输电线除冰机器人多f々感器信息融合障碍物信息柃测方i:左
Abstract
Multi-sensor
techniquesforobstacleinfomationdetectionanddatafusionisononeofthekeyofautonomousdeicingrobothighVoltagetransmissionline.The
suspensiononobstaclessuch嬲
Voltagecounterweight,lineshouldinsulator,stminclampandmeasuredhightotl‘ansmission
andbeaVoideff色ctiVeIyand
onf.ordeicingAccordingrobottointelligentlyautonomouslycrossobstacles.thecharacteristicsanddistributionofobstacles220KVtransmissionline,aseriesof
onmethodsf.orobstacleinfbrmationdetectionanddata如sionbased
systemaremulti.sensorfollows:putforwardinthisthesis.Themainstudiesinthispaper
onareasFirstly,adetailedanalysiskeyissuesofmethodsforobstacleinf.011口1ation
aredetectionanddata缸sionofautonomousdeicingrobot
a玎daputforwardinthisthesis,multi—sensorobstacledata如sionsystemisdeVeloped.
onSecondly,themodelformulti—sensorobstacledata向sionsystemofrobot
transmissionlineisbuilt.Theobstacledetectionsystemandequationofsystemstate
aredesigned.
’I’hirdly,amethodofobstacledatafhsionbasedonneuralnetworkisresearched.
sensorMethodsoff.eatureextractionanddataassociationf.ordataprocessingare
giVen.Aappropriateneuralnetwork
withadaptiVeconVergencerate
accuratedataonaredata向sionmodelanditsBPlearningalgorithmaproposed.goodconvergenceresultsandareseriesofMulti-sensorobstaclef.usion
wayobtained.based
atonAtlast,anotherisofinfo咖ation向sionextendedKalmanfilterare(EKF)alsostudiedinthispaper.Aimed
statethenonlinearmulti-sensoronsystem,themodeland
forwardtoreceiVenewequationofobstacledata如sionbasedEKFareputdatafilterandforecastprocessmethod.Multi—sensorobstacle
fhsiondatawithhigherintegrationaccuracyandmorerapidconVergencethanneural
networkafeobtained.
Keyword:DeicingRobot;Multi—sensorData
Network:ExtendKalmanFilterFusion;ObstacleDetection;Neural
lIl