Acoustic vector sensors increasing UAV’s situational awareness

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