Utafiti huu unachunguza usanifu wa cascade control wenye tabaka mbili wa LAMDA-LAMDA unaowezesha magari matatu ya angani yenye mikono ya roboti kufuatilia trajectory ya pande tatu iliyowekwa huku yakidumisha kwa pamoja formation ya pembetatu. Outer LAMDA loop hudhibiti jiometri ya formation na mwendo wa center point, huku inner LAMDA loop ikijaribu kufidia disturbances za dynamic kulingana na makosa ya velocity. Sifa kuu ya mbinu hii ni kwamba, tofauti na njia za classical inverse-dynamics, haihitaji kutumia explicit na complete model ya dynamics changamani kati ya aerial platform na robotic arm ndani ya control law. Katika simulation comparisons, LAMDA-LAMDA ilitoa total error metrics za chini zaidi kati ya mbinu tano zilizotathminiwa, ikiwa na mean ISE = 0,702 na IAE = 1,652. Hata hivyo, utafiti hauwasilishi flight au load-carrying experiment kwenye aerial robots halisi; matokeo yote ya performance yanategemea simulation environment na hasa parametrically disturbed scenario iliyofanywa kwa makusudi kuwa kali ili kupima robustness.

Tatizo la control si tu kufanya magari matatu yasogee katika mwelekeo mmoja. Flight platform ya kila aerial manipulator na robotic arm iliyoko juu yake zinaweza kuathiriana dynamically. Mwendo wa arm unaweza kubadilisha inertia distribution na behavior ya mfumo; external loads na model uncertainties pia vinaweza kudhoofisha control performance. Kwa sababu hii, watafiti walitumia hierarchical structure inayotenganisha high-level control ya formation geometry na low-level control inayofidia dynamic error.

Formation ya aerial manipulators watatu inafafanuliwa kwa edge lengths mbili \(d_1\) na \(d_2\), internal angle \(\beta\), na coordinates \(x_F\), \(y_F\), \(z_F\) za formation center. Katika simulation, formation ilireconfigured katika sekunde ya 40; desired edge lengths zilibadilishwa kutoka 3 m hadi 2 m, na angle kutoka 45° hadi 60°. Kisha katika sekunde ya 70, inertia, Coriolis na gravity terms zilibadilishwa kwa kiasi kikubwa ili kupima robustness ya control architectures dhidi ya dynamic model mismatch.

Matokeo yanaonyesha kwamba mbinu kama SMC-ID na SMC-SMC zinazotumia model information moja kwa moja zinaweza kuathiriwa na model mismatch kubwa, huku muundo wa LAMDA wa tabaka mbili ukiweza kupunguza formation na position errors kwa haraka zaidi chini ya simulation conditions zilizochunguzwa. Kutokuwepo kwa high-frequency oscillations katika control signals pia kulitafsiriwa na waandishi kama mojawapo ya faida muhimu za mbinu ya LAMDA-LAMDA.

Kwa mtazamo wa Uturuki, utafiti unaweza kutoa mfano wa kimethodolojia kwa research ya control ya UAV swarms, multi-robot systems, aerial maintenance, cooperative load transport na aerial robots zinazofanya physical interaction. Hata hivyo, superiority ya ISE/IAE iliyopatikana katika simulation haionyeshi kwamba performance hiyo hiyo itapatikana kwenye UAV platforms halisi nchini Uturuki. Kwa matumizi halisi, experimental validation inahitajika ikiwa ni pamoja na sensor noise, communication delay, real rotor dynamics, actuator saturation, wind, load variations na collision avoidance.

Tatizo kuu la utafiti ni nini?

Aerial manipulator mmoja ni mchanganyiko wa multirotor aerial platform na robotic arm iliyowekwa juu yake. Katika mfumo kama huu, movement za arm hazibadilishi tu position ya end effector; pia zinaweza kuathiri dynamics ya aircraft. Aerial manipulators wengi wanaposogea pamoja, si stability ya kila robot pekee inayopaswa kuhifadhiwa, bali pia geometric formation ya kundi.

Utafiti unasisitiza hasa kwamba control methods zinazohitaji dynamic parameters kujulikana kwa usahihi zinaweza kuwa sensitive kwa tatizo hili: katika real system, mass, inertia, load, Coriolis effects na dynamic components nyingine zinaweza kutofautiana na model. Kwa hiyo watafiti wanapendekeza dual-layer inference-based control system ambayo haitegemei sana explicit full-dynamic-model compensation.

Formation inafafanuliwaje?

End-effector positions za aerial manipulators watatu zinafafanuliwa kama:

\[ \xi_1=(\xi_{x1},\xi_{y1},\xi_{z1}), \quad \xi_2=(\xi_{x2},\xi_{y2},\xi_{z2}), \quad \xi_3=(\xi_{x3},\xi_{y3},\xi_{z3}) \]

.

Edge lengths mbili kuu za formation geometry ni:

\[ d_1=\|\xi_1-\xi_3\| \]

na

\[ d_2=\|\xi_1-\xi_2\| \]

. Geometric variable ya tatu ni internal angle \(\beta\) ya triangle. Overall position ya formation katika space inaelezwa kwa center point, ambayo ni average ya end effectors watatu:

\[ x_F=\frac{\xi_{x1}+\xi_{x2}+\xi_{x3}}{3}, \quad y_F=\frac{\xi_{y1}+\xi_{y2}+\xi_{y3}}{3}, \quad z_F=\frac{\xi_{z1}+\xi_{z2}+\xi_{z3}}{3} \]

Hivyo overall control objective hukusanywa katika formation vector yenye variables sita:

\[ \Gamma = [d_1,\ d_2,\ \beta,\ x_F,\ y_F,\ z_F] \]

Mgawanyo huu ni muhimu: \(d_1\), \(d_2\) na \(\beta\) zinawakilisha shape ya group; \(x_F\), \(y_F\) na \(z_F\) zinawakilisha center position ya group katika space.

Inapitwaje kutoka robots binafsi hadi common formation model?

End-effector velocity ya kila aerial manipulator inaunganishwa na control velocities kupitia Jacobian matrix yake:

\[ \dot{\xi}_i=J_i\mu_i \]

Robots watatu wakichukuliwa pamoja, block-diagonal collective Jacobian \(J_R(q)\) hutumiwa kupata:

\[ \dot{\xi}_R=J_R(q)\mu_R \]

.

Derivative ya formation variables kwa end-effector positions inaonyeshwa kwa formation Jacobian \(J_F(\xi_R)\):

\[ \dot{\Gamma}=J_F(\xi_R)\dot{\xi}_R \]

Expressions hizi mbili zikichanganywa, common kinematic model kuu ya utafiti hupatikana:

\[ \dot{\Gamma} = J_F(\xi_R)J_R(q)\mu_R \]

Equation hii inaeleza jinsi movements za robots binafsi zinavyobadilishwa kuwa formation shape na center motion.

Kwa nini classical matrix inverse haiwezi kutumika?

Jacobian matrices \(J_R\) na \(J_F\) katika utafiti kwa kawaida si square. Kwa hiyo classical matrix inverse haiwezi kutumika moja kwa moja. Watafiti hutumia Moore–Penrose pseudoinverse katika control laws.

Ili pseudoinverse iwe well-defined, utafiti unadhania kwamba Jacobians zina full row rank katika operating region husika:

\[ \mathrm{rank}(J_R)=m, \qquad \mathrm{rank}(J_F)=p \]

Chini ya assumption hii, desired velocities katika formation level zinaweza kurudishwa kwenye robot control inputs kwa least-squares sense. Kwa hiyo stability na control laws za mbinu hutegemea Jacobians kudumisha rank conditions zinazohitajika katika operating region husika.

Dynamic model ina terms gani?

Ili kutathmini comparative control methods na kujenga simulation system, dynamic model ya aerial manipulators imeelezwa kama:

\[ \mu_{\mathrm{ref}} = M(q)\dot{\mu} + C(q,\mu)\mu + g(q) \]

imeelezwa katika namna hii.

• \(M(q)\): block-diagonal matrix yenye inertia matrices za aerial manipulators watatu.
• \(C(q,\mu)\): inawakilisha Coriolis na centrifugal effects.
• \(g(q)\): inawakilisha gravity forces.
• \(\mu_{\mathrm{ref}}\): reference input inayozalishwa na control system.

Dai kuu la proposed LAMDA-LAMDA control law ni kwamba si lazima ku-invert na kufidia wazi parameters za full dynamic model hii wakati wa control. Hata hivyo, virtual aerial manipulator na model-disturbance experiments katika simulation bado zilifanywa juu ya dynamic structure iliyoainishwa katika utafiti.

Ni control approaches gani tano zinazolinganishwa?

Layers mbili za LAMDA-LAMDA architecture zinafanya nini?

Outer loop inafuatilia error kati ya actual formation na desired formation. Lengo ni kuzalisha formation reference velocities kiasi kwamba:

\[ \tilde{\Gamma}=\Gamma_d-\Gamma \rightarrow 0 \]

.

Inner loop inafuatilia error kati ya reference velocity inayozalishwa na outer loop na actual robot velocities:

\[ \tilde{\mu} = \mu_{\mathrm{ref}}-\mu \rightarrow 0 \]

Layer hii inalenga kupunguza dynamic uncertainties na disturbances katika velocity level kwa LAMDA inference. Hivyo outer loop hushughulikia swali “group iende wapi na katika shape gani?”, huku inner loop ikishughulikia “robots zifanyeje desired motion hii licha ya dynamic uncertainties?”.

Kwa nini sliding surfaces zinatumika?

Katika inner LAMDA loop, sliding surface ya velocity error inafafanuliwa:

\[ s_L = \lambda_{FpL}\tilde{\mu} + \lambda_{FiL}\int\tilde{\mu}\,dt + \lambda_{FdL}\dot{\tilde{\mu}} \]

.

Katika outer loop, structure hiyo hiyo inatumika kwa formation error:

\[ s_{LF} = \lambda_{FpE}\tilde{\Gamma} + \lambda_{FiE}\int\tilde{\Gamma}\,dt + \lambda_{FdE}\dot{\tilde{\Gamma}} \]

Proportional term huleta instantaneous error, integral term accumulated error, na derivative term rate of change ya error kwenye control surface. LAMDA hu-classify features zinazotengenezwa kutoka surfaces hizi na derivatives zake ili kutoa control action.

MAD concept ya LAMDA ni nini?

LAMDA huamua kiwango ambacho kila input descriptor inafanana na class fulani kupitia Marginal Adequacy Degree (MAD). Gaussian similarity katika utafiti limetolewa kama:

\[ MAD = e^{ -\frac{1}{2} \left( \frac{o_{ij}-\rho}{\sigma_{ij}} \right)^2 } \]

.

Hapa \(o_{ij}\) ni input feature inayotathminiwa, \(\rho\) ni mean ya training samples za class, na \(\sigma_{ij}\) ni standard deviation. Utafiti umetumia \(\sigma_{ij}=0,3\).

MAD values nyingi kisha huunganishwa ndani ya Global Adequacy Degree (GAD) kwa kutumia T-norm na S-norm components. Hivyo degree ya membership ya input kwenye fuzzy class fulani huamuliwa na corresponding control output hupatikana.

Fuzzy classes zinafafanuliwaje?

Unique classes tano zinazotumiwa katika chanzo ni:

• PB — Positive Big
• PS — Positive Small
• ZE — Zero
• NS — Negative Small
• NB — Negative Big

Source note: katika sentensi moja ya method text, baada ya kusema “five classes”, Positive Small (PS) imeandikwa mara mbili. Table 1 katika section hiyo hiyo inaonyesha wazi unique classes tano hapo juu. Verianla haibashiri sababu ya repetition hii kwa niaba ya chanzo.

Stability inashughulikiwaje?

Utafiti umejenga Lyapunov candidate functions kwa sliding-mode controllers na LAMDA-based layers. Katika inner LAMDA loop, kwa mfano:

\[ L_L= \frac{1}{2}s_L^Ts_L \]

imetumika, na control gains zikitimiza appropriate bound conditions, \(\dot L_L<0\) hupatikana na convergence ya \(s_L\rightarrow0\) hutetewa.

Kwa outer loop pia formation sliding surface inalengwa kufikia zero asymptotically. Theoretical analysis hii imetolewa chini ya bounded uncertainties na rank/gain assumptions zilizobainishwa katika utafiti; si stability experiment kwenye real hardware.

Source-internal algebraic issue katika inner-loop stability condition

Katika source, ndani ya norm bound inayoelekea Equation 70, contributions mbili za LAMDA control zinaonekana kama negative terms:

\[ -\|\cdot\|K_{aL} - \|\cdot\|K_{dL} \]

. Hata hivyo, sufficient-condition text inayofuata huandika gain combination kama \(K_{aL}-K_{dL}\). Similar derivation ya outer loop katika source hutumia contributions hizo kwa addition. Verianla haisahihishi kimya kimya tofauti hii ya algebra na inaonyesha kwamba sehemu hii ya stability proof iliyoandikwa katika source inahitaji kukaguliwa tofauti.

Reference change iliyojaribiwa ni ipi?

Katika sekunde ya 40 ya simulation, desired triangular formation ilibadilishwa kutoka:

\[ [d_{d1},d_{d2},\beta_d] = [3\,\mathrm{m},3\,\mathrm{m},45^\circ] \]

hadi:

\[ [2\,\mathrm{m},2\,\mathrm{m},60^\circ] \]

.

Transition hii ya ghafla haijaribu tu kama control systems zinaweza kudumisha static formation, bali pia jinsi zinavyojibu reconfiguration ya formation geometry wakati wa operation.

Kwa nini disturbance ya sekunde ya 70 ni kali sana?

Dynamic model katika sekunde ya 70 ilibadilishwa kutoka nominal:

\[ [M(q),C(q,\mu),g(q)] \]

hadi:

\[ [2.4M(q),\,1.2C(q,\mu),\,100g(q)] \]

.

Mabadiliko haya, ikiwemo kuongeza gravity term mara 100, hayakutumika kuwakilisha normal flight condition. Watafiti wanaeleza wazi katika conclusion kwamba disturbance scenario hii ni extreme robustness stress test. Kwa hiyo performance baada ya sekunde ya 70 haiwezi kutafsiriwa kama “flight chini ya gravity mara 100 katika real world”.

Desired center trajectory inafafanuliwaje?

Three-dimensional reference trajectory ya formation center ni:

\[ x_{Fd}=3\cos(0.04t) \]

\[ y_{Fd}=3\sin(0.08t) \]

\[ z_{Fd}=\sin(0.2t)+1 \]

.

Source inatoa derivatives za \(x\) na \(y\) components mtawalia kama:

\[ \dot{x}_{Fd}=-0.12\sin(0.04t) \]

na:

\[ \dot{y}_{Fd}=0.24\cos(0.08t) \]

.

Source-internal mathematical inconsistency: katika section hiyo hiyo, derivative ya \(z_{Fd}=\sin(0.2t)+1\) imeandikwa kama \(\dot z_{Fd}=0.2\sin(0.2t)\). Trajectory expression iliyotolewa na derivative iliyotolewa si representations mbili zinazolingana za same mathematical function. Verianla haibadilishi expression hii kwa niaba ya source.

Figures 7 na 8 zinaonyesha nini?

Figure 7 inaonyesha pamoja desired three-dimensional center trajectory na calculated trajectory ya LAMDA-LAMDA controller pekee. Simulation inapohamisha formation kutoka starting point hadi ending point, inajaribu kuhifadhi triangular geometry ya aerial manipulators watatu.

Figure 8 inalinganisha motion hiyo hiyo kwa controllers tano. Katika nominal segment methods zote kwa ujumla zinaweza kufuatilia desired trajectory, huku tofauti kuu ikijitokeza baada ya large model disturbance ya sekunde ya 70. Source inaripoti kwamba LAMDA-LAMDA curve inaonyesha behavior iliyo karibu zaidi na desired trajectory baada ya disturbance.

Figures 9–11: Formation-shape errors

Figures 9 na 10 zinaonyesha distance errors \(\tilde d_1\) na \(\tilde d_2\) mtawalia, na Figure 11 inaonyesha angle error \(\tilde\beta\). Critical time regions mbili zimezoomwa kwenye plots: formation change ya sekunde ya 40 na dynamic disturbance ya sekunde ya 70.

Kinematic-SMC ina transient oscillations zilizo dhahiri zaidi kwa sababu haina dynamic compensation layer. SMC-ID hujibu vizuri reference change, lakini baada ya model mismatch permanent offset na oscillations hutokea. LAMDA-LAMDA nayo ina transient oscillations baada ya large disturbance, lakini inazipunguza errors tena kuelekea zero.

Source text inaeleza wazi kwamba angular formation error ya Kinematic-SMC inaweza kufikia amplitude takriban 0,25 rad.

Figures 12–14: Position errors za formation center

Plots hizi zinaonyesha errors za center coordinates \(\tilde x_F\), \(\tilde y_F\) na \(\tilde z_F\). Source inasema Kinematic-SMC inaweza kuzalisha oscillations karibu 0,05 m hasa baada ya sekunde ya 70.

LAMDA-LAMDA curves zinaonyesha smaller steady-state errors na faster recovery baada ya reference au disturbance transitions. Hata hivyo, plots ni simulation results na si physical flight test yenye real GPS, visual positioning, inertial measurement au communication noise.

Figures 15–20: Control signals na joint velocities

Figures 15–17 zinaonyesha LAMDA-LAMDA control actions katika directions \(x\), \(y\) na \(z\) kwa robot platforms tatu. Ingawa kuna transient changes zilizo wazi karibu na sekunde 40 na 70, source inasisitiza kwamba continuous high-frequency oscillations hazikuonekana.

Figures 18–20 zinaonyesha joint velocities za aerial manipulators watatu. Wakati wa severe disturbance katika sekunde ya 70, short-duration velocity peaks hutokea, lakini source inaripoti kwamba hazikubadilika kuwa unstable growth na joint velocities zilirudi kwenye regular levels.

Matokeo yanayoungwa mkono na utafiti

• LAMDA-LAMDA ilitoa mean ISE na IAE za chini zaidi kati ya control approaches tano katika simulation scenario iliyochunguzwa.
• Dual-layer structure hutumia LAMDA inference katika formation level na velocity/dynamic level.
• Baada ya abrupt formation change katika sekunde ya 40, LAMDA-LAMDA ilileta errors tena kuelekea zero.
• Chini ya severe parametric disturbance ya sekunde ya 70, proposed method ilionyesha lower overall error kuliko mbinu nyingine zilizolinganishwa.
• Lyapunov analysis iliyotolewa katika source inalenga asymptotic convergence chini ya rank, bound na control-gain conditions fulani.
• Katika simulation, LAMDA-LAMDA control actions hazikuonyesha continuous high-frequency chattering behavior.

Matokeo ambayo utafiti hauungi mkono au haujajaribu

• Utafiti haukufanya flight experiment kwa aerial manipulators halisi.
• Real cooperative load-transport success haijaonyeshwa.
• Wind, sensor noise, actuator saturation na physical communication delay hazijajaribiwa kwenye real hardware.
• 100g(q) disturbance katika sekunde ya 70 haijawasilishwa kama realistic operational condition; ni robustness stress test.
• Low simulation ISE/IAE values si guarantee ya real flight safety.
• Haijaonyeshwa kwamba LAMDA-LAMDA itakuwa superior kwa aerial-manipulator geometries zote na load conditions zote.
• Utafiti hautoi real-time embedded processor load, computational delay au energy-consumption measurement.

Mbinu na Matokeo ya Utafiti

Simulation design

Utafiti unatathmini formation yenye aerial manipulators watatu. Kila agent ina multirotor aerial platform na light robotic manipulator iliyowekwa juu yake. Mfumo umemodeliwa ili formation shape na center trajectory vidhibitiwe kwa wakati mmoja.

Matukio mawili makuu ya disturbance katika simulation ni:

Initial end-effector positions

Initial end-effector coordinates zilizotolewa katika source kwa robots watatu ni:

\[ \xi_1=[2,\ 2.78,\ 0.88]^T \ \mathrm{m} \]

\[ \xi_2=[4,\ 1.78,\ 1.88]^T \ \mathrm{m} \]

\[ \xi_3=[7,\ 2.78,\ 1.88]^T \ \mathrm{m} \]

.

Performance metrics

Control approaches zililinganishwa kwa integrated error metrics mbili.

Integral of Squared Error (ISE):

\[ ISE=\int error^2\,dt \]

Inapenalize errors kubwa kwa nguvu zaidi kwa quadratic form.

Integral of Absolute Error (IAE):

\[ IAE=\int |error|\,dt \]

Inapima accumulated absolute error magnitude katika simulation yote.

Katika metrics zote mbili, lower value inaonyesha lower total tracking error.

ISE results

LAMDA-LAMDA ilitoa mean ISE ya chini zaidi, 0,702. Source inaripoti tofauti hii kama %35,3 ISE reduction ikilinganishwa na second-lowest model-based result, SMC-SMC yenye 1,086.

Haiwezi kusemwa kwamba LAMDA-LAMDA ilitoa lowest result kwa kila subvariable. Kwa mfano, katika \(\beta\) ISE, SMC-SMC ina 0,0485, ambayo ni chini kuliko 0,074 ya LAMDA-LAMDA. Superiority ya proposed method inaonekana katika overall average ya error metrics sita.

IAE results

Proposed method pia ilitoa mean IAE ya chini zaidi, 1,652. Source inalinganisha LAMDA-LAMDA na SMC-SMC ya 2,432 na kuripoti %32,1 improvement.

Comparison note: Ukiangalia raw mean IAE katika Table 4 pekee, lowest value nje ya LAMDA-LAMDA ni 2,353 ya Kinematic-SMC; SMC-SMC ina 2,432. Kwa hiyo figure ya %32,1 katika source inalingana na comparison dhidi ya SMC-SMC na haipaswi kusomwa kama “second-lowest IAE among all reference methods”.

Tofauti kati ya SMC-LAMDA na double LAMDA

SMC-LAMDA inaacha inner dynamic loop pekee kwa LAMDA huku ikitumia SMC katika outer formation loop. LAMDA-LAMDA hutumia LAMDA inference mechanism katika layers zote mbili.

Average values zilizotolewa katika source ni:

Comparison hii ndiyo msingi mkuu wa watafiti kuhusisha performance gain si tu na ukweli kwamba “LAMDA inatumika”, bali na kuwekwa kwa LAMDA katika hierarchical control layers zote mbili.

Tofauti ya LAMDA definition ndani ya source

Introduction ya utafiti inafafanua LAMDA kama “supervised classification and inference algorithm”. Katika results analysis, expression “unsupervised learning algorithms such as LAMDA” inatumika. Classifications hizi mbili hazilingani terminologically ndani ya text moja. Badala ya ku-reclassify learning paradigm kwa niaba ya source, Verianla inawasilisha MAD/GAD-based LAMDA inference mechanism kama ilivyotumika katika utafiti.

Notation problem katika control parameters

Katika section ya LAMDA-LAMDA outer-loop parameters, source kwanza inatoa:

\[ K_{aLF}=\mathrm{diag}([0.1,\ 0.1,\ 0.1,\ 0.1,\ 0.1,\ 0.1]) \]

na baadaye, kwa symbol ileile \(K_{aLF}\):

\[ K_{aLF}=\mathrm{diag}([20,\ 20,\ 35,\ 20,\ 20,\ 20]) \]

. Zaidi ya hayo, diagonal matrices hizi zinaonyeshwa kwenye line hiyo hiyo kuwa sawa na \(I_{6\times6}\). Expressions hizi, kama zilivyoandikwa, si sawa. Kwa kuwa source haisahihishi wazi kwenye line hiyo ikiwa symbol ya pili inawakilisha gain tofauti, Verianla haibadilishi kwa kubashiri.

Strength kubwa zaidi ya utafiti

Strength kubwa zaidi ya utafiti ni kwamba hauonyeshi tu proposed controller ikifuatilia trajectory, bali unalinganisha architectures tano tofauti chini ya same reference change na same disturbance. Kutoa shape errors, center-position errors, control actions, joint velocities na integrated error metrics mbili kwa pamoja kunaruhusu tofauti kati ya mbinu kutathminiwa kutoka pembe nyingi.

Zaidi ya hayo, utafiti hauishii kwenye numerical simulation result kwa proposed controller; pia unatengeneza Lyapunov-based convergence analysis. Hata hivyo, kutokana na algebraic na notation inconsistencies zilizotajwa hapo juu, stability derivations ni muhimu kukaguliwa upya kwa kujitegemea.

Limitation muhimu zaidi

Limitation kuu ni kwamba validation yote imefanywa katika simulation. Watafiti wanakubali hili wazi katika conclusion na kupendekeza future work yenye experimental validation kwenye real aerial-manipulator platforms pamoja na sensor noise, actuator limits na more realistic disturbance models.

Kwa hiyo values 0,702 ISE na 1,652 IAE zilizoripotiwa kwa LAMDA-LAMDA si real field-performance metrics, bali ni matokeo ya simulation environment iliyofafanuliwa na utafiti.

Dokezo la Chanzo na Mbinu

Jina kamili la kazi asilia: Dual-Layer LAMDA-Based Cascade Control for Cooperative Formation of Aerial Manipulators

Waandishi na mpangilio: Gabriela M. Andaluz; Luis Morales; Paulo Leica; Guillermo Palacios-Navarro.

Corresponding authors: Gabriela M. Andaluz na Paulo Leica.

Equal contribution / co-first authorship: Hakuna tamko kama hilo katika kazi iliyochunguzwa.

Taasisi: Departamento de Automatización y Control Industrial, Escuela Politécnica Nacional, Quito, Ecuador; Department of Electronic Engineering and Communications, University of Zaragoza, Teruel, Spain. Gabriela M. Andaluz ana affiliations mbili, Luis Morales na Paulo Leica wamehusishwa na Escuela Politécnica Nacional, na Guillermo Palacios-Navarro na University of Zaragoza.

Aina ya source / peer review: Peer-reviewed research article. Scientific validation method si real-robot experiment bali simulation-based control comparison.

Journal: Actuators.

Publisher: MDPI, Basel, Switzerland.

Bibliographic information: Actuators 2026, 15(4), 204.

DOI: 10.3390/act15040204.

Submission / revision / acceptance / publication: 17 Februari 2026 / 29 Machi 2026 / 31 Machi 2026 / 4 Aprili 2026.

Official link:https://doi.org/10.3390/act15040204

License: Creative Commons Attribution (CC BY). Source ni open access.

Funding: Utafiti uliungwa mkono na Escuela Politécnica Nacional chini ya project PIS-23-09.

Data availability: Source inatumia kauli “Data sharing is not applicable to this article”.

Conflict of interest: Waandishi wanasema hakuna conflict of interest.

Ethics committee na informed consent: Zote zimeripotiwa kama “Not applicable”.

Author contributions: Conceptualization G.M.A., P.L. na G.P.-N.; methodology P.L.; software G.M.A.; validation G.M.A. na P.L.; formal analysis L.M.; investigation G.M.A. na L.M.; resources P.L.; data curation G.M.A.; original draft G.M.A. na L.M.; review/editing G.M.A., P.L. na G.P.-N.; visualization G.M.A.; supervision P.L. na G.P.-N.; project administration P.L. na G.P.-N.; funding acquisition P.L.

Scientific-validation boundary: Matokeo yote yanategemea simulation. Mabadiliko M → 2,4M, C → 1,2C na g → 100g yaliyotumika katika sekunde ya 70 ni, kwa maneno ya waandishi, stress test kwa extreme conditions na hayawakilishi realistic standard operating environment. Real-platform validation imeachwa kwa future work.

Source-internal mathematical inconsistency: Baada ya kutoa trajectory \(z_{Fd}=\sin(0.2t)+1\), source inaandika derivative yake kama \(\dot z_{Fd}=0.2\sin(0.2t)\). Trajectory na derivative expressions hazilingani mathematically. Verianla text hii haisahihishi kimya kimya.

Source-internal stability-derivation note: Katika inner LAMDA Lyapunov bound, control gains mbili zinaonekana kama negative contributions katika Equation 70, lakini sufficient condition inayofuata hutumia \(K_{aL}-K_{dL}\). Katika outer-loop derivation ya source, corresponding gains zinaonekana kwa addition. Algebraic difference hii haijasahihishwa kwa niaba ya source.

Source-internal control-parameter note: Diagonal gain matrices mbili tofauti kwa LAMDA-LAMDA outer loop zimetolewa kwa symbol ileile \(K_{aLF}\), na kila moja inaonyeshwa kuwa sawa na \(I_{6\times6}\). Kama zilivyoandikwa, notations hizi hazilingani mathematically.

Source-internal terminology note: Introduction inafafanua LAMDA kama supervised classification and inference algorithm, wakati results analysis inatumia expression unsupervised learning algorithm. Terminological difference hii haijaondolewa kimya kimya.

Content method: Mbinu, equations, control architectures, simulation conditions, error metrics na scientific conclusions katika makala hii ya Verianla zinategemea tu kazi iliyochunguzwa. External sources zilitumika tu kwa bibliographic verification ya DOI, official publication identity, journal na peer-review status; hakuna external control performance au experimental result mpya iliyoongezwa.