Yini Okwenza Ukuthi Iteknologiya Yokuvikela Ngeendlela Zokuvikela Amadroni Ekhona Ngempilo Kukusebenza Kwesikhathi Samva?

Ukugcina Iinkokhelo Eziningi Ukufunda Ukuthi Idrone Ayisemthwakathi

Ukuthi ukuthola i-drone yokubona kuyadinga ukugcina iinkokhelo eziningi—ukubhekisa izinto ezitholakala kwi-radar, i-RF scanners, i-electro-optical/infrared (EO/IR) cameras, kunye ne-acoustic sensors ukwenza isilinganiso esisodwa sokuthola i-drone ngexesha lesinye. Lokhu kukhuphela izimbono ezingenakho ngokubhekisa iziqondiso zokuthi yiliphi iqembu: i-radar ithola ukudlula nokude, i-RF ithola izinkambo zokuxhumana, i-EO/IR inika ubuchule bokubona nokuthola okungakhangeleki, kunye ne-acoustic sensors ithola izimbono zokudlula kwemoto yokuqinisa. Umzekelo, xa i-radar ithola into ephambili ngemvelo yayo yokuqinisa ngokwe-60 kph, ukukhulu kwayo kuphela akakwazi ukwahlula phakathi kwe-drone ne-bird. Ukuthola kwe-RF ngokwesinye kwezinkambo zokuthi yiliphi iqembu (umzekelo, 2.4 GHz noma 5.8 GHz) kanti ukuhambisana kwesono sokuphathwa kwe-propeller kusiza ekubeni kubonakala ngokuphelele okungenani kwe-95%. Okulungile kakhulu, lokhu kusiza ekubeni kunezindlela eziningi zokusebenza uma isenso e yilinye iphatha—i-EO/IR ibalulekile kusasa noma kusolwandle, ngakho-ke izinto zokusho zokusho zizokwazi ukusebenza kusolwandle lapho izinto zokusho zokusho zingaphansi.

I-Radar, I-RF, I-EO/IR, ne-Acoustic Synergy ukuncima izimboni ezizingabiyo

Indima yonke yesensu isebenza ngemiphumela eyodwa. I-radar ibone amadroni okugcina kude—kufika ku-7.5 km kuma-droni weklasi 1—kepha isebenza ngendlela engeyinkimbinkimbi noma kwezindawo eziphansi kakhulu. Amasensu we-RF abonisa izinyathelo zokuphakamisa kude kwe-~3 km kepha azidinga ukubona ngokwesibonelo (line-of-sight) futhi azisewo kuma-droni owenziwe ngokuphelele ngokwesibonelo (fully autonomous drones). Amakhamera we-EO/IR abonisa ukubonwa ngokwesibonelo nokuhlanzwa kwesithunzi kude kwe-2 km, ngakho-ke amagumbi we-acoustic abalula kude kwe-~1 km futhi abhekene ngendlela enenhle kumaqondo awaphambili, amaqondo apho akukho GPS, noma kumaqondo apho ukubonwa kuyabulawa ngokwesibonelo. Ukugcina lezi zinto kusobala kunciphisa iziphuzo zezinye ngaphezu kwe-0.1%, ngakho-ke kune-~12% kuma-sistimu we-radar oyedwa. Izifundo zesistimu zokugcina eziningi—kufaka izifundo zokugcina zika-Kalman ezilungisekile kanye nezifundo zokugcina zokusebenzisa i-AI—zihlanganisa izinsu ngokudlula ngokwesibonelo: ngemvulo yomvula omkhulu, isistimu isizwa amakhamera we-EO/IR futhi isebenza ngokwesibonelo kahle kwe-radar ne-RF. Ngokuthi kubonisiwe kuma-ERDC kwezizwe zaseNingizimu Afrika, lokhu kucindezelela ukusebenza kwe-99.5% kwezizwe zaseNingizimu Afrika kwezinkinga zokugcina kwezinhlobo zokugcina kanye nezimo ezingavumelekile.

Ukuhlaziya iziDroni kusukela kwezilwane, amakhompyutha, nokuthakazelwa kweminye imithombo ekhaya lezinhloko

Izisistemi zezindikimiso eziningi zisebenzisa izinsiza zokuphathwa kanye nezinsiza zokuziphatha ngokwanele ukuqondisa izidwaba kusukela kwezinto ezingenakwazi ukuthi ziyiziphi. Ukuchaza kwe-radar kwe-micro-Doppler kubona izifreqwensi zokuzungulwa kwe-propeller—izidwaba zokugcina amasimu (quadcopters) zihlanganisa izifreqwensi zokuzungulwa zokuphakanyiswa (harmonics) ezingama-200–600 Hz, ngakumbi izinyoni zihlanganisa izinsiza zokuphakanyiswa (flapping) ezinamafreqwensi angaphezu kwe-20 Hz. Ukuvumela kwe-RF kubona izinsiza zokusebenza okufanayo nendlela yokusebenza yomthwalo, njengokuthi i-DJI™ inqubu yamafreqwensi okuzungulwa (frequency-hopping sequences) noma izithembiso zokuphatha (telemetry) esemthweleni wokuphatha (military-grade encrypted telemetry). Ukuvumela kwesound kubona izifreqwensi zokuphakanyiswa kwe-blade kanye nezinsiza zokuphakanyiswa (spectral envelopes), kubonisa izinsiza zokuphakanyiswa (harmonics) zezidwaba zokugcina amasimu (Phantom-class) kusukela kwezinye izinto zokuphakanyiswa (urban noise) noma izigwinya zokuphakanyiswa (wind gusts). Izinhlobo zezibhalo zokufunda (Neural networks) ezifundwe ngokusebenzisa izilayishini (datasets) eziphakathi kwe-NATO STO-TR-HFM-298 benchmark ziqondisa ngokuphakanyiswa ukuvumela izinto ezintsha kusukela kwezinto ezinamandla (evolving threats)—njengokuthi izinyoni ezihlanganile, izibhola zokuphakanyiswa (weather balloons), kanye nezinto zokuphakanyiswa (airborne debris). Kuma-deployments ema-urban apho izinyoni zihlola ama-65% wezilayishini zokuvumela kwe-radar, izindlela zokuvumela (fusion logic) zihlola ngokuzenzakale izinto ezingavumelekile (targets) ezinakho ukuthi zingaphansi kwe-RF telemetry noma umthwalo wokuphatha (digital command structure). Ngokufunda ngokuphakanyiswa, izidwaba ezintsha zivumelwa kanye nezivumelwa ngokuphakanyiswa ngama-72 amahour emva kokubona kokuqala—ngaphandle kokuthi kubhekwe ngokuphakanyiswa (manual model retraining).

Ukubhalwa kwezinkinga ngokusebenzisa i-AI nokuhlinzekwa kwezinkinga kuTeknoloji Yokuvikela Idrone Emkhosini

Imodeli yase-AI ephindwe kwe-edge ekwenza ukuphendula ngemizuzwana engaphezu kweyodwa

Izisistemi zesifundazane sezinkulungwane zokusebenzisa i-AI kuzenza ngokugcwele kwezinhloso zokusebenza kwe-edge hardware—njengoko kunezinhloso zika-NVIDIA Jetson AGX Orin noma izinhloso zika-Xilinx Versal ACAP—ukuprosesa idatha yezinhloso ezihlanganisiwe ngendlela eyakhiwe phakathi. Lokhu kususa umkhawulo wokusebenza kwemvelo (cloud latency) futhi kuvumela ukuthi izinqumo zizamekile ngomzuzu omncane (<300 ms end-to-end), okuyisisekelo sokubhekana nezinkinga zokusebenzisa i-FPV noma izinkinga zokusebenzisa izinhloso zokuhlanganisa (swarm-based threats). I-AI isebenza ngokufaka izindlela zokufaka ezihlanganisiwe kusuka kwezinhloso zokubheka, RF, EO/IR, nokushaywa kwemoto, ukuchaza izinto ngokubhekana nezinhloso zokudlala, ubukhulu, isigaba sesithunzi sesithunzi, nokufaka kwe-RF—ukuhlanganisa izinhloso zokusebenzisa i-quadcopter zokuziphatha ngokwenyama kumaqhudelwane othile wezinkulungwane zokubheka noma izinhloso zokuhamba kwezinyoni. Iziqu zokudlala zokubhekana zibonisa izinkulungwane ezingenayo—ukuhlala emhlabeni owenzekile ngokungafanele, ukuguqula ukuphakamisa okuphakade, noma ukuhlanganisa ngokuphakade kwezinhloso zokuhlanganisa—futhi zinika isikhalo sesikhundla sokubhekana esihlanganisiwe. Ukufunda ngokuphakade kwe-online kusobala imodeli ngokushesha kusasa kubonakala kwezinhloso zokusebenzisa izinhloso ezintsha, ukufaka ukubuyekeza kuka-operators nokubuyekeza kwezindaba zokuphinda kusasa. Ukuphakanyiswa kwezifundazane kusobala kwezinhloso zokusebenzisa i-U.S. SOCOM’s Project Convergence 2023 kuthi ukuchaza kwezinhloso zokusebenzisa i-AI kwe-edge kubeka ubuncinane boka-operators ngama-70% futhi kubeka umkhawulo wokusebenza ngama-4.2× ngaphezu kwezinhloso zokusebenzisa izinhloso zokubhekana zokuziphatha ngokwenyama.

Ukuchaza Okwaziwe Ngokwemithambo Yokuqala Kwemithambo Yokugcina Amadroni Okwamanje

Ngaphandle kokuhlukaniswa ngokomzimba, i-AI yenza ukuhlaziywa okujulile, okuqondayo kwe-protocol kwe-command-and-control (C2) traffic kufaka phakathi i-Wi-Fi, i-LTE / 5G, kanye nama-protocol omsakazo njenge-OcuSync noma i-Light Isebenzisa izinjini ezincane zokukhipha amaphakethe ezisebenza kuma-CPG e-FPGA, uhlelo luchaza isikhathi sokushaya izandla, isakhiwo somthwalo, nokuziphatha kokumodula ngesikhathi sangempela. Ihlobanisa okutholakele nemitapo yolwazi yezinsongo egunyaziwe egcinwa yiNational Cybersecurity Center of Excellence (NCCoE) kanye namathuluzi okudlulisa ulwazi omthombo ovulekile afana ne-DroneDB. Lokhu kwenza kube nokwenzeka ukuhlukanisa ngokunembile: ukuhlukanisa ukuhlolwa kwezindiza zobungane nokuhlola okuphikisana ngokusekelwe okhiye bokubethela, isikhathi seseshini, kanye ne-entropy yesiteshi sokulawula. Uhlelo luphinde lubike ukuziphatha okuvimbela ukuphazamiseka ukugxuma kwemvamisa, ukudluliswa kwe-spread-spectrum, noma ukucindezelwa kwe-beacon okuhlobene kakhulu nenhloso enonya ngokwe-DoD Directive 3000.09. I-protocol telemetry ingena ngqo enjini yokulinganisa izinsongo, ikhulise ukuzethemba kwama-drone abonisa ukusakazwa kwevidiyo + i-C2 + geofence overridesignature indicators of malicious payloads. Lolu ungqimba lunciphisa ukuxhomeka ekuqapheni kwesigcawu somsakazo ngesandla futhi lwenza ukuhlonza okuzenzakalelayo, okuvikeleka ngokomthetho okuhambisana nenqubomgomo ye-DoD's Electronic Warfare Execution Policy (EWP).

Ukubalansisa Kwezindlela Zokunqoba Ngokuphelele: Ukubalansisa Izindlela Zokunqoba Ezincane Nezinkulu

Ukubhekisana Kwemithambo Kuyaphatha izindlela zokunqoba ezincane nezinkulu ukuze zihambisane nehlobo lezinkinga, isikhundla, nokuthakazeka kwezinhloso—ukuhlonipha ukunqoba okuphelele ngokwesibonelo esifanayo ngaphandle kokukhula umsindo wokusebenza noma ukugcina ukuzimisela kwemithambo yezomnotho.

Ukunqoba Ngokwesibonelo Noma Ngaphandle Kwesibonelo Ngemizindvo YaseNtloko Noma Emizindvini Yokugcinwa Kwemithambo Yekhulekhole (EM)

Abahlawuli baya kufanele bakhokhise indlela yokunqoba nesikhundla, ubude bewesiqondiso, nokugcinwa kwemithambo yekhulekhole (EM):

• Ukunqoba ngokwesibonelo —kungahlanganisa izidrano zokunqoba, izinqamuzi zokubhala amanethi, noma izinkampani zokuphatha ngokwesibonelo—zisiza ekunqobeni kahle ngaphandle kokuthi izidrano ziyasebenza ngokuphelele noma zinazo izindlela zokuvikela phambi kokuvikela. Kodwa, emizindvini yaseNtloko, isibonelo sokususa kuyabika imiphumela eyodwa: I-Ponemon Institute (2023) iyathola ukuthi isilinganiso sesilinganiso sokugcinwa kwezinto ezingavumelekileyo kuyisikhundla sokunqoba ngokwesibonelo kuyisikhundla sokugcinwa ngokuphelele $740K.
• Ukunqoba ngaphandle kwesibonelo , njengoko kubhekwa kwe-RF noma ukufaka isithembiso se-GNSS, kusinda ukuxhumana noma ukuphatha ukuthambeka ngokungafanele kohlobo lwesibonelo—kuyisifundo sokulondoloza abantu basebenzi, izinkampani, noma izindawo zezopolitiki eziphakeme. Umphumela wakho uyisiphelo esingaphansi kwezinkinga zokusebenza kuzo izi-drone ezihlukene ngokuphelele ezisebenzisa imishini yazo eyodwa ngaphandle kokuxhumana nge-C2 ngokushesha.

Isimo sesikhundla sokuphendula—sokwamukwa ngalesi 2024 ngaleli raporti ya RAND Corporation Ukuphendula Izithembiso Zokuhamba Ngokuphelele Kulezinyawo —sikhipha ukuphendula ngokungafanele njengendlela yokuphendula okuphakeme, kusika izinketho zokuphendula ngokugcina kwezinkampani eziphakeme noma izindlela ezinokuthi ukuphendula ngokungafanele akwazi ukwenza (isibonelo, izi-drone ze-FPV ezisebenzisa izilayini zokubhala ngokungafanele esizenzileyo esingasenzi ukuthi kubhekwe ngokungafanele). Ukusebenzisa okuphelele kufuneka kubhekwe ngokuphathelene nokubhekwa kwezinhlobo zemvelo zomoya emthwakathi wokuxhumana—ukubona izibandlana eziphakeme ezisebenzisa izinkampani zokulungisa noma ukuphatha ukuthambeka ukuze ukungabhekwe ngokungafanele futhi ukubona izindlela ezinokuthi zisebenze ngokuphelele ekuphenduleni ngokungafanele.

Ukuxhumana Nokuphatha Ngokuphelele Ukuphatha Okuphathelene Nokuxhumana Ngokushesha

Itheknoloji yomkhosi wokuphatha amadroni yasezinsizini iyisisekelo sezinkundla zokuphatha nokubambisana (C2) esingathandabuzi—esakhiwe ukubambisana ukuthi kufunyanwa, kubambisana nokulawula ngokwesibonelo esihlukene. Kuyakhiwe emisebenzini eyodwa (MOSA, STANAG 4586, no-IEEE 1394.2), izinkundla zokuphatha zama-C2 zilanda nokubambisana ngezikhathi izindlela zokufunda ezifana ne-radar, RF, EO/IR, kunye nezindlela zokufunda ngokusha, zithuthukisa isibonelo esinye sokuhleli kwesikhaleni. Abasebenzi bafuna ukuzinikela ngokushesha kwesimo sokuhleli, ukubhekisa okuthile okungahleli ngokushesha, nokubambisana ngokuzenzakalelayo kwezinhlayo zokuphatha—okuthi kubhekiswe ukuphatha okungakhulu kumaqondiso awaphansi okungahleli, noma kubhekiswe ukuphatha okuphambili uma kubhekiswe ngokwenyama noma ingxenye yokusebenza. Ngokubambisana kwezinhlayo zokuphatha ngendlela enye yendlela yokusebenza, isistimu isula izinhlayo zokuphatha ezingasombululi futhi isula ukuphatha okungafanani (isibhalo: ukuphatha ngokusha ngokufanana nokuphatha kwe-GNSS). Ngokubonwa kwezindlela zokuphatha zasezinsizini zonke (JADC2), isistimu sokubambisana sokuphatha sishintshisa isikhathi sokuphatha kusukela ku-12 sekondi ku-2.5 sekondi—futhi sisebenza ngokuphelele nakuba kunezinhlayo zokufunda ezimbili ezinjalo. Isiphumo kukhona isistimu esihleli, esihlanganisiwe, futhi esihloniphi umuntu okuphatha, esinokuthintwa ngokushesha ngokwesikhathi sokuhleli kwesikhaleni esisha.

Isikhombisi Sokuqala

Ukuthi ukugcina izinhlobo eziningi zemithambo kuyini ngokugcina amadroni?

Ukugcina izinhlobo eziningi zemithambo kuchazela imininingwane emithambeni emibili njenge-radar, RF, EO/IR, nezinhlobo zokusho isono ukuze kubonekwe indlela eyodwa yokuqonda amadroni ngokushesha kanye nokuphepha kakhulu. Lokhu kunciphisa izinkinga zokubona okungasemthwakweni futhi kubhekise ubukhulu bokuphepha.

Ukuthi ukuthintana kwe-AI kubhekise indlela yokugcina amadroni kuyini?

Ukuthintana kwe-AI kuchazela ngokushesha umsebenzo wamadroni, ubukhulu bawo, imisebenzi yabo, nokuthintana kwezinhlobo zokubeka izinkulumo ukuze kubhekiswe nokuhlanganiswa izinkinga. Lokhu kunciphisa umsebenzi wobunyazi futhi kubhekise ukuthi kuzothathelwa isikhathi esincane sokuthabatha isibonelo.

Izindlela zokugcina ezingenzi isifo noma ezenzi isifo ziyini?

Izindlela zokugcina ezingenzi isifo zivame ukuthintana ngokungafani nge-RF jamming noma GNSS spoofing, ngakolunye uhlangothi izindlela zokugcina ezenzi isifo zivame ukusebenzisa izinhlobo zokuthintana zokuphinda phansi noma izinhlobo zokuthintana zokuphinda phansi zokuthintana ngokuphinda phansi ukuze kugcine amadroni ngokuphinda phansi.

Ukuthi ukugcina okuphakathi kufuna ukuncipha ukuthwasa kweminye imithwakuli njengemizini kuyini?

Amathuluzi okugcina amasiko akhona ngokuthula ukukhathazeka kwezinto zokugcina ngokuphatha ngokuphelele futhi abuyisela izinkokhelo zokugcina ngokuphatha ngokuphelele kubizo labo obizwa ngokuthi bingcono kakhulu noma imikhombiso ekufuneka kuzwakalisa ngokuphelele.

Kungani isikhalazo sokugcina nokumemeza kusobala?

Izinsiza zokugcina nokumemeza zokugcina zihlanganisa ukubonwa, ukujabulisa, nokugcina ngokuphatha ngokuphelele phakathi kwamaphethini amaningi, ekuvumelana ngokushesha kanye nokuhlanganisa imiphumela ngokuphatha ngokuphelele nokunciphisa iziphulo.