1 Gabatarwa
Ra'ayin lissafi ko'ina (pervasive computing), wanda W. Mark ya gabatar a shekara ta 1999, da kuma Abubuwan Intanet (IoT), wanda Kevin Ashton ya ƙirƙira a wannan shekarar, sun sami ci gaba sosai. IoT tana haɗa abubuwa na zahiri zuwa Intanet don yin mu'amala da yanke shawara ta atomatik. Duk da haka, na'urorin IoT sau da yawa suna da ƙarancin albarkatun lissafi da makamashi, wanda ke sa sarrafa abubuwa masu sarkakiya ya zama ƙalubale. Lissafi a bakin cibiyar sadarwa (Edge Computing) ya fito a matsayin mafita ta hanyar kawo lissafi da ajiyar bayanai kusa da tushen bayanan, yana rage jinkiri da amfani da ƙarfin watsa bayanai. Kasuwar lissafi a bakin cibiyar sadarwa a duniya an kimanta ta da dalar Amurka biliyan 11.24 a shekara ta 2022 kuma ana hasashen za ta yi girma da matsakaicin ci gawa na shekara-shekara (CAGR) na 37.9% daga 2023 zuwa 2030.
2 Tsarin Lissafi don IoT
Akwai tsare-tsare da yawa na lissafi waɗanda ke tallafawa aikace-aikacen IoT, kowannensu yana da halaye da amfani daban-daban.
2.1 Lissafi a Girgije (Cloud Computing)
Sarrafawa ta tsakiya a cikin cibiyoyin bayanai masu nisa. Yana ba da albarkatu masu yawa amma yana haifar da jinkiri ga aikace-aikacen IoT masu mahimmanci na lokaci.
2.2 Lissafi a Hazo (Fog Computing)
Yana faɗaɗa iyawar girgije zuwa bakin hanyar sadarwa, yana ƙirƙira wani yanki tsakanin na'urorin IoT da girgije. Yana ba da matsakaicin sarrafawa da ajiya.
2.3 Lissafi a Bakin Cibiyar Sadarwa (Edge Computing)
Yana tura lissafi da ajiyar bayanai zuwa ƙarshen hanyar sadarwa, wato akan ko kusa da na'urorin IoT da kansu. Yana rage jinkiri sosai kuma ya dace da sarrafa bayanai cikin gaggawa.
Hangar Kasuwa
Kasuwar Lissafi a Bakin Cibiyar Sadarwa a Duniya (2022): Dalar Amurka Biliyan 11.24
Hasashen Matsakaicin Ci Gawa na Shekara-Shekara (CAGR) (2023-2030): 37.9%
Tushe: Hasashen binciken kasuwa da aka ambata a cikin daftarin.
3 Tsarin Aikin Lissafi a Bakin Cibiyar Sadarwa
3.1 Lissafi a 'Cloudlet'
'Cloudlets' ƙananan cibiyoyin bayanai ne na gida waɗanda aka sanya su a bakin hanyar sadarwa, sau da yawa kusa da masu amfani (misali, a cikin gini ko harabar jami'a). Suna ba da albarkatun lissafi masu ƙarfi tare da ƙarancin jinkiri fiye da girgije masu nisa, suna aiki a matsayin masu tsaka-tsaki don sauke ayyuka daga na'urorin wayar hannu/IoT masu ƙarancin albarkatu.
3.2 Lissafi a Bakin Cibiyar Sadarwar Wayar Hannu (MEC)
MEC, wanda yanzu ake kira da Lissafi a Bakin Cibiyar Sadarwa ta Hanyoyi Daban-daban (Multi-access Edge Computing), yana haɗa albarkatun lissafi kai tsaye cikin hanyar sadarwar samun rediyo (RAN), kamar a tashoshin tushen wayar salula. Wannan tsari yana da mahimmanci ga hanyoyin sadarwa na 5G, yana ba da damar aikace-aikace masu ƙarancin jinkiri kamar motocin da ke gudanar da kansu da kuma haɓaka gaskiya (augmented reality).
4 Tsarin Gine-ginen IoT Wanda Ya Dogara akan Lissafi a Bakin Cibiyar Sadarwa
4.1 Tsarin Gine-gine Mai Matakai Uku
Tsarin gine-gine na yau da kullun ya ƙunshi:
- Yanki na Na'ura/Hankali: Ya ƙunshi na'urori masu auna yanayi (sensors), na'urori masu motsawa (actuators), da na'urorin IoT waɗanda ke tattara bayanai.
- Yanki na Bakin Cibiyar Sadarwa (Edge Layer): Ya haɗa da nodes na bakin cibiyar sadarwa (ƙofofin shiga (gateways), sabobin aiki (servers), 'cloudlets') waɗanda ke aiwatar da sarrafa bayanai na gida, tacewa, da bincike.
- Yanki na Girgije (Cloud Layer): Girgije na tsakiya don bincike mai nauyi, ajiyar dogon lokaci, da gudanarwa na duniya.
4.2 Manyan Fa'idodi
- Rage Jinkiri: Sarrafawa na gida yana kawar da tafiya zuwa girgije mai nisa.
- Ingantacciyar Amfani da Ƙarfin Watsa Bayanai: Ana aika bayanan da suka dace ko waɗanda aka tattara kawai zuwa girgije.
- Ƙarfafa Sirri & Tsaro: Ana iya sarrafa bayanai masu mahimmanci a gida.
- Ingantacciyar Amincewa: Yana aiki da kansa ɗan lokaci yayin matsalolin haɗin kai da girgije.
5 Fasahohin da ke Ba da Dama
5.1 Hankalin Wucin Gadi a Bakin Cibiyar Sadarwa
Gudanar da samfuran AI (misali, don gano abin da bai dace ba, kulawa na gaba, hangen nesa na kwamfuta) kai tsaye akan na'urorin bakin cibiyar sadarwa. Wannan yana buƙatar dabarun inganta samfura kamar yanke reshe (pruning), ƙididdigewa (quantization), da narkar da ilimi (knowledge distillation) don dacewa da ƙuntatawar albarkatu. Ana iya wakiltar tsarin ƙididdigewa (inference) kamar nemo $y = f_{\theta}(x)$, inda $f_{\theta}$ ke wakiltar samfurin da ba shi da nauyi da aka tura zuwa bakin cibiyar sadarwa.
5.2 Sauƙaƙan Tsarin 'Virtualization'
Fasahohi kamar kwantena na Docker da unikernels suna ba da muhallin aikace-aikace masu keɓewa, masu ɗaukar kaya tare da ƙaramin nauyi idan aka kwatanta da na'urori masu kama da na'ura ta zahiri (VMs), wanda ya sa su dace da tura ƙananan ayyuka (microservices) akan nodes na bakin cibiyar sadarwa.
6 Nazarin Lamura & Aikace-aikace
6.1 Lafiya
Kula da marasa lafiya cikin gaggawa ta hanyar na'urori masu auna yanayi da ake sawa. Nodes na bakin cibiyar sadarwa suna bincika alamun rayuwa (yawan bugun zuciya, SpO2) a gida don kunna faɗakarwa nan da nan don yanayi mai mahimmanci, suna tabbatar da shiga tsakani cikin lokaci yayin aika taƙaitaccen rahoto zuwa girgije.
6.2 Masana'antu
Kulawa na gaba a masana'antun masu hankali. Na'urori masu auna girgizawa da zafin jiki akan injina suna aika bayanai zuwa ƙofar shiga ta bakin cibiyar sadarwa. Samfuran AI na gida suna hasashen gazawar kayan aiki, suna ba da damar kulawa kafin lalacewa ta faru, suna rage lokacin da aikin ya tsaya.
6.3 Noma
Noma mai daidaito ta amfani da na'urori masu auna yanayi na IoT don danshin ƙasa, zafin jiki, da lafiyar amfanin gona. Na'urorin bakin cibiyar sadarwa suna sarrafa waɗannan bayanan don sarrafa tsarin ban ruwa da kansu kuma cikin gaggawa, suna inganta amfani da ruwa.
6.4 Sufuri
Motocin da ke gudanar da kansu da kuma sarrafa zirga-zirga. Motoci suna amfani da lissafi a bakin cibiyar sadarwa da ke cikin su don sarrafa bayanan LiDAR da na'urar daukar hoto don yanke shawara nan da nan game da hanyar tafiya, yayin da sabobin aiki na bakin cibiyar sadarwa a mahadar hanyoyi ke inganta tsarin fitilun zirga-zirga bisa gudun zirga-zirga na gaskiya.
7 Ƙalubalen Bincike & Hanyoyin Gaba
Ƙalubale: Daidaita tsarin gine-ginen bakin cibiyar sadarwa, tsaron nodes masu rarrabawa, ingantaccen sarrafa albarkatu a cikin na'urori iri-iri, da kuma gudanar da bayanai a cikin muhallin masu ruwa da tsaki da yawa.
Hanyoyin Gaba: Haɗawa da hanyoyin sadarwa na 6G, ci gaba a cikin AI na asali na bakin cibiyar sadarwa (misali, koyon haɗin gwiwa a bakin cibiyar sadarwa), haɓaka dandamali masu ƙwarewa (kamar KubeEdge), da bincika lissafi a bakin cibiyar sadarwa don metaverse da tagwayen dijital.
8 Nazarin Fasaha & Fahimta
Ra'ayin Manazarcin: Rarraba Haɗin Kai na Edge-IoT
Babban Fahimta: Wannan daftarin yana sanya lissafi a bakin cibiyar sadarwa ba kawai a matsayin reshen fasaha na girgije ba, amma a matsayin gyaran gine-gine da ya wajaba don sabani na girman IoT. Tsarin girgije na tsakiya, duk da cewa yana da ƙarfi, yana haifar da toshewa na asali ga aikace-aikacen IoT masu mahimmanci na jinkiri, masu ƙwazo na ƙarfin watsa bayanai, da kuma masu kula da sirri. Takardar ta gano daidai cewa ainihin ƙimar IoT ba ta cikin samar da bayanai ba, amma a cikin aiwatarwa cikin gaggawa, na gida—wani aiki wanda girgije ba shi da tsarin gine-gine don ba da shi yadda ya kamata. Kamar yadda aikin mahimmanci kan Tsarin Jiki na Cyber (CPS) na Edward Lee da Seshia ya tabbatar, haɗin kai na lissafi tare da hanyoyin jiki yana buƙatar ƙayyadaddun lokaci, wanda girgije masu nisa ba za su iya tabbatar da shi ba.
Kwararar Hankali & Ƙarfafawa: Tsarin babin yana da ma'ana, yana motsawa daga tsare-tsare zuwa gine-gine zuwa tabbatarwa na gaske. Ƙarfinsa yana cikin rarrabe Cloudlet da MEC sosai—wani bambanci da ake yawan watsi da shi. Mai da hankali kan sauƙaƙan tsarin 'virtualization' yana da hangen nesa; tsarin kwantena (Docker) da fasahohin microVM (Firecracker) hakika su ne ƙa'idodin da ake amfani da su don tura aiki zuwa bakin cibiyar sadarwa, kamar yadda ake gani a dandamali kamar AWS IoT Greengrass da Azure IoT Edge, suna ba da damar tsarin "rubuta sau ɗaya, tura a ko'ina" wanda ke da mahimmanci ga bakin cibiyoyin sadarwa iri-iri.
Kurakurai & Rashin Ambato: Daftarin, duk da cewa yana da cikakken bayani, bai yi ƙasa da babban ƙalubalen tsarawa ba. Gudanar da dubban nodes na bakin cibiyar sadarwa masu rarrabawa, masu ƙarancin albarkatu, da kuma waɗanda za su iya motsawa yana da sarkakiya sosai idan aka kwatanta da sarrafa girgije na tsakiya. Ayyuka kamar KubeEdge da OpenYurt suna magance wannan, amma har yanzu babban shamaki ne ga amfani da kamfanoni. Bugu da ƙari, tsarin tsaro ana bi da shi cikin bege. Rarraba bakin cibiyar sadarwa yana faɗaɗa yankin harin sosai; kowane node ya zama wurin shiga mai yuwuwa, yana buƙatar tsarin gine-ginen rashin amincewa (zero-trust) waɗanda har yanzu suna girma.
Fahimta Mai Aiki: Ga masu aiki, abin da za a ɗauka a bayyane yake: Yi ƙira don rashin daidaituwa. Kar ku tura aikace-aikacen girgije guda ɗaya kawai zuwa bakin cibiyar sadarwa. Yi amfani da dabarun matakai: yi ƙididdigewa na gaskiya ($y = \text{SamfurinBakinCibiyarSadarwa}(x)$) da sarrafawa nan da nan a bakin cibiyar sadarwa, yayin da kuke aika sabuntawar samfura kawai da tsarin bayanai marasa daidaituwa ($\Delta \theta$, $x_{anomaly}$) zuwa girgije don sake horarwa da fahimtar duniya. Yaƙin gaba ba zai kasance a cikin ƙarfin lissafi na asali a cibiyar ba, amma a cikin tsarawar software mai hankali a duk faɗin hanyar daga na'ura zuwa girgije. Zuba jari a cikin ƙwarewa don dandamali kamar K3s (Kubernetes mara nauyi) da fahimtar dandamalin koyon haɗin gwiwa zai zama mahimmanci. Hasashen CAGR na 37.9% ba almubazzaranci bane; yana nuna wannan canjin gine-gine ya zama wajibi na masana'antu.
Cikakkun Bayanai na Fasaha & Tsarin Lissafi
Muhimmin ingantawa a cikin AI na bakin cibiyar sadarwa shine jinkiri na samfura da ciniki na daidaito. Don samfuri tare da sigogi $\theta$, jinkirin ƙididdigewa $L$ akan na'urar bakin cibiyar sadarwa tare da ƙarfin lissafi $C$ ana iya ƙirƙira shi azaman aiki na sarkakiyar samfura: $L \propto \frac{|\theta|}{C}$. Dabaru kamar ƙididdigewa (quantization) suna rage daidaiton sigogi (misali, daga floats na 32-bit zuwa integers na 8-bit), yana rage $|\theta|$ sosai don haka $L$, sau da yawa tare da ƙaramin asarar daidaito. Ana iya tsara matsalar ingantawa kamar haka:
$$\min_{\theta'} \, \mathcal{L}(f_{\theta'}, \mathcal{D}) \quad \text{subject to} \quad \text{Jinkiri}(f_{\theta'}) \leq T_{max}, \, \text{Ƙwaƙwalwar Ajiya}(f_{\theta'}) \leq M_{max}$$
inda $\theta'$ suke wakiltar sigogin da aka inganta, $\mathcal{L}$ shine aikin asara, $\mathcal{D}$ shine tarin bayanai, kuma $T_{max}$, $M_{max}$ su ne ƙuntatawar jinkiri da ƙwaƙwalwar ajiya na na'urar.
Tsarin Bincike: Lamarin Kulawa na Gaba
Yanayi: Binciken girgizawa don lafiyar famfo na masana'antu.
Aiwatar da Tsarin (Ba Code ba):
- Tushen Bayanai: Na'urar auna saurin gudu (accelerometer) akan famfo (sampling a 1 kHz).
- Sarrafa Bakin Cibiyar Sadarwa (Ƙofar Shiga):
- Mataki na 1 (Tacewa): Aiwatar da tacewa mai girma (high-pass filter) don cire ƙaramar girgiza injina.
- Mataki na 2 (Cire Fasali): Lissafa fasali na lokaci (RMS, Kurtosis) da fasali na mitar (manyan mitoci ta hanyar FFT) akan tagogin daƙiƙa 1.
- Mataki na 3 (Ƙididdigewa): Shigar da vector ɗin fasali cikin samfurin da aka riga aka horar, mara nauyi na Dajin Bazuwar (Random Forest) ko samfurin CNN 1D da aka tura a cikin kwantena akan ƙofar shiga ta bakin cibiyar sadarwa. Samfurin yana fitar da "makin lafiya" (0-1).
- Mataki na 4 (Aiwatarwa): Idan makin lafiya < 0.3, kunna faɗakarwa na gida da kuma tsara tikitin kulawa. Idan maki tsakanin 0.3-0.6, ƙara yawan sa ido.
- Haɗin Kai da Girgije: Ƙofar shiga tana aika jerin lokutan makin lafiya kawai da vectors ɗin fasali don maki < 0.6 zuwa girgije kowace rana don sake horar da samfura da binciken duk rundunar.
Sakamako: Jinkiri don faɗakarwa bai wuce daƙiƙa ɗaya ba. Amfani da ƙarfin watsa bayanai an rage shi da kusan 99% idan aka kwatanta da watsa bayanan girgizawa na danye. Samfurin girgije yana ci gaba da inganta ta amfani da fahimta da aka samu daga bakin cibiyar sadarwa.
Hangen Aikace-aikace & Hanyoyin Gaba
Gaba ɗaya (shekaru 1-3): Yaduwa a cikin Birane Masu Hankali don inganta zirga-zirga cikin gaggawa da binciken bidiyo na tsaron jama'a. Ci gaba a cikin Hanyoyin Wutar Lantarki Masu Rarrabawa don sarrafa ƙananan hanyoyin wutar lantarki da tashoshin cajin motocin lantarki. Faɗaɗawa a cikin Kasuwanci don ƙwarewar cikin kantin sayar da kayayyaki da sarrafa kaya.
Tsakanin lokaci (shekaru 3-5): Haɗuwa tare da Abun Ciki da AI Ya Ƙirƙira (AIGC) don ƙirƙira kafofin watsa labarai na gida, masu ƙarancin jinkiri (misali, masu tace AR, kayan wasa). Tashin Metaverse na Asali na Bakin Cibiyar Sadarwa, inda ake kiyaye tagwayen dijital na muhallin zahiri kuma ake mu'amala da su a bakin cibiyar sadarwa don tabbatar da amsa.
Dogon lokaci (shekaru 5+): Tushe don Komai Mai Gudanar da Kansa (motoci, jirage marasa matuka, mutum-mutumi) waɗanda ke buƙatar haɗin kai na fahimta da yanke shawara tsakanin na'urori (motar zuwa komai, V2X). Haɗawa tare da Hanyoyin Sadarwa na Gaba (6G+) don tallafawa sadarwar holographic da hankali ko'ina. Juyawa zuwa "Yadin Lissafi" inda albarkatu daga na'urori, bakin cibiyoyin sadarwa, da girgije ake tattara su cikin sauri kuma ana sarrafa su azaman amfani guda ɗaya, mara tsangwama.
9 Nassoshi
- Ashton, K. (2009). Wannan 'Abubuwan Intanet' Abun. RFID Journal.
- Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016). Lissafi a Bakin Cibiyar Sadarwa: Hangen Nesa da Ƙalubale. IEEE Internet of Things Journal.
- Satyanarayanan, M. (2017). Fitowar Lissafi a Bakin Cibiyar Sadarwa. Kwamfuta.
- ETSI. (2014). Lissafi a Bakin Cibiyar Sadarwar Wayar Hannu (MEC); Tsarin Gine-gine da Tsarin Nassoshi. ETSI GS MEC 003.
- Lee, E. A., & Seshia, S. A. (2017). Gabatarwa ga Tsarin Da aka Haɗa: Hanyar Tsarin Jiki na Cyber. MIT Press.
- Rahoton Binciken Kasuwa akan Lissafi a Bakin Cibiyar Sadarwa (2023). [Ambaton hasashe don bayanan kasuwa].
- Han, S., Mao, H., & Dally, W. J. (2016> Matsi Mai Zurfi: Matsa Cibiyoyin Jijiyoyi Masu Zurfi tare da Yanke Reshe, Ƙididdigewa da aka Horar da Kuma Lambobin Huffman. ICLR.
- Morabito, R. (2017). Tsarin 'Virtualization' akan Na'urorin Bakin Cibiyar Sadarwa na Abubuwan Intanet tare da Fasahohin Kwantena: Kimanta Aiki. IEEE Access.
- KubeEdge. (2023). Dandamalin Lissafi a Bakin Cibiyar Sadarwa na Asali na Kubernetes. https://kubeedge.io
- McMahan, B., Moore, E., Ramage, D., Hampson, S., & y Arcas, B. A. (2017). Koyon Cibiyoyin Jijiyoyi Masu Zurfi Mai Ingantacciyar Sadarwa daga Bayanan Da aka Rarraba. AISTATS.