The commercial and urban application of UAVs has increased over the last years. As many more potential applications will likely emerge in the future, minimizing the impact of the use of UAVs on its environment is important. With noise emission being one of these challenges, acoustic analysis on UAVs has become more relevant [1, 2]. With the goal of continuously developing quieter UAVs for urban and commercial applications, the mechanism behind noise must be understood. The localization and quantification of the spatial and temporal characteristics of noise will provide insight into sources and the mechanisms behind the generation. Furthermore, we have been witnessing the emerging of novel UAV designs, for instance the ones in [3, 4], aimed at creating robots that can safely and effectively interact with the environment and with humans, thanks to their higher manoeuvrability compared to standard underactuated quadrotors. The acoustic image of these platforms has the potential to become a key design factor.
Acoustic imagery using tomographic reconstruction is a novel approach for the localization and quantification of noise sources in 3D [5]. Using experimental data from a phased microphone array and beamforming techniques [6], noise sources can be identified, analyzed, and localized. However, acoustic tomographic reconstruction is computationally expensive. This computational constraint limits the applicability to, e.g., wind tunnel tests where several large datasets need to be processed within a relatively short amount of time.
With the application of AI emerging in several similar fields, such as computational fluid dynamics and dynamic mesh solving, the question raises whether data-driven techniques could be a useful tool for acoustic localization, especially acoustic tomographic reconstruction [7]. Resolving the computational bottleneck that is present in acoustic imagery using tomographic reconstruction, will allow significant acceleration of analysis and has the additional potential of improving the current spatial resolution limitation of two-dimensional acoustic imaging.
The main research questions are formulated as follows:
-Can machine learning methods be used for acoustic tomographic reconstruction of noise measurements of UAVs?
-Can this approach improve the capabilities to more accurately localize and quantify differently configured drones and their different components?
The following key aspects will be treated in this research.
-Understanding the aeroacoustics of different types of UAVs and their components.
-Investigate the best machine learning approaches to apply to acoustic tomographic reconstruction.
-Creating a framework for large-scale numerical datasets of microphone phased array measurements to apply machine learning.
-Validating the developed machine learning tool with noise measurements of UAVs and/or UAV components in a wind tunnel test.
References
[1] Cussen, K., Garruccio, S., & Kennedy, J. (2022, March). UAV noise emission—A combined experimental and numerical assessment. In Acoustics (Vol. 4, No. 2, pp. 297-312). MDPI.
[2] Pang, E., Cambray, A., Rezgui, D., Azarpeyvand, M., & Showkat Ali, S. A. (2018). Investigation towards a better understanding of noise generation from UAV propellers. In 2018 AIAA/CEAS Aeroacoustics Conference (p. 3450).
[3] Aboudorra, Y., Gabellieri, C., Brantjes, R., Sablé, Q., & Franchi, A. (2024). Modelling, Analysis, and Control of OmniMorph: an Omnidirectional Morphing Multi-rotor UAV. Journal of Intelligent & Robotic Systems, 110(1), 21.
[4] Afifi, A., Corsini, G., Sable, Q., Aboudorra, Y., Sidobre, D., & Franchi, A. (2023, June). Physical human-aerial robot interaction and collaboration: Exploratory results and lessons learned. In 2023 International Conference on Unmanned Aircraft Systems (ICUAS) (pp. 956-962). IEEE.
[5] Tuinstra, M., & van der Meulen, M. J. (2019). Acoustic location by tomographic reconstruction. In 25th AIAA/CEAS Aeroacoustics Conference (p. 2409).
[6] Sijtsma, P. (2010). Phased array beamforming applied to wind tunnel and fly-over tests.
[7] Kujawski, A., Pelling, A. J., Jekosch, S., & Sarradj, E. (2023). A framework for generating large-scale microphone array data for machine learning. Multimedia Tools and Applications, 1-21.