高频状态估计使用低成本惯性光学流量测量单元琼joan sol天使讲稿.pdf

2015IEEE/RSJInternationalConferenceonIntelligentRobotsandSystems(IROS)

CongressCenterHamburg

Sept28-Oct2,2015.Hamburg,Germany

High-frequencyMAVstateestimation

usinglow-costinertialandopticalflowmeasurementunits

`

AngelSantamaria-Navarro,JoanSolaandJuanAndrade-Cetto

—Thispaperdevelopsasimpleandlow-costmethod

for3D,high-ratevehiclestateestimation,speciallydesignedfor

free-flyingMicroAerialVehicles(MAVs).Wefuseobservations

frominertialmeasurementunitsandtherecentlyappearedlow-

costopticalflowsmartcameras.Thesesmartcamerasintegrate

asonaraltimeter,atriaxialgyrometerandanopticalflow

sensor,anddirectlyprovidemetricego-motioninformationin

theformofbodyvelocitiesandaltitude.Comparedtostate-

of-the-artvisual-inertialodometrymethods,weareableto

drasticallyreducethecomputationalloadinthemainprocessor

unit,andobtainanaccurateestimationofthevehiclestateat

ahighupdaterateof100Hz.WethutendthecurrentuseFig.1.QuadrotorusedintheexperimentswiththeIMUandopticalflow

ofthesesmartcamerasfromhoveringpurposestoodometrymeasurementunitsattachedbelow.

estimation.Inordertoproposeasimplealgorithmicsolution,

weinvestigatetheperformancesoftwoKalmanfilters,intheCombinationsofIMUandvisualsensors(especiallyin

extendedanderror-stateflavors,alongsidealargenumberof

monocularmode)arebecomingverypopular,thankstotheir

algorithmvariations,usingsimulationsandrealexperiments

withprecisegr