Optical detection of promising landing sites for helicopter-type unmanned aerial vehicle using kohonen self-organizing MAPS

Machine-building Engineering and Machine Science


DOI: 10.34759/vst-2022-3-209-221

Аuthors

Al'khanov D. S.1*, Kuzurman V. A.**, Gogolev A. A.2***

1. Bauman Moscow State Technical University, MSTU, 5, bldg. 1, 2-nd Baumanskaya str., Moscow, 105005, Russia
2. Moscow Aviation Institute (National Research University), 4, Volokolamskoe shosse, Moscow, А-80, GSP-3, 125993, Russia

*e-mail: daniil.alhanov@yandex.ru
**e-mail: kuzurmanva@student.bmstu.ru
***e-mail: kirbizz8@yandex.ru

Abstract

The subject of the article being presented is a helicopter-type unmanned aerial vehicle (UAV) with a coaxial rotors design. The research issue is landing procedures automation on a site unprepared with respect to engineering. The purpose of the work consists in developing a set of basic requirements for an air-defined landing site based on current aviation standards, as well as implementing neural network classifier of the underlying surface. The authors considered the existing methods of landing performing for the UAV. As the result of the analysis, the method of autonomy enhancing by implementation of information systems and sensors of various operation principles was defined as the most promising. With account for acting Federal Aviation Regulations (FAR), as well as norms adopted by the International Civil Aviation Organization (ICAO) and European Aviation Safety Agency (EASA), the list of requirements for the prospective landing zone characteristics, accounting for the specifics of the UAV studied in the work, was developed. The main complexity here consists in the lack of the standardized regulations of performing landing procedures for the UAVs of this weight class of 325 kilos. The review of the conventional methods for the underlying surface quality determining was conducted. By reason of small overall sizes of the aerial vehicle being studied, meso- and micro-relief of the terrain are of special interest. The authors decided to split the algorithm for appropriate landing site determining into the two logical stages. Optical survey of the terrain and determination of several optimal prospective landing zones based on color semantics, characteristic structure patterns, presence of obstacles and proximity of the terrain regions transition are being executed at the first stage. Next, the descent to the most optimal site to the altitude exceeding the critical decision point is being performed, and relief scanning by the compensated laser-radar system is being executed to obtain the relief model and determine the soil characteristics. Both technique and software development was being performed in the course of this work for the first stage of the underlying surface primary inspection. The main problem of the video fixation cameras application onboard of aerial vehicles consists in strong dependence of the obtained data processing results on the environment state. Variability of both weather conditions and Earth surface lighting conditions may exert drastic parasitic effect the result of the algorithm execution. Various methods of preliminary image processing, such as contrast ratio improving, segmentation and noise filtering, allow partially solving this problem. However, the greatest invariance to the shooting conditions can be achieved using neural network methods for image analysis. The authors proposed an optical recognition method of the prospective landing zones employing self-organizing Kohonen maps. The neural networks of this kind advantage is the simplicity of the training sample preparing, as well as simplicity of the synoptic weights distribution process in the course of the casual observer training. The selected approach allows evaluating not only the color specters distribution on the image, bug tracking characteristic patterns of the texture as well. The training sample contained 2700 fragments of the terrain topographic snapshots, and the neural network training time was 10,000 epochs. Computer tests revealed 21% of the alpha errors and 0% of the beta errors, which is specific for the neural networks of this class as well. The results obtained in the course of this work are simultaneously indicative of this approach exploitability to the underlying surface clustering and the need for further research on the considered issue.

Keywords:

unmanned aerial vehicle, landing procedures automation, underlying surface analysis, Kohonen self-organising maps

References

  1. Byun J.-H., Lee G.-M., Kim S.-H. Design of Experiments for Optimization of Helicopter Flight Tests. Transactions of the KSME C Industrial Technology and Innovation, The Korean Society of Mechanical Engineers, 2014, vol. 2, no. 2, pp. 113-124. DOI: 10.3795/KSME-C.2014.2.2.113

  2. Gulevich S.P. Letnye ispytaniya skorostnykh bespilotnykh letatel’nykh apparatov (Flight tests of high-speed unmanned aerial vehicles), 2nd ed, Moscow, Pervyi tom, 2018, 288 p.

  3. McAteer T., Rice C., Gavin C. Flight test safety – The U.S. navy approach. IEEE Aerospace Conference (28 June 2018; Big Sky, MT, USA). DOI: 10.1109/AERO.2018.8396745

  4. Gregory J.W., Liu T. Introduction to Flight Testing (Aerospace Series). John Wiley & Sons Ltd, 2021, 352 p.

  5. Costello D.H., Jewell J., Xu H. Autonomous Flight-Test Data in Support of Safety of Flight Certification. Aerospace Research Central, 2020, vol. 29, no. 2. DOI: 10.2514/1.D0220

  6. Hallberg E., Kaminer I., Pascoal A. Development of a flight test system for unmanned air vehicles. IEEE Control Systems Magazine, 1999, vol. 19, no. 1, pp. 55-56. DOI: 10.1109/37.745769

  7. Alaimo A., Esposito A., Orlando C., Tesoriere G. A Pilot Mental Workload Case Study in a Full Flight Simulator. Aerotecnica Missili & Spazio, 2018, vol. 97, pp. 27–33.

  8. Kantowitz B.H., Campbell J.L. Pilot Workload and Flightdeck Automation. In: Automation and Human Performance: Theory and Applications. CRC Press, 2019, 20 p.

  9. Tang L., Si J., Sun L., Mao G, Yu S. Assessment of the mental workload of trainee pilots of remotely operated aircraft using functional near-infrared spectroscopy. BMC Neurol, 2022, no. 1: 160. DOI: 10.1186/s12883-022-02683-5

  10. Evdokimenkov V.N., Kim R.V., Vekshina A.B., Yakimenko V.A. Study of pilot’s control actions personality during landing based on neural network models. Aerospace MAI Journal, 2015, vol. 22, no. 3, pp. 17-29.

  11. Semkin V., Huhtinen I. Millimeter-wave UAV-based Channel Measurement Setup. IEEE 93rd Vehicular Technology Conference (25-28 April 2021; Helsinki, Finland). DOI: 10.1109/VTC2021-Spring51267.2021.9448737

  12. Miao Y., Fan W., Rodríguez-Piñeiro J., Yin X., Gong Y. Emulating UAV Air-to-Ground Radio Channel in Multi-Probe Anechoic Chamber. IEEE Globecom Workshops (09-13 December 2018; Dhabi, United Arab Emirates). DOI: 10.1109/GLOCOMW.2018.8644381

  13. Burhanuddin L.A., Liu X., Deng Y., Challita U., Zahemszky A. QoE Optimization for Live Video Streaming in UAV-to-UAV Communications via Deep Reinforcement Learning. IEEE Transactions on Vehicular Technology, 2022, vol. 71, no.5, pp. 5358 – 5370. DOI: 10.1109/TVT.2022.3152146

  14. Zhou H., Hu F., Juras M., Mehta A.B., Deng Y. Real-Time Video Streaming and Control of Cellular-Connected UAV System: Prototype and Performance Evaluation. IEEE Wireless Communications Letters, 2021, vol. 10, no. 8, pp. 1657–1661. DOI: 10.48550/arXiv.2101.10736

  15. Pogosyan M.A., Vereikin A.A. Position and motion control of aerial vehicles in automatic landing systems: analytical review. Aerospace MAI Journal, 2020, vol. 27, no. 3, pp. 7-22. DOI: 10.34759/vst-2020-3-7-22

  16. Ivanov P.I., Kurinnyi S.M., Krivorotov M.M. Parameters liable to be defined while a multi-dome parachute system flight-testing for its efficiency estimation. Aerospace MAI Journal, 2020, vol. 27, no. 3, pp. 49-59. DOI: 10.34759/vst-2020-3-49-59

  17. Nikolaev E.I., Nedelko D.V., Shuvalov V.A., Yugai P.V. External airbags application onboard a helicopter. Aerospace MAI Journal, 2019, vol. 26, no. 3, pp. 91-101.

  18. Zhurin S.V. Parachute-jet soft landing system with elastic linkage. Aerospace MAI Journal, 2016, vol. 23, no. 1, pp. 107-114.

  19. Grlj C.G., Krznar N., Pranjić M. A Decade of UAV Docking Stations: A Brief Overview of Mobile and Fixed Landing Platforms. Drones, 2022, vol. 6, no. 1: 17. DOI: 10.3390/drones6010017

  20. Bagare S.V., Mirza K., Sharma M. et al. Design of Mobile Docking Mechanism for Unmanned Aerial Vehicles capable of Vertical Take-off and Landing. AIAA Aviation Forum, 2022. DOI: 10.2514/6.2022-4063

  21. Stewart W., Ajanic E., Müller M., Floreano D. How to Swoop and Grasp Like a Bird With a Passive Claw for a High-Speed Grasping. IEEE/ASME Transactions on Mechatronics, 2022. DOI: 10.1109/TMECH.2022.3143095

  22. Kramar V., Kabanov A., Dudnikov S. A Mathematical Model for a Conceptual Design and Analyses of UAV Stabilization Systems. Fluids, 2021, vol. 6., no. 5: 172. DOI: 10.3390/fluids6050172

  23. Theodore C., Rowley D., Hubbard D. et al. Flight Trials of a Rotorcraft Unmanned Aerial Vehicle Landing Autonomously at Unprepared Sites. San Jose State Foundation, Jet Propulsion Laboratory, US Army AFDD, 2006.

  24. Takahashi M., Abershitz A., Rubinets R., Whalley M. Evaluation of safe landing area determination algorithms for autonomous rotorcraft using site benchmarking. 67th American Helicopter Society International annual forum (03-05 May 2011; Virginia Beach, Virginia, USA).

  25. Sanfourche M., le Besnerais G., Fabiani P., Piquereau A., Whalley M. Comparison of terrain characterization methods for autonomous UAVs. 65th American Helicopter Society International annual forum (27-29 May 2009, Grapevine, Texas, USA).

  26. Krizhevsky S., Sutskever I., Hinton G.E. ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems. 2012.

  27. Lunev E.M. Improving the accuracy of uav navigation parameters during landing on the basis of photogrammetric measurements. Aerospace MAI Journal, 2011, vol. 18, no. 2, pp. 150-159.

  28. Merkishin G.V., Slivin I.V. Optoelectronic altimeter for aircraft landing system. Aerospace MAI Journal, 2001, vol. 8, no. 2, pp. 75-82.

  29. Garcia-Pardoa P.J., Sukhatmeb G.S., Montgomery J.F. Towards vision-based safe landing for an autonomous helicopter. Robotics and Autonomous Systems, 2002, vol. 38, no. 1, pp. 19-29. DOI: 10.1016/S0921-8890(01)00166-X

  30. Federal’nye aviatsionnye pravila “Trebovaniya k posadochnym ploshchadkam, raspolozhennym na uchastke zemli ili akvatorii” (Federal Aviation Regulations “Requirements for landing sites located on a plot of land or water area”), 2011.

  31. Normy letnoi godnosti grazhdanskikh vertoletov SSSR. NLGV-2 (Airworthiness Standards of the USSR civil helicopters. NLGV-2). Moscow, 1987, 410 p.

  32. European Union Aviation Safety Agency, Certification Specifications and Guidance Material for the design of surface-level VFR heliports located at aerodromes that fall under the scope of Regulation (EU) 2018/1139 (CS-HPT-DSN). 2019.

  33. International Civil Aviation Organization, Physical Characteristics  Surface-level Heliports. ICAO Regional Workshop. 2016.

  34. Nath S.S., Mishra G., Kar J. et al. A survey of image classification methods and techniques. International Conference on Control, Instrumentation, Communication and Computational Technologies (10-11 July 2014; Kanyakumari, India). DOI: 10.1109/ICCICCT.2014.6993023

  35. Gonzalez R.C., Woods R.E. Digital Image Processing. 4th edition. Pearson, 2017, 1192 p.

  36. Tajbakhsh N., Shin J.Y., Gurudu S.R. et al. Convolutional neural networks for medical image analysis: Full training or fine tuning? IEEE Transactions on Medical Imaging, 2016, vol. 35, no. 5, pp. 1299-1312. DOI: 10.48550/arXiv.1706.00712

  37. Zhao Z., Ma Q. A novel method for image clustering. 10th International Conference on Natural Computation (19-21 August 2014; Xiamen, China), pp. 648-652.

  38. Bernard Y., Hueber N., Girau B. Novelty Detection with Self-Organizing Maps for Autonomous Extraction of Salient Tracking Features. Advances in Self-Organizing Maps, Learning Vector Quantization, Clustering and Data Visualization, 2019, pp 100–109.

  39. Malone J., Mcgarry K., Wermter S., Bowerman C. Data mining using rule extraction from Kohonen self-organising maps. Neural Computing and Applications,. 2006, vol. 15, no. 1, pp. 9-17. DOI: 10.1007/s00521-005-0002-1

  40. Kabra A., Iwahori Y., Usami H. et al. Polyp Classification and Clustering from Endoscopic Images using Competitive and Convolutional Neural Networks. 8th International Conference on Pattern Recognition Applications and Methods (19-21 February 2019, Prague, Czech Republic), pp. 446-452. DOI: 10.5220/0007353204460452

  41. Kohonen T. Essentials of the self-organizing map. Neural networks, 2013, vol. 37, pp. 52-65. DOI: 10.1016/j.neunet.2012.09.018

  42. Parfent’ev K.V. Molodezhnyi nauchno-tekhnicheskii vestnik, 2013, no. 11, pp. 34.

  43. Gavrilov A.I., Parfent’ev K.V. Materialy XLI Akademicheskie chteniya po kosmonavtike (24–2727 yanvarya 2017; Moskva), p. 413

mai.ru — informational site of MAI

Copyright © 1994-2024 by MAI