE UAV, ACOUSTIC BEACON ON GROUND

A beacon on the ground can be used to guide a UAV to a landing site. The AVS in the UAV localizes the beacon and use its location for navigation.

To be able to apply the sensor in a flying UAV it must be used in the presence of flow that is caused by the aircraft speed. Apart from that, the size and mass are important figures.

In a Dutch National R&D project the Microflown sensor is made suitable for this. An AVS was tested up to 70m/s in a nose cone packaging [8], see Fig. 10.

Initial tests are done with an AVS inside a UAV wing. No airborne tests are done yet but are scheduled in the year 2009.

The development of the 3D monolithic sound chip (see Fig. 4) lowered the mass and the size of the system.

Fig. 10: An AVS in a 15mm in diameter nosecone tested up to 70m/s.

With two of these nosecone AVS it must be possible to 3D locate a source on the ground. This idea is not tested yet but is a topic for future research.

ATOL III: SENSOR IN THE UAV, ACOUSTIC SIGNAL FROM THE UAV

If the UAV makes noise itself or if a noise source in the UAV can be applied it is possible to use this signal and localize the reflections of the ground. Such realization can be used if the landing site is unknown.

The properties of the soil can be determined (the reflection coefficient) and it can be sensed if the landing site is flat or if obstacles are blocking the landing.

An AVS noise mapping algorithm (e.g. beam forming) must be used for this to be able to get an ‘acoustic reflection picture’ of the landing area. This is a technique similar to sonar. The AVS based beam forming algorithm is used for underwater applications (see e.g. [15]), and with the Microflown these algorithms can be applied for air acoustics.

ACOUSTIC ANTI COLLISION SYSTEM

The AVS sensor resistance to the high wind conditions, the small sensor size and low sensor mass, and the ability to find acoustic sources in 3D space are the ingredients for future R&D towards an acoustic anti collision system for UAV’s.

The difference from automatic take off and landing systems is that now multiple sources and sources of unknown signature have to be located.

Modern aircraft can use several types of collision avoidance systems to prevent unintentional contact with other aircraft, obstacles, or the ground. Radar or electro-optical systems the most used techniques for this.

Radars or electro-optical systems are large, heavy, costly and usually does not have a full spherical coverage.

Acoustic vector sensors are not affected by night or day conditions, rain, fog or clouds. However, there is a dependency upon meteorological conditions such as wind and temperature. These can be accounted for to a large extent.

Acoustic vector sensors can help to steer other types of sensors that have limited angle of aperture (narrow field of view) but maybe a higher spatial resolution. Thus, a light weight sensor configuration with steerable sensors becomes possible, e.g. a laser to confirm the range detected.

Acoustic systems are used for UAV’s situational awareness Sound pressure microphone based arrays have been discussed as a method for acoustic sensing in several papers, see e.g. [7].

The traditional systems are based on a limited number of sound pressure microphones see e.g. Fig. 11.

Fig. 11: Acoustic systems on UAV’s.

Due to the limited number of sound pressure sensors it is only possible to find a few sources in a limited bandwidth.

The use of AVS (as opposed to sound pressure microphones) in these cases will increase the number of sources that can be detected by at least a factor of 4, and if the sources are broad banded, possibly a factor of

8. With the use of AVS it is possible to also estimate the range of an acoustic source [9], [10], [12]. The bandwidth will be increased and the detection area will be the full sphere. See also Fig. 12.

Fig. 12: sound pressure sensors on a UAV can determine the bearing of a single source in a limited bandwidth. Right: with 4 AVS it is possible to find the bearing elevation and range of multiple (up to 30) sources.

It is possible to find two uncorrelated sources with a single AVS and with multiple (n) spaced AVS it is possible to find 4n-2 uncorrelated sources in 3D space [3], [4], [5], [6].

In the example shown in Fig. 13 four sources are switched on and measured with two spaced AVS. With a MUSIC algorithm the sources are found.

Fig. 13: left: measurement setup, right MUSIC results using two AVS. In this example 4 sources are found.

If the sources are broadband it is in principle possible to discriminate 8n-2 sources in 3D space. For this a PARAFAC algorithm is used. These algorithms are still under investigation [14]. A preliminary result is shown below: only a partial AVS is used (3 signals) and three sources are found.

−180 −135 −90 −45 0 45 90 135 180

−90

−45

0

45

90

Fig. 14: PARAFAC results using 20 frequencies based on vx, vy and p. Exact (○) and Calculated (×)

It is envisioned that for inflight Sense & Avoid purposes, a system of acoustic vector sensors:

• has a full and complete spherical field of detection

• is passive and can not be detected itself

• can be used autonomously

• can detect in the broad banded audio range

• can detect multiple intruders simultaneously

• can detect at larger range when airborne as compared to on the ground

• works under all weather conditions (day, night, rain, clouds)

• works without a line of sight

• can cue/steer other narrow field of view sensors, such as laser

• requires little power consumption and footprint

• has limited weight

ACOUSTIC GROUND SURVEILLANCE

Once a set of AVS is applied in a UAV it opens the way for acoustic ground surveillance.

The application of gun shot localization is tested in the field. Shots from a 9mm handgun are easily found at 1km distance [4].

Also ground vehicles that produce low frequency harmonic signals can be tracked. Arrays based on sound pressure microphones have difficulties to locate low frequency sources. AVS do not have this disadvantage. They can locate sources independent of frequency [6], [9].

Both source localization techniques and ‘noise mapping’ techniques (like beam forming) can be used simultaneously.

Source localization techniques will pinpoint sources like armored vehicles or gunshots but does not give any information regarding noise levels in between the sources.

Noise mapping techniques like beamforming produce an ‘acoustic picture’ of a surrounding. These results will show the dominant sources that are also found with source localization but with less accuracy. Advantage of the noise mapping technique is that it will give a complete ‘acoustic picture’ of the surrounding.

REFERENCES

1. H.E. de Bree, P. Leussink, T. Korthorst, H. Jansen,

T. Lammerink and M. Elwenspoek, “The Microflown:

a novel device measuring acoustical flows”, Sensors

and Actuators A-Physical, 54, pp 552-557, 1996.

2. D.R. Yntema et al, Fully integrated three dimensional

sound intensity sensor. MEMS 2007, Kobe, Japan,

2007

3. Malcolm Hawkes and Arye Nehorai, Wideband

Source Localization Using a Distributed Acoustic

Vector-Sensor Array; IEEE transactions on signal

processing, vol. 51, no. 6, JUNE 2003

4. HE de Bree, A Perspective on Acoustic Vector

Sensors in Passive Surveillance - Real World

Measurements, Algorithms and Applications, Aero

India, Bangalore, India, 2009

5. Jelmer W. Wind, Emiel Tijs and Hans-Elias de Bree,

Source localization using acoustic vector sensors, a

MUSIC approach, NOVEM, Oxford UK, 2009

6. T. Basten, HE de Bree, and WF druyvesteyn,

Multiple incoherent sound source localization using a

single vector sensor, ICSV16, Krakow, Poland, 2009

7. Acoustic sensing for urban battlefield applications;

Tien Pham, US Army Research Laboratory; ICA

2007, Madrid, Spain

8. Hans-Elias de Bree and Emiel Tijs, PU nosecone

intensity measurements in a wind tunnel, DAGA,

2009

9. T. Basten, HE de Bree, and E. Tijs, Localization and

tracking of aircraft with ground based 3D sound

probes, ERF33, Kazan, Russia, 2007

10. T.G.H. Basten, H.E. de Bree, S. Sadasivan, Acoustic

eyes, a novel sound source localization and

monitoring technique with 3D sound probes, ISMA,

2008

11. T.G.H. Basten, H.E. de Bree, D.R. Yntema, An

orientation calibration procedure for two acoustic

vector sensor configurations, ICSV, Krakow, 2009

12. H-E de Bree, C. Ostendorf, Tom Basten, An acoustic

vector based approach to locate low frequency noise

sources in 3D, DAGA, 2009

13. H-E de Bree and W.F. Druyvesteyn, An acoustic

vector sensor based method to measure the bearing,

elevation and range of a single dominant source as

well as the ground impedance, Euronoise, Edinburg,

2009

14. H.E. de Bree, E. Tijs, W.F. Druyvesteyn, J. W. Wind,

broad banded acoustic vector sensors for passive

monitoring of aircraft, DLRK, Aachen, Germany,

2009

15. Joseph A. Clark and Gerald Tarasek, Localization of

Radiating Sources along the Hull of a Submarine

Using a Vector Sensor Array, IEEE, 2006