This study covers three aspects of acoustic localisation of drones using a microphone array. First, it assesses a grid-free approach, using differential evolution, to estimate the three-dimensional position of a drone. It is found that this is indeed possible for the drone in the near-field. For larger distances, it still provides the angular position of the drone. Second, the study emphasizes the essence of localisation over small frequency bands with the bands jointly spanning a large frequency range to reveal the presence of multiple sound sources and maximise the drone localisation range. Third, it addresses the localisation ranges for six different drones.

Authorities are starting to pay attention not only to the noise levels of Unmanned Aerial Vehicles (UAVs) but also to their quality for acceptance. This manuscript presents a study of four types of propeller-driven UAVs (single-propeller quadcopter, coaxial-propeller quadcopter, quadplane eVTOL (electric vertical take off and landing) and tailsitter eVTOL) to assess their acoustic and psychoacoustic signatures. Experimental outdoor recordings are conducted under realistic flyover conditions. An acoustic analysis showed that quadcopters present higher noise levels compared to the eVTOLs, where the coaxial-propeller configuration revealed to be the noisiest and the quadplane the quietest. A psychoacoustic analysis demonstrated that the coaxial-propeller quadcopter was roughly three times more annoying than its single-propeller counterpart, whereas the quadplane and tailsitter eVTOLs showed similarly lower annoyance values. Additionally, the coaxial-propeller quadcopter exhibited the highest levels of loudness and impulsiveness, while the tailsitter had the lowest. Conversely, the tailsitter exhibited contrasting behavior in terms of sharpness. Regarding tonality, the quadplane was the most tonal, and the tailsitter eVTOL the least. In terms of modulation frequency characteristics, the single-propeller UAV emitted the harshest and most pulsating sound, while the tailsitter had lower values.

The use of acoustics for drone localisation has gained more interest in recent years. Traditionally, acoustic localisation is done using microphone arrays. The data is processed with methods such as a time-difference-of-arrival approach or beamforming. It has however not been investigated yet if alternative noise monitoring systems can be used for this purpose. The aim of this research is therefore to investigate if noise monitoring systems which output noise directivity can be used for drone localisation. Four SV200A noise monitoring systems were used during a real-life experiment. These systems output the directivity in the XY-plane from 0◦ to 360◦ and in the Z-plane from 0◦ to 180◦ in bins with a width of 11.25◦. During the experiment, a DJI Phantom quadcopter flew a variety of manoeuvres. For performance evaluation, two flight patterns are distinguished: horizontal and vertical flight. Initial results showed limited performance. In the XY-plane, the percentage of time where the correct bin was estimated varied between 20% to 49% depending on the manoeuvre and microphone, whereas in the Z-plane this was only 0.3% to 14%. However, when including the angular bins adjacent to the correct bin, the performance increased significantly. In the XY-plane, the percentages increased to 63% to 98% and in the Z-plane to 29% to 49%. A reason for this increase can be that the GPS receiver of the drone has limited accuracy, which leads to a mismatch between the true flown position and the logged position. Furthermore, due to an offset in internal clocks of the drone and microphones, some microphone estimations were slightly ahead or behind the actual drone movement. In conclusion, these results show that the noise monitoring systems can estimate the location of the drone if some margin in the estimation is taken into account.

This study investigated the acoustic and psychoacoustic properties of five quadcopters drones during realistic flyover scenarios, utilizing a 64-microphone array for outdoor recordings. Acoustic analyses encompassed signal-to-noise ratio (SNR) values, time-frequency sound pressure levels, and noise spectra at overhead positions. An analysis based on A-weighted SNR revealed discernible drone noise despite background noise. Significant noise levels were observed up to 12 kHz. Harmonics of blade passage frequencies were evident, influencing noise spectra up to 1 kHz. Unlike traditional aircraft, drones' proximity to the ground limits the atmospheric absorption effects of high-frequency noise. A psychoacoustic analysis focused on sound quality metrics (SQMs) and annoyance assessment. SQMs exhibited consistent patterns across attributes, such as sharpness, tonality, roughness, and impulsiveness, with notable drone-specific perceptions. Different annoyance models indicated varying degrees of annoyance perception, with the Autel EVO II drone (lowest installation ratio, defined as the ratio between the drone diagonal size and the propeller diameter) perceived as the most annoying and the DJI Phantom 4 (heaviest) as the least one. Propeller positioning, represented by the parameter of installation ratio, correlated significantly with annoyance levels, suggesting an influence on both noise signature and psychoacoustic response. These findings highlight the importance of understanding the acoustic and psychoacoustic impact of drones, particularly in urban environments.

This manuscript summarizes the main recent research efforts at Delft University of Technology in the field of drone and urban air mobility (UAM) vehicle noise. Illustrative examples are showcased, specifically in terms of acoustic measurements (both in-field and in wind-tunnel facilities), noise modelling (both data-driven and physics-based), and human perception of these sounds. In particular, the measurements feature microphone arrays and acoustic imaging to detect, localize, and isolate drone noise emissions. Regarding drone noise modelling, the proposed approaches cover noise generation, propagation, and acoustic footprint calculation. The evaluation of the human perception of drone noise and the perceived annoyance is another crucial aspect. To this end, psychoacoustic listening experiments are conducted in laboratory conditions and the results are analyzed using perception-based sound metrics. Data from aeroacoustic measurements and synthetic sound auralizations are considered. Combining these three main approaches holistically, the perception-driven design and assessment can be performed by targeting the minimization of the perceived noise annoyance, rather than merely reducing sound pressure levels.

Correction Notice • Place s5 in front of the symbols ‘greater than or equal to (≥)’ and ‘less than (<)’ in Equation 2, as shown below. (Formula Presented). • Replace the correlation coefficient ‘0.886’ showed in Table 6 to ‘-0.886’. This correction also apply to the paragraph below Section 3 (Correlation between Annoyance and Drone Characteristics) and in the last paragraph of the Conclusions. This negative value indicates an inverse relationship between the installation ratio (d/D) and the annoyance computed using the More PA model. • Replace ‘0.337 (Zwicker PA model)’ by ‘0.362 (Willemsen PA model)’ in the paragraph below Section 3 (Correlation between Annoyance and Drone Characteristics). This value (0.362) is also reported in Table 6.

Loss of control (LOC) is a prevalent cause of drone crashes. Onboard prevention systems should be designed requiring low computing power, for which data-driven techniques provide a promising solution. This study proposes the use of recurrent neural networks (RNNs) for LOC prediction. Four architectures were trained in order to identify which RNN configuration is most suitable and if this model can predict LOC for changing aerodynamic characteristics, wind conditions, quadcopter types, and LOC events. One-hundred and seventy-two real-world LOC events were conducted using a 53 g Tiny Whoop, a 73 g URUAV UZ85, and a 265 g GEPRC CineGO quadcopter. For these flights, LOC was initiated by demanding an excessive yaw rate (2000 deg/s), which provokes an unrecoverable upset and subsequent crash. All RNNs were trained using only onboard sensor measurements. It was found that the commanded rotor values provided the clearest early warning signals for LOC because these values showed saturation before LOC. Moreover, all four architectures could correctly and reliably predict the impending LOC event 2 s before it actually occurred. Furthermore, to investigate generality of the methodology, the predictors were successfully applied to flight data in which the quadcopter mass, blade diameter, and blade count were varied.

Threats posed by drones urge defence sectors worldwide to develop drone detection systems. Visible-light and infrared cameras complement other sensors in detecting and identifying drones. Application of Convolutional Neural Networks (CNNs), such as the You Only Look Once (YOLO) algorithm, are known to help detect drones in video footage captured by the cameras quickly, and to robustly differentiate drones from other flying objects such as birds, thus avoiding false positives. However, using still video frames for training the CNN may lead to low drone-background contrast when it is flying in front of clutter, and omission of useful temporal data such as the flight trajectory. This deteriorates the drone detection performance, especially when the distance to the target increases. This work proposes to pre-process the video frames using a Bio-Inspired Vision (BIV) model of insects, and to concatenate the pre-processed video frame with the still frame as input for the CNN. The BIV model uses information from preceding frames to enhance the moving target-to-background contrast and embody the target’s recent trajectory in the input frames. An open benchmark dataset containing infrared videos of small drones (< 25 kg) and other flying objects is used to train and test the proposed methodology. Results show that, at a high sensor-to-target distance, the YOLO algorithms trained on BIV-processed frames and concatenation of the BIV-processed frames with still frames increase the Average Precision (AP) to 0.92 and 0.88, respectively, compared to 0.83 when it is trained on still frames alone.

Due to technological advances in the drone industry, security threats induced by unmanned aerial vehicles (UAVs) are becoming more relevant. Fast and accurate localisation systems need to be designed. One approach is localisation of UAVs by their sound using acoustic techniques. So far, a systematic performance assessment of acoustic techniques for drone localisation, based on real-world data, is lacking. This work presents a comparison of selected techniques using real-world measurement data. The achieved performance serves as a baseline for future design of novel localisation methods. Three techniques are chosen. The first technique estimates the time-difference-of-arrival (TDOA) using generalised cross-correlation with phase transform weighting (GCC-PHAT). The second technique is differential evolution, which approaches the localisation task as a global optimisation problem. The third technique is conventional frequency domain beamforming. Real-world data of 5 quadrotor UAVs were used acquired with an acoustic microphone-array. The performance of the techniques is assessed using the absolute error between the estimated source location and the true source location obtained from the onboard GPS tracker of the drones. GCC-PHAT and differential evolution attempt to estimate the drone position in one or few steps. They have a much shorter runtime than beamforming, which is an exhaustive grid search algorithm. However, these techniques result in lower detection ranges and accuracy compared to beamforming.