Browsing by Author "Tedesco, Francesco"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Cooperative Distributed Command Governor Strategies for Surface Marine Vehicles
    (Università della Calabria, 2025-02-26) ElQemmah, Ayman; Fortino, Giancarlo; Tedesco, Francesco
    Thedeploymentofswarmingsurfacemarinevehicleoffersnumerousad- vantagesinaquaticenvironments,includingoperationalflexibility,taskand environmentaladaptability,resilience,andscalability.Inthiscontext,coor- dinationtogetherwiththesatisfactionofactuator,state,safety constraints becomesessentialtoenhanceautonomy.Meetingthisdemandischallenging: whileadvancedtechniqueshavebeenproposed,manyofthemarecomplex and/ornecessitatefrequentreplanning,whichcannegativelyimpactperfor- manceandcomputationalintensity.Therefore,readilyimplementableandlow computationallydemandingstrategiesarerequired. Thisdissertationpresents: i) aclassofnoveldistributedsupervisionstrate- giesappliedtomarinesurfacemulti-vehiclessystemstoaddresscoordination problems, ii) anovelsupervisionschemethatbypassestheneedforexplicit modelingand iii) experimentalvalidationofdistributedsupervisionstrategies. ByresortingtoCommandGovernor(CG)ideas,predictivesupervision algo- rithmsareemployedtoeffectivelydealwiththeabovementionedsupervision problems. WhiletraditionaldistributedCGstrategieshaveanon-cooperativeap- proach,inthisdissertationtwodistributedschemesthatpromote cooperation amongagentsareintroduced.Initially,consideringthecomputationallyde- mandingnatureofcooperativeapproaches,themainideaconsists ofleverag- inganon-cooperativenon-iterativesolution,andenhancingitto inducecoop- erationbetweenagents,thuspreservingthelowcomputationalburden. Forthis reason,theresultingschemeisreferredtoas“Cooperation-inducing.”Sub- sequently,inspiredbyadistributedoptimizationalgorithmfrom literature, a moregeneraldistributedapproachthatallowsagentstocomputeanear- optimalcentralizedsolution,isfullydescribedandanalyzed. Inthisrespect, wheretheresultingschemeisreferredtoas“Cooperative,”agentsexchange onlyauxiliaryvariables,thusensuringprivacypreservation.Thedevelopment ofbothschemesenablestheselectionofthemostsuitableapproachbased on specificapplicationrequirements. Concurrently,toaddressthechallengeofsolvingtheCGdesignproblem whenanexplicitmodelrepresentationisnotavailable,anapproachreferredto as“Data-Driven,”ispresented.Thisalternativetothemodel-basedapproach avoidstheneedfortime-consumingmodeling,identification,and validation phasesand,unlikeotherexistingapproaches,doesnotrequireanewdata collectionamidstcontrollermodifications. Finally,withtheaimofevaluatingcomputationalperformances andim- plementabilityofCG-baseddistributedstrategiesonmarinesurfacevehicles, severalexperimentalresultsinvariousenvironmentsarepresented.These re- sultsareobtainedusingprototypesofhighlymaneuverablemarinesurface vehicles,followingacomprehensivedesign,identification,validation,control, anddeploymentprocess.
  • Thumbnail Image
    Item
    Sensor Selection and Reconciliation Methodologies for Fault-Tolerant Estimation
    (Università della Calabria, 2025-01-30) Torchiaro, Franco Angelo; Fortino, Giancarlo; Casavola, Alessandro; Tedesco, Francesco
    This dissertation delves into the development of novel state estimation architectures and sensor selection methodologies to enhance fault tolerance in state estimation schemes. The research addresses a twofold objective: designing faulthiding architectures, and determining the optimal number/placement of sensors to efficiently support these architectures. The 昀椀rst objective advances the state-of-the-art in fault-tolerant estimation schemes by introducing novel architectures based on the Sensor Reconciliation methodology. Both centralized and distributed approaches are explored. A centralized method is proposed that combines a Luenberger observer and a reconciliator unit into a merged component that is used to identify and mitigate faults. Recognizing the limitations of centralized solutions for large-scale systems, a distributed sensor reconciliation architecture is also proposed, based on the decomposition of a centralized steady-state Kalman 昀椀lter. This latter approach allows for distributed state estimation by fusing local measurements and estimates from neighboring agents. Both methods leverage sensor redundancy to enhance robustness against sensor faults. Secondly, the dissertation explores the optimization of sensor selection. A Mixed-Integer Semide昀椀nite Programming (MISDP) framework is de昀椀ned and presented to systematically optimize sensor selection and design the corresponding L1 optimal observer. This approach minimizes the number of sensors required for full state reconstruction while optimizing speci昀椀c state reconstruction metrics to enhance efficiency, reliability, and robustness. Theoretical validation of the sensor placement’s optimality is also provided. The e昀昀ectiveness of all proposed methodologies is rigorously assessed through extensive simulations. In particular, a generic road network traffic 昀氀ow monitoring scenario serves as a practical testbed to demonstrate the potential applications of this research.