Content uploaded by Shiwu Zhang
Author content
All content in this area was uploaded by Shiwu Zhang on Sep 12, 2022
Content may be subject to copyright.
Earthshaker: A Mobile Rescue Robot for Emergencies and Dis-
asters through Teleoperation and Autonomous Navigation
YuZhang1,3,YuxiangLi2,3,HefeiZhang1,YuWang2,ZhihaoWang2,YinongYe1,YongmingYue1,NingGuo1,
WeiGao1
✉
,HaoyaoChen2
✉
,andShiwuZhang1
✉
1Department of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230027, China;
2College of Mechanical Engineering and Automation, Harbin Institute of Technology Shenzhen, Shenzhen, Guangdong 518055, China;
3Co-first authors
✉
Correspondence: WeiGao, Email: weigao@ustc.edu.cn; HaoyaoChen, Email: hychen5@hit.edu.cn; ShiwuZhang, Email: swzhang@ustc.
edu.cn
©2022TheAuthor(s).ThisisanopenaccessarticleundertheCCBY-NC-ND4.0license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Graphical abstract
Overview of the rescue robot Earthshaker, the first place in the Advanced Technology & Engineering Challenge (A-TEC) champion-
ships.
Public summary
■Earthshaker,amobilerescuerobotthatcombinesatrackedchassis,aroboticarmandgripper,andvarioussensorsand
controllers,hasbeencreatedtodealwithvariousemergenciesanddisasters.
■Earthshaker’smultimodalteleoperationsystemcan adapt to differenttransmissionconditions,anditcan achieve both
semi-autonomousmanipulationwithitsarmandgripperandautonomousnavigationinunknownareas.
■EarthshakerwonthefirstA-TECchampionships,standingoutof40robotsfromtheworld,showingtheefficacyofthe
systemintegrationandthecontrolphilosophybehindit.
http://justc.ustc.edu.cn
Citation: ZhangY,LiYX,ZhangHF,etal.Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonom-
ousNavigation.JUSTC,2022,52(0):.DOI:10.52396/JUSTC-2022-0066
Just Accepted
Earthshaker: A Mobile Rescue Robot for Emergencies and Dis-
asters through Teleoperation and Autonomous Navigation
YuZhang1,3,YuxiangLi2,3,HefeiZhang1,YuWang2,ZhihaoWang2,YinongYe1,YongmingYue1,NingGuo1,
WeiGao1
✉
,HaoyaoChen2
✉
,andShiwuZhang1
✉
1Department of Precision Machinery and Precision Instruments, University of Science and Technology of China, Hefei 230027, China;
2College of Mechanical Engineering and Automation, Harbin Institute of Technology Shenzhen, Shenzhen, Guangdong 518055, China;
3Co-first authors
✉
Correspondence: WeiGao, Email: weigao@ustc.edu.cn; HaoyaoChen, Email: hychen5@hit.edu.cn; ShiwuZhang, Email: swzhang@ustc.
edu.cn
©2022TheAuthor(s).ThisisanopenaccessarticleundertheCCBY-NC-ND4.0license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
CiteThis:JUSTC,2022,52(X):(12pp) ReadOnline SupportingInformation
Abstract: Todealwithemergenciesanddisasterswithoutrescueworkersbeingexposedtodangerousenvironments,this
paperpresentsamobilerescuerobot,Earthshaker.Asacombinationofatrackedchassisandasix-degree-of-freedomro-
boticarm,as wellasmiscellaneoussensorsandcontrollers,Earthshaker iscapableoftraversingdiverseterrains andful-
fillingdexterousmanipulation.Specifically,Earthshakerhasauniqueswingarm–dozerbladestructurethatcanhelpclear
upcumbersomeobstaclesandstabilizetherobotonstairs,amultimodal teleoperation system that can adapt to different
transmissionconditions, adepthcameraaidedroboticarm andgripperthatcanrealizesemi-autonomous manipulation,a
LiDARaidedbasethatcanachieveautonomousnavigationinunknownareas.Itwasthesespecialsystemsthatsupported
EarthshakertowinthefirstAdvancedTechnology&EngineeringChallenge(A-TEC)championships,standingoutof40
robotsfromtheworldandshowingtheefficacyofsystemintegrationandtheadvancedcontrolphilosophybehindit.
Keywords: Rescue robot; Autonomous navigation; Semi-autonomous manipulation; Multimodal Teleoperation; System
integration
CLC number: TP242Document code: A
1 Introduction
Rescueworkers’livesareoftenunder threatduringtheirres-
cueworkin and after emergencies anddisasters. Sometimes
even casualties have to be suffered unfortunately. With the
development of robotics in general, robots have seen their
prosperityinreplacing human beings tofulfillmiscellaneous
tasksinthosedangerousscenarios[1−4].
Duringrescuework,itisoftenrequiredtotraverseunstruc-
tured and complicated terrains, even climb up and down
stairs, while carrying miscellaneous equipment and sensors
fordealingwithdangeroussituationsandclearingupcumber-
someobstacles[5].Therefore,mostrescuerobotshavebeende-
velopedbaseduponleggedortrackedroboticplatformstoen-
suremobility.Todate,plentyofleggedrobotshavebeende-
veloped by different research organizations and industrial
companies[6,7], and some of them have shown up in various
competitions like DARPA Robotics Challenge[8], DARPA
SubterraneanChallenge and so on[9]. To further improve the
mobilityofleggedrobots,therehavealsobeenhybridlegged
robotsthathave wheels or tracksattachedattheendoftheir
legs to replace the feet, e.g. RoboSimian[10], Momaro[11],
CHIMP[12],etc. However,theserobotshaveverycomplicated
structuresandlow-levelcontrolsthatconsume a lot of com-
putationpowerandcontroltime,thusresultinginarelatively
fragilesystemunder consistent large workloadduringrescue
work.Sofar, only thequadrupedal robot ANYmal hasbeen
successfully deployed in real rescue scenarios[13]. As with
tracked robots, popular ones are often equipped with swing
armsthat can help cross diverse obstacles, afford high pay-
load,and perform stable locomotion. After the 2016 earth-
quakeinItaly,a such tracked robot TRADR wasusedtoin-
spect damaged buildings[14]. In the ARGOS Challenge[15],
TeamArgonautsalsousedatrackedrobotwithswingarmsto
win the championship[16]. Other groups have even realized
autonomousnavigationfor such trackedrobots when climb-
ingstairs[17] andslipperyslopes[18].However,due totheexist-
ence of the swing arms, those robots lose the capability of
clearingupcumbersomeobstacles.Toovercomethosedraw-
backs in this work, we have designed a unique structure to
provide the tracked system the capability of both climbing
stairsandclearingobstacles.
Besidesthoughtful structural designs, autonomous opera-
tioncanalsogreatlyimproverescuerobots’efficiencyindif-
ferentrescueworks.A typical application scenario would be
exploringsignalblockedareasafteremergenciesordisasters
havehappened. To overcome the loss of telecommunication
betweentherobotsandtheoperators,autonomousnavigation
is potentially desired for rescue robots[19,20]. Miscellaneous
sensors can be taken advantage of to conduct simultaneous
localization of the robot and mapping of the unknown
area[21−23].WehaveintegratedaLight Detection and Ranging
http://justc.ustc.edu.cn
–1 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
(LiDAR) and an Inertia Measurement Unit (IMU) to build
gridmapsofthesurrounding environment, evaluate position
and pose of the robot and realize autonomous navigation.
Note that no Global Positioning System (GPS) is needed in
thisprocess,makingitparticularlysuitableforsignalblocked
areas.
Evenwithautonomousnavigation,therobotstillneedsthe
operators’ help when it comes to dexterous manipulation
taskslikeopeningdoors.Therewasevenacasewhereseven
operatorswereneededtocooperateoncontrollingarobot[24].
To release the operation burden, we have developed depth
cameraaided semi-autonomousmanipulationforroboticarm
indooropeningtasks that can quickly locatetheposition of
thedoorhandle. Thewholeoperationprocessjustneedstwo
operators’ cooperationto control the base and the arm, re-
spectively,largelyreducingthe operationcomplexityandin-
creasingoperationefficiency.
Consequently, the teleoperation system on a rescue robot
becomesquite critical for successful rescue works. The ef-
fectivityand reliability of the teleoperation system determ-
ines the lower boundary of the rescue robot’s performance.
Therefore, a multimodal teleoperation system to provide
enoughredundancy and deal with different conditions be-
comenecessary.
Basedupontheabovearticulation,wepresentinthispaper
ournewlydesignedrescuerobot,whichhassuccessfullyad-
dressed the aforementioned four points of functionality. We
havenamedit Earthshaker, not onlybecause it “shakes” the
ground when it moves around, but also because we hope it
canbringearthshakingimprovementontheroleofrobotsin
real rescue work. An overview of Earthshaker is shown in
Figure1.Theremainderofthepaperfirstintroducesthevari-
oussystemsofEarthshakerinSection2,includingthetracked
chassis, the robotic arm and gripper, the perception system,
theteleoperationsystem,andtheirmechatronicintegration.In
Section3, control frameworks of the multimodal teleopera-
tion,thedepth camera aided semi-autonomousmanipulation,
andtheLiDARaidedautonomousnavigationarepresentedin
detail.Section4 summarizestheperformanceofEarthshaker
inthefinalsofthe2020A-TECchampionshipsastheexperi-
mentalvalidationofthesystemintegrationandcontrolphilo-
sophy. The experience obtained from the competition and
possiblefuturedirectionsarediscussedinSection5.
2 Earthshaker
2.1 Tracked chassis for robust locomotion
The tracked chassis supports all the other systems onboard
with corresponding mechanical and electronic interfaces to
formtherobotasawhole.Itdeterminestheupperlimitofthe
wholesystem’smobility[25].Thetrackedchassisismadeofal-
loysteelthroughcastingandwelding.Itcombinesthedesign
ofChristiesuspensionandMatildasuspensiontoachieveex-
cellent traversing capability. The vibration and impact from
rough terrains can be effectively absorbed by the chassis to
maintainastable operation environment for the systems on-
board.Thechassisisdrivenbytwo1000wattsbrushlessmo-
tors,whichcansupportamaximumrunningspeedof1.6m/s
and a maximum climbing inclination of 40 degrees. Four
packsof LiPo batteries inside the chassis can power Earth-
shakerto continuously work for 3 hours at medium work-
loads.Eachbatterypacksupportsanindividualsystemtoen-
sure power isolation and security, namely, one 48 volts 60
ampere-hourpackfor the chassis, three24 volts 16 ampere-
hourpacksfor the manipulation system, the perception sys-
temandtheteleoperationsystem,respectively.Intheend,the
Earthshakeris 0.72 m wide and1.22 m high, and itslength
canvaryfrom1.33mto1.49m,withatotalweightofabout
250kg.
Topromotethe robot’s capability ofclearingcumbersome
obstacles and climbing up and down stairs, a swing arm –
dozer blade structure has been designed and attached to the
rearend of the chassis. The structure consists of an electric
linear actuator, two tracked swing arms, and a dozer blade.
Theelectriclinearactuatorcanbecontrolledunderteleopera-
tionto rotate theswing arms, thusadjusting the poseof the
dozerbladefrom65degreesto-45degreeswithrespecttothe
horizontaldirection.Onflatterrains,thestructureisfoldedto
reducemotionresistance and increaseagility,whileonstairs
Fig. 1.OverviewoftherescuerobotEarthshaker.
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–2 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
itcan be used to adjust pitch angle of the robot to improve
stability,asshowninFigure 2. When there are cumbersome
obstaclesintheway,thedozerbladecanbeputverticaltothe
ground to push them away obstacles efficiently, as long as
theyareunder75kg.
2.2 Robotic arm and gripper for dexterous manipulation
Withoutdexterous manipulation, taskslike pressing buttons,
opening doors, turning off valves, picking up small objects,
movingaround wounded victims, etc., cannot be accom-
plished.EarthshakerhasbeenequippedwithaUR5erobotic
armandanAG95two-fingergripperforthosepurposes.The
armcanrealizedexterousmanipulationwithinaradiusof750
mm,withamaximumpayloadof5kg[26].Thearmisinstalled
atthefrontend of the robot toguaranteeenough workspace
andbalancetheextra weight introduced bytheswingarm –
dozerbladeat the rear end. The Original Equipment Manu-
facturer(OEM)control box of the armhas been customized
tosavespace on the robotand can work under24volts DC
powerinsteadof220voltsACpower.Thevelocitycontrolof
eachjointonthearmandthegripperismappedtotheremote
controller, thus precise impedance control can be achieved.
Also, to facilitate semi-autonomous manipulation, an Intel
D435iRGBD camera has been mounted on the gripper, the
useofwhichisdiscussedlaterinSection3.2.
2.3 Sensors for diverse perception
Earthshakerhasaplatformforsensorinstallationbetweenthe
roboticarm andtheswingarm.Foursides oftherectangular
platform have four wide-angle cameras, which are headed
slightlydownwards to provide a panorama of the environ-
ment surrounding the robot. Thus, the remote operator can
plan paths and avoid obstacles accordingly. There are also
two infrared cameras on the sensor platform that can help
identifyobjectsinsmokyenvironment.Thetwoinfraredcam-
erasareplacedoppositetoeachother,withone pointingfor-
wardandonepointing backward[27]. At the frontpanelofthe
chassis,thereisanothermicrocamerathatcanprovideawide
view of the environment in front of the robot. With further
helpfrom the lasers onboth sides ofthe robot, the operator
canpreciselydriveEarthshakertopassthroughnarrowdoors
orcorridorswithoutanyproblem.
2.4 Teleoperation and communication
EarthshakeristeleoperatedbytwooperatorsusingtwoAT9S
remotecontrollers,oneforthetrackedchassisandoneforthe
roboticarm and gripper. Each controller has 12 channels to
transmitdigital commands via 2.4 GHz communication fre-
quencytothereceiverontherobot.ASTM32F091basedmi-
crocontrolleris then utilizedtodecodethesignalsto achieve
closed-loopcontrolofthechassis, as well as otherperipher-
alsliketheswingarm–dozerblade,thelasers,theLEDsetc.
Meanwhile,thesignalstothereceiverfortheroboticarmand
gripper are translated into specific actions by an Intel NUC
minicomputer,which hasaRAMof 16GBandanIntelCore
i7-1165G7 CPU with a maximum clock frequency of 4.7
GHz.Ontheotherhand,thevideoimagestransmittedbackto
theoperatorsconsistofimagesfromthewide-anglecameras,
theinfraredcameras,themicrocameraandtheoperatingsys-
temscreenoftheNUCminicomputer.Theseeightimagesare
selectedandcombinedintoonesingleimagefortransmission
tosavebandwidth.
Besidesthe2.4 GHz direct communication,there are also
tworedundant communication pathson Earthshaker, the 1.8
GHz MIMO-mesh radios[28] andthe 4G/5G mobile telecom-
munication.Theseadditionalpathscanovercometherelative
shortcommunicationdistanceofthe2.4GHzsignalsanden-
suretherobustnessofteleoperationforEarthshaker.
2.5 Mechatronic integration
Figure 3 summarizes the major mechatronic components of
Earthshaker,aswellasthesignalpathsformultimodalteleop-
eration.NotethattheNUCminicomputercanalsocontrolthe
chassis, depending on its priority comparing with the
STM32F091microcontrollerontheCANbus.Consequently,
theswitchbetween teleoperation and autonomousnavigation
canbeorganized.Toaccomplishautonomousnavigationand
dexterous manipulation, the NUC is also connected to the
Fig. 2.Demonstrationoftheswingarm–dozerbladestructure.Thestructurecanbefoldedorextendedbasedonneeds.
Zhangetal.
–3 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
LiDAR,the RGBD cameraand the grippervia USB cables,
andtotheroboticarmviaaswitch.The same switch is also
connectedtotheMIMO-meshradioandthe4G/5Grouter.As
aresult,theswitchbuildsa100Mbpsnetworkwiththeoper-
ators’ computer, the signal quality of which affects the
latencyofteleoperation.
3 Control
3.1 Control logic of the base
Inrealrescuework,itisinevitabletofaceenvironmentswith
detrimentalmagneticfieldsorpoorsignaltransmissioncondi-
tions,where theregular2.4GHzteleoperationsignals would
decaygreatly with reduced signal-to-noise ratio and in-
creaseddata package loss. To maintain robust signal trans-
missionbetween the operators andEarthshaker for real-time
teleoperation,a framework for multimodal teleoperation has
beendeveloped to ensure the communication path is un-
blocked,asshowninFigure4(a).
Withintheframework,whenEarthshakerisclosetotheop-
eratorsuch that theAT9S remote controllers can talkto the
receivers on the robot directly, the 2.4 GHz communication
frequencyisused. Once thedistanceinbetweenincreasesor
forsomereasonthesignals get blocked to a point where the
directcommunicationfails,the1.8GHzcommunication fre-
quency would be adopted and the signals are transmitted
throughtheMIMO-meshradios.Whennecessary,therobotic
armcanevendropanextrarelayradioonboardtofurtherin-
crease the communication distance and quality. Multiple
MIMO-meshradioscanformadistributednetworkwithvari-
ous forms, e.g., a line, a star, a net, and even a mixture of
those.Thenetworkcanflexiblyadapttofastnodemovement
andnode-to-nodesignalqualityvariation,realizinghighqual-
itysignaltransmissionconsistently.To ensure the teleopera-
tioncommunicationincaseeventheMIMO-meshradiosfail,
one more redundant communication path realized by the
4G/5GrouterhasbeenaddedtoEarthshaker.Theroutercan
eitherconnecttonearbybasestationsfromtheselectedInter-
netServiceProviderorberelayedbynearbyUnmannedAeri-
alVehicles,tobuildanetworkwithapresetcloudserver.The
operators can then access the server, monitor the real-time
datafromtherobotandgivecorrespondingcommands.
Jnorm
Earthshaker checks control signals from these three paths
accordingtotheir prioritylevelsandsignalquality. Ifeffect-
ive data are not received within a prescribed time, the path
withalowerprioritylevelwouldbechecked.Ifallthreepaths
fail, the program would determine whether to enter the
autonomous navigation mode or an emergency stop mode.
Onceanycommunicationpathissuccessfullyestablished,the
remotecontroller inthe baseoperator’s handscan drivethe
robottomoveforward,backwardandrotatearounditscenter
point.Themicrocontroller on therobot first unifies the joy-
stickvaluesobtainedfromtheremotecontrollerto ,
Jnorm =2Jinput −Jmax −Jmin
Jmax −Jmin
(1)
Jinput =[linput ,rin put ]T
Jnorm =[linorm,rnor m]T
Jmax
Jmin
v
ω
where , representing the values from each
joystick, ,representing the unified joystick
commands,and and denotethemaximumandminim-
um values of the joysticks, respectively. When the value is
zero,therobotisstill.Theunifiedvaluesaretheninterpreted
intothebase’slinearvelocity andangularvelocity as
v=0,i f lnorm
3<0.001,
lnorm
3∗vmax ,otherwise,(2)
and
ω=0,i f |rnorm
3|<0.001,
rnorm
3∗ωmax ,otherwise,(3)
Fig. 3.ElectricalschematicsofEarthshaker.
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–4 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
vmax
ωmax
ωl
ωr
where and arethemaximumlinearandangularvelo-
citiessupportedbythebase.Throughthekinematicmodelof
differentialdrive,theangularvelocities of left and right mo-
torsofthebase, and ,canbecalculatedas
ωl=2v−lω
2r,
ωr=2v+lω
2r,
(4)
l
r
ωl
ωr
where isthedistancebetweentracksand representsthera-
diusofthedrivewheel.Thecalculated and arethensent
tothemotorsascontrolcommands.
3.2 Control logic of the arm and gripper
Theprograms for teleoperation of the arm and gripper con-
sist of an operation assisting module for door opening task
andseveral interface modules for maintaining communica-
tionbetweentheNUCminicomputer and the other compon-
ents,includingthesignal receiver, the UR5e arm,theAG95
gripperandtheD435iRGBDcamera.Insidetheseprograms,
anetworksocketisfirstcreatedaccordingtothearmcontrol-
ler’s IP address and port number, such that the built-in
input/output functions can be called to read or write to the
sockettointeract with the armcontroller. At the sametime,
theserialports connected tothesignalreceiverandthe grip-
perareinitialized in the programs throughRS485protocols.
Once the NUC receives remote control instructions through
thesignalreceiver,itparsesthemintothepositionsandvelo-
citiesforeachjointofthearm,aswellastheopeningangle
andholdingforceforthegripper.
K
Tofacilitate semi-autonomous door opening for Earth-
shaker,theoperationassistingmoduleisdevelopedusingthe
depth camera in an eye-to-hand manner. This module, as
showninFigure4(b),cangreatlysimplifytheprocessofdoor
opening,avoidingthepotential mistakes that could beintro-
duced by the operator through teleoperation. In the module,
thecoordinatesofthecameraandthearmarefirstcalibrated
toobtain thetransformationrelationshipbetweenthem.With
theintrinsicparameter matrix ,thepixelsonthedepthim-
agesobtainedfromtheRGBDcameracan be converted into
three-dimensionalpointcloudas
Get
joystick
value
4G/5G
network
1.8GHz
MIMO-
Mesh radio
2.4G
remote signal
received?
1.8G
radio signal
received?
4G/5G
network
connected?
2.4GHz
remote &
receiver
Operator Side Remote
Side
E-Stop / Start autonomous
navigation
Y
N
N
N
Y
Y
Linear velocity&
angular velocity
Motor Speed
Nonlinear
mapping
Normalized
Kinematics
RGB Image 3D point
cloud
Depth Image
RealSense
D435i
Arm &
Gripper
Extracting the
door planar
Gripping
position
Clustering the
door handle
Gripping
orientation
Visually
feedback
control
External
parameter
calibration
EKF
Scan-
matching
Lidar
Odometry
LiDAR
Inertial
Odometry
Pre-
integration
IMU Laser-Inertial
Odometry
Occupancy
Grid Map
RRT
Expansion
NBV Path
Tracking
(a) Flow chart for control logic of the multimodal teleoperation.
(b) Flow chart of the semi-autonomous door opening algorithm.
(c) Flow chart of the autonomous navigation algorithm.
Fig. 4.FlowchartsofthecontrolalgorithmsforEarthshaker.
Zhangetal.
–5 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
P=DK P′,K=
1
fx
0
0
0
1
fy
0
0
0
1
,(5)
P
D
P′
(θ, n)
where is the coordinates of the 3D point, denotes the
depthmeasuredontherayofthepixel, isthecoordinatesof
thepoint intheimage.Next,objects canbeidentifiedwithin
thepointcloudconvertedfromthedepthimage.Specifically,
inthetaskofdooropening,thepositionandorientationofthe
doorandthe handle should beestimatedtoserveasthegoal
forpath planningofthearm andgripper.Theposition ofthe
dooris determined through fittingplanes to thepoint cloud.
Before the operation assisting module is started, the robot
needstobeinfrontofthedoorsuchthatthedoorisinsidethe
FieldofView(FOV)oftheRGBDcamera.Consequently,the
pointcloud corresponding to the doorcan be recognized by
planarsegmentation.Tofigureouttheorientationofthedoor,
the Principal Component Analysis (PCA) method[29] is ex-
ploitedtocalculatethenormalvectorofthedoorplaneinthe
point cloud. Then the axis-angle of the door’s normal
directioncanbecalculatedas
θ=acos(a·x),
n=a×x,(6)
θ
n
x
x
n
a
x
q
where denotestheanglebetweenthenormalvector ofthe
doorplaneand the unitvector of -axis, and represents
therotationaxisfromthe vector to the vector . Thus, the
rotationmatrix can befurther calculated with Rodrigues’s
Formula[30],
q=cosθI+(1−cosθ)nnT+sinθn∧.(7)
I
p
q
p
where isthe identity matrix. Subsequently, DBSCAN al-
gorithm[31]is usedtoclusterthecloud pointsthatarecloseto
thedoor plane.Theclusterwith apropersizeisidentifiedas
thepointcloudofthedoorhandle,andtheclustercenter is
calculatedas themeanofallthecluster pointsandsetasthe
targetpositionfor arm to grip.As a result, theorientation
andposition serveasthetargetposewhenapproachingthe
handle.However,duetotheobservationmodeloftheRGBD
sensor, the depth measurement error is proportional to the
squareofdistance.The eye-to-hand method leadsto a relat-
ivelylongseparationbetweenthetargetandthesensor,inev-
itablycausingobservationerrorsforthegrippingpose.Addi-
tionally, the vibration introduced by the movement of the
basealsomakesit difficult to realizevisuallyfeedback con-
trolofgripping.Hence,atthispoint,thealgorithmisonlyad-
optedtoprovideaninitialposeforthedooropeningtask.The
remainingoperationstillneedstobecompletedbytheoperat-
ors. Even with this level of semi-autonomy, the operating
steps have been greatly simplified and the operation burden
ontheoperatorsissufficientlyreleased.
3.3 Autonomous navigation
Whenautonomousnavigation is desired, the control author-
ityofEarthshakercouldbegiventotheNUC.Thishelpsthe
robotexploreunknownandsignalblockedareasactivelyand
searchforanexittowardsadesireddirection,thealgorithmof
whichcanbefoundinFigure4(c).Oncetheautonomousnav-
igationisstarted,theNUCanalyzesthedatascanned by the
XL
c
LiDARtobuildagridmapofcurrentenvironmentandestim-
ate its ego-motion simultaneously. Feature matching based
methodsuchas LOAM[32] isa popularposeestimationmeth-
odthat demonstratesrobustnessandefficacy incomplexoff-
roadenvironment.Therefore,scanmatchingisalsoincorpor-
ated into Earthshaker’s autonomous navigation algorithms.
Featuresareextracted from each frameoftheLiDARsweep
forthesmoothness as
c=1
|S|XL
(i)
j∈S,j,i
(XL
(i)−XL
(j))(8)
XL
i
i
S
i
S
T
where denotes the -th point within the sweep, and
definesasetofconsecutivepointsobtainedbythesamelaser
beamnearpoint .Thepointnumberwithin is empirically
setto10.Bysettingathresholdforsmoothness,thecurvecan
be determined as edge feature with greater smoothness and
planarfeaturewithlesssmoothness.Thentheedgeandplanar
featuresofconsecutiveframescanberegisteredseparatelyto
restore the motion between frames using Iterative Closest
Point(ICP) algorithm[33].The object functionfor the ICP al-
gorithmisset tominimizethecostwithrespecttotheestim-
atedtransformation as
f(T)=min f
n
i=1
d∈i(T
+
m
j=1
dH j(T)) (9)
m
n
d
dH
where and denote the numberofmatched edge features
and planar features, respectively, represents the distance
betweentwomatchededgefeaturesand representsthedis-
tancebetweentwomatchedplanarfeatures.
Due to the vibration of the base caused by tracks and
ruggedterrains,IMUpre-integration[34]is introduced into the
system to further improve the robustness of the localization
results. As shown in Figure 4(c), the Extended Kalman
Filter[35]isusedtoinferthestateoftherobot,fusingthescan-
matchingresultsandtheIMUpre-integrationresultsinatight
couplingmanner.
Withthehigh precision Laser-Inertial odometryestimated
fromEKFfusion,thelaserscansarethenmergedintotheoc-
cupancygridmap.Ingeneral,theexplorationtaskistomax-
imize the covered area on the grid map. Herein, a frontier
basedmethod[36]isusedtoguidetherobottoexplorealongthe
boundary between unknown area and free area on the grid
map.In the method, random tree incrementally expands to-
wardboundariesduringthe exploration process bysampling
viewpointsasnewnodes.Thenewlyaddednodesin theran-
domtree are then evaluated with information gain and tra-
versingcostas
s(x)=g(x)·exp(−λ·c(x)) (10)
g(x)
x
c(x)
x
λ
where is the expected information gain in position ,
is thedistancecostbetween robotandposition ,and
denotesacoefficientthatcontrolsthepenaltyonthedistance
cost.By selecting the branch with maximum score, the first
edgeofthisnodeissetasthenextbestviewtonavigate.The
move_basenavigation module providedby Robot Operating
System (ROS)[37] is employed to calculate the shortest path
basedontheDijkstraalgorithm[38].Therobotfollowsthegen-
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–6 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
erated path to explore the environment gradually. Once the
targetpoint is reached, the next round of exploration plan-
ningcontinues.Thewholeprocessgetsrepeateduntilthero-
botcoversthewholeareaorfindstheexit.
4 Experimental Validation
ToexaminethefunctionalityofEarshakeranddemonstrateits
superiority,itwas sent toattend the first A-TEC champion-
ships in 2020 as experimental validation. The competition
was held by the government of Shenzhen in Guangdong,
Chinatofurtherenhancerobotictechniquesandseekindustri-
al opportunities[39]. In the finals of the championships, the
competitionwasdivided into five sessionsandall the teams
wererankedbasedontheirperformanceinthosesessions,in-
cludingtask difficulty, task completion, and time consump-
tion.Thefivesessions,inturn,weretraversingroughterrains,
clearingcumbersome obstacles and opening doors,climbing
up and down regular stairs, passing through signal blocked
areas,andsearching and rescuing in smoky indoor environ-
ments,asillustratedinFigure5.Specifically,passingthrough
signalblockedareasrequiredtherobotstoautonomouslynav-
igateinside a mazeandsearchforthe only exit,whileinthe
othersessionstherobotswereremotelycontrolledbyoperat-
ors in first person point of view from hundreds of meters
away.Thesediversesessionsexaminedthecapabilityofpar-
ticipatedrobotsinlocomotion,manipulation,perception,tele-
communication, etc.[40−42] Robust and consistent performance
inallsessionsbecame more important thanoutstanding per-
formanceinanysinglesession[43].
During the intense championships, 15 teams globally
entered the finals in total. Out of those teams, Earthshaker
tookthefirstplacewithascoreof109points,whereasthero-
bots from Tsinghua University and Chongqing University
tookthe second and thirdplaces with scoresof 79 and 70.5
points,respectively.
Compared to the Seeker robot from Tsinghua University,
theMIST-RobotfromChongqingUniversity,andmanyoth-
errobotsfromtherestteams,Earthshakerrealizedtransform-
ationofthetracked system for climbing stairsand improve-
mentoncumbersomeobstacleclearingcapabilityinthemost
economical way, the swing arm – dozer blade structure.
Earthshakerwonthecompetitionalsobythediversesensors
integratedintotherobotthatallowedtherobottoberobustly
teleoperated and even achieve autonomous navigation. The
followingsub-sections describe the performance of Earth-
shakerineachsessionofthefinals.
4.1 Traversing miscellaneous terrains
Thissession requiredtherobotto firsttraversea30m-by-3m
roughterrainthatcouldbecoveredbyrubbles, bricks, or ir-
regular concrete debris depending on the selected difficulty
by each team. Following that, the robot needed to pass
throughan area covered by large immobile obstacles, climb
up and down slopes of 36 degrees at most, and travel on a
bridge tilted to the side by 27 degrees. Even though these
taskswererelativelyeasy, they relied heavily ontherobots’
speedandagility.Becausetheroboticarmdidnotneedtobe
operatedduringthissession,the corresponding operator was
abletofly a DJI MavicUnmannedAerialVehicle(UAV)to
provideaglobalviewofthefieldfromabove,whichallowed
thebaseoperatortoplanoperationbeforehandandgreatlyre-
duced the time consumption. Benefited from the great
horsepower and well-designed suspension of the chassis,
Earthshaker performed excellently in these tasks and was
rankedthefirstplaceamongalltherobots.
Besidesthe aforementioned regular tasks, there were also
challengetasksinthissession,wheretherobotsneededtotra-
versemuddyterrainswithpotholes,flatterrainswithtrenches
of various widths, and pools filled with water of different
heights. Earthshaker accomplished these challenging tasks
successfully,asshown in Figure 6. Specifically,whenfaced
Fig. 5.OverviewofthesessionsofA-TECchampionships.
Zhangetal.
–7 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
with the trenches, Earthshaker put down the swing arm –
dozerblade to increasethe body lengthof the chassis. As a
result,itcrossedthetrenchwithawidthof600mm.Aswith
thewaterpools,becausethewholebodyofEarthshakerwas
waterproofofthelevelIP64andthechassiswaseven water-
proofof the level IP66, Earthshaker was capable of dealing
withthepoolwithwaterdepthof500mm.Itisworthnoting
thatEarthshakerprepared for thepossiblerainyweather dur-
ingthe competition,whereasmanyotherrobots didnothave
thispreparation.Consequently,somerobotssufferedfromthe
rainy weather with their naked electronic interfaces, and
endedupnotbeingabletofinishthecompetition.
4.2 Approaching buildings
In this session, the robots were required to first clear up a
10m-by-20mareaby moving obstacles todesignated places,
andthen open and enter a door with automatic closers. The
obstacles included hollow steel tubes as light as 5 kg, and
steelbeams and concrete blocks as heavy as 50 kg. Earth-
shaker successfully utilized the dozer blade to push all the
obstaclestothetargetpositions.
Thereweremultipledifficultylevelsfordooropening,with
differenttypes of doors and door handles. Options are uni-
foldorbifold doors withspherical handles, L-shape handles
orvalves.Themostchallengingcombination,a unifold door
witha spherical handle, was selected for Earthshaker in the
competition.Becauseofthedoorcloser,therobotneededto
rotatethehandleandmaintain therotationwhileopeningthe
door.Asaresult,thetwooperatorsneededtocooperateinthe
process. One operator needed to first align the 0.8 m wide
Earthshakerwiththe1m wide door frame under the help of
the equipped laser pointers, and then keep commanding the
baseto moveforwardslowlyas thedoorhandlewas rotated,
untilthefrontendofthechassiswaspushedagainstthedoor
andthehandlecouldbereleasedbythegripper.Theotherop-
erator needed to fine tune the robotic arm and the gripper
aftertheinitial semi-autonomous manipulation, grip and ro-
tatethedoor handle as thechassiswasapproachingthedoor
untilthehandlecouldbereleasedfromthegripper.Figure7(a-
e)showssomesnapshotsofthewholeprocessofthissession.
Earthshaker finished this session within 31 minutes and 12
seconds.
4.3 Manipulation inside buildings
Robotsinthissession needed to climb up to and downfrom
theplatform as shownin Figure 7(f-h). Optional wayswere
throughvertical laddersorregularstairs.The trackedchassis
determinedthatEarthshakercouldonlypicktheregularstairs,
which was the common choice among all the robots in the
competition.The stairshad 24steps one-way,every step of
whichhadadepthof300mmandaheightof175mm.Thus,
theinclinationanglewasabout34degrees.Therewasaturn-
ingplatformbetweentwosectionsofstairs.
When Earthshaker was climbing up the stairs, the swing
arm– dozer bladestructure was adjustedto provide enough
contact length for the chassis and help the robot move
smoothly.However,theswingarmwasnotputfullyflatdue
to the detrimental friction generated by the passive arm
tracks,whichwouldhindertherobot from thrusting upward.
Theangleoftheswingarmwasempiricallysettojustenough
tosupporttherobottoclimbupthestairs.Ontheotherhand,
whentherobot was climbing down the stairs,theswingarm
couldbe putfullyflatto takeadvantageofits lengthandthe
passive friction generated to increase stability. Earthshaker
was able to finish this session within 6 minutes and 13
seconds,wheretheclimbingupprocesstookthemajorityof
theconsumedtime.Comparedtotheothersmall tracked ro-
botsinthecompetition,Earthshakerwasslowerduetoitsrel-
ativelycumbersomebodyonthestairway.
4.4 Autonomous navigation
Theautonomousnavigation session tested therobot’sintelli-
Fig. 6.SnapshotsofEarthshakerinSession1.(a-c)Earthshakerwaspassingthroughapoolfilledwithwaterof50cmindepth.(d)Earthshakerwastra-
versingmuddyterrains.(e)Earthshakerwascrossingatrenchof60cminwidth.
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–8 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
gence in building maps and finding exits within unknown
areaswithout human’shelp.Tosimulatethesituation ofsig-
nallossinreality,during the competition, the refereeturned
on the signal blocker once the robot entered the maze. The
operators inside the control room were also not allowed to
touchtheremotecontrollersduringthisperiod.Themazehad
threepossibleentrancesandthreepossibleexits.Whenthero-
botarrivedatthemaze,onlyoneentrancewouldbeopen,and
also only one exit would be usable. To be fair, inside the
maze,thereweremoveabledoorsthatwereadjustedforeach
robottoformadifferentunknownstructure.Earthshakerwas
abletoshowupattheexitwithin 41.13 seconds in this ses-
sion,rankedthesecondfastestamongalltherobots.Tocheck
thebuiltmapforthemaze,thepointcloudstoredintheNUC
hasbeenextractedafterthecompetition,asshowninFigure8.
Theleft halfof thefigure demonstrates the map built when
therobotjustenteredthemazefromthebottomleftentrance,
wherethewhitelineconnectsthe robot to its target position
on the far side of the maze. The right half of the figure
presentsthe map builtwhentherobotsuccessfully found the
exitonthetopleftandthepathitfollowedinthemazeindic-
atedbyaredsolidline.
4.5 Search and rescue in smoky environment
The last session of the competition involved indoor rescue
work. The robot was supposed to enter dense black smoke
filledroomsand searchforafiresource andawoundedper-
son.Thesmokewasreal,spreadbysomesmokegenerators.
However,thefiresourcewasrepresentedbyanelectricoven,
and the wounded person was actually a sand bag in human
shape.Thedummyweighedabout50 kg. To simulate a real
person,clotheswere putonforthedummy thatcouldgener-
ateheatforaperiodoftime.Therewereintotaleightsimilar
rooms.The fire source and the wounded person were ran-
domlydistributed among them. There were also other com-
mon items like tables, chairs, cabinets, etc., inside those
rooms,justlike regular roomspeople can find intheir daily
life.Therobotneededto find the fire source andturnitoff,
andalsoneededtofindthewoundedpersonandcarryitout
oftheroomtoadesignatedarea.Thesmokewasquitedense
andthevisible distance was lessthanhalfameterinsidethe
rooms.Earthshakerhadtosearcheveryroomunderteleopera-
tiontolocatethe wounded person andtheovenwithtwoin-
fraredcameras,thenuse the gripper to turntheovenoffand
carrythewounded person out. This again required coopera-
tionbetweenthetwooperators.Tocarrythewoundedperson
outoftheroom,acustomizedlassowasinstalledontoEarth-
shakerbeforesetoff. Once the woundedperson was located,
theroboticarmand gripper picked up thelassousing preset
controltrajectoriesandputthelassoaroundthewoundedper-
son’sarm through teleoperation. The lasso then automatic-
allylockeduponcetherobotstartedtodragthewoundedper-
son.Figure9showsscenesfromthissession.Earthshaker fi-
nallyspent11minutesand36secondsfinishingallthetasks.
4.6 Summary
Earthshakerperformedreasonablywellineachsessionofthe
competition, even ranked first in two of the five sessions.
That eventually allowed Earthshaker to take the first place
Fig. 7.SnapshotsofEarthshaker in Session 2&3. (a-b)Earthshakerwasclearing a light obstacleontheleftand a heavy oneontheright.(c-e) Earth-
shakerwasopeningaunifolddoorwithasphericaldoorhandle.(f-h)Earthshakerwasclimbingupanddownthestairs.
Zhangetal.
–9 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
among all the robots. The overall score table is shown in
Table 1. As a demonstration of dominance, Earthshaker got
109 points in the finals, whereas the runner-up only got 79
points.Earthshakerstoodoutbyitsmultimodalteleoperation,
itsmodular and waterproof mechatronic design, and suffi-
cient experiments and practice before the ultimate test. The
competitionrequiredacompleteand robustrescuerobotasa
whole, not just any advanced individual module of it.
However,some of the Earthshaker's shortcomings were re-
flectedinthecompetition.Theexcessivesizelimiteditsflex-
ibility of movement, making it hard to pass through certain
narrowspacesinactualuse.Atthesametime,thepayloadto
themanipulatorislimited, thus it cannot complete dexterous
manipulationtasks with large loads. Even though Earth-
shakerstillhadalotofroomtoimprove,itwastheexcellent
mechatronicintegration andtheadvancedcontrolphilosophy
thatmadeitthewinneroftheA-TECchampionships2020.
5 Conclusions
This paper introduces a rescue robot Earthshaker, including
Fig. 8.Themapbuiltforthemazefromthecompetition.Theredslimlinerepresentsthepaththattherobotfollowed.Thegrayareaindicatestheaccess-
iblepartofthemap,whilethecyanareaswithdarkredboundariesindicatetheinaccessibleparts.
Fig. 9.SnapshotsofEarthshakerinSession5.(a)Theinfraredthermalimageofthesmokyenvironment;(b-e)Thesceneandthestrategyusedtorescue
thedummy.
Table 1.FinalscoretableoftheA-TECchampionships2020
Team/Robot
Name
Traversing
miscellaneous
terrains
Approaching
buildings
Manipulation
insidebuildings
Autonomous
navigation
Searchandrescuein
smokyenvironment
Additional
challenge
Time
score Re-challenge Total
score
EarthShaker 25 12 12 18 25 7 10 0 109
Seeker
12
15 18 1 18 8 7 0 79
MIST-Robot
10
2 10 25 8 6.5 9 0 70.5
TeamofJingpin
18
10 25 10 12 8 6 -20 69
TeamofDream 8 25 8 0 15 8 8 -20 52
TeamofShentuo
Tech 6 4 4 4 4 6.5 5 0 33.5
FerociousLionof
Tsinghua 4 18 15 12 0 0 0 -20 29
TeamofWalkers 0 0 6 8 1 0 0 0 15
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–10 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
the system integration and the control algorithms of it. The
performanceof the robothas been evaluatedto be excellent
duringtheA-TECroboticchampionshipsin2020.Theunique
swing arm – dozerblade structure of Earthshaker helps ex-
tendsthecapabilityofconventionaltrackedchassis,improv-
ingitsperformanceoncumbersomeobstacleclearingandreg-
ular stair climbing. The multimodal teleoperation system
providestherobotredundancyandrobustnesswhentheoper-
atorscannotshowuponsite.Thefiniteautonomyintheoper-
ationofthe roboticarmandgripperhelps releasetheoperat-
ors’workburdentoasuitableextent.Whenteleoperationsig-
nalsarelost,therobotcouldalsoentertheautonomousnavig-
ationmode to search for anexit by itself and giveback the
controlauthoritytotheoperators.Overall,the championship
that Earthshaker earned has shown the efficacy of all the
aforementionedefforts.Itcan play a huge role insearchand
rescue in disaster scenarios such as nuclear accidents, toxic
gasleaks, and fires, where human workers cannot be de-
ployedduetoradiation, dangeroftoxiccontaminationorar-
chitecturecollapse. Future efforts canbe put into improving
therobot’sautonomyinmanyforeseeabletasksforemergen-
ciesanddisasterstofurtherincreaseitsefficiencyandrobust-
ness. More earthshaking endeavors in helping the human
communitycanbeexpectedfromEarthshaker.
Acknowledgments
This work is supported by the National Natural Science
Foundation of China (No. U21A20119, No. 62103395) and
the championship prize funded by Shenzhen Leaguer Co.,
Ltd.TheresearchofWGisalsosupportedinpartbytheFun-
damentalResearchFundsfortheCentralUniversities.
Conflict of interest
Theauthorsdeclarethattheyhavenoconflictofinterest.
Biographies
Yu ZhangYuZhang received the BEdegreeinengineering from the
UniversityofScienceandTechnologyofChina(USTC)andiscurrently
aMasterstudentintheBio-Inspired Robotics Laboratory in the USTC.
Hiscurrentresearchinterestsincludesystemdesignforspecialrobotsand
leggedrobots.
Yuxiang LiYuxiang Li received the MS degree in Instrument and
MeterEngineeringfromZhengzhouUniversity.He iscurrently pursuing
hisPhD degreeinHarbinInstituteof TechnologyShenzhen.His current
researchinterestsincluderoboticsandartificialintelligence.
Wei GaoiscurrentlyanassociateresearchfellowintheDepartmentof
PrecisionMachinery and Precision Instrumentationat the University of
Scienceand Technology of China(USTC).Hereceivedhis Bachelor of
Engineeringdegree from the Departmentof Mechanical Engineering at
NorthwesternPolytechnicalUniversityin2011, andhis DoctorofPhilo-
sophydegreefromtheDepartmentofMechanicalEngineeringatFlorida
StateUniversity(FSU)in2019.Hewasapostdoctoralresearchfellowin
FSUfrom2019to2020,andinUSTCfrom2020to2022.Hiscurrentre-
searchfocusesondynamiccontrolofmobilerobots.
Haoyao Cheniscurrentlya Professor in Harbin Institute of Techno-
logyShenzhen andtheState KeyLaboratoryof RoboticsandSystem of
China.He receivedtheBachelor’ sdegreeinMechatronics andAutoma-
tion from the University of Science and Technology of China in 2004,
andthe PhDdegreeintheRobotics fromboththe UniversityofScience
andTechnologyofChinaandtheCityUniversityofHongKongin2009.
Hewasworking asavisitingscholarintheAutonomousSystemsLabin
ETHz,Switzerland.Hisresearchinterestsincludeaerialmanipulationand
transportation,roboticperceptionandcognition,multi-robotsystems.
Shiwu Zhangis currently a professor in the Department of Precision
MachineryandPrecisionInstrumentation,USTC.HereceivedhisB.Eng.
degree in Mechanical and Electronic Engineering from USTC in 1997,
and his Ph.D. degree in the Precision Instrumentation and Machinery
from USTC in 2003. He has been a visiting scholar in University of
Wollongong,Australia in2016andin theOhiostate university,USAin
2012,respectively.His researchinterestsincludeamphibiousrobots,soft
robots,leggedrobots,liquidmetalrobotsandrescuerobots.
References
DEGREEFFJ,MIOCHT,VANVUGHTW,etal.Persistentrobot-
assisted disaster response[C]. Companion of the 2018 ACM/IEEE
International Conference on Human-Robot Interaction, 2018: 99-
100.
[1]
MATSUNOF,SATON,KONK,etal.Utilizationofrobotsystems
indisastersitesofthegreateasternjapanearthquake[C].Fieldand
servicerobotics,2014:1-17.
[2]
QUERALTA J P, TAIPALMAA J, PULLINEN B C, et al.
Collaborativemulti-robotsearchandrescue:Planning,coordination,
perception,andactivevision.Ieee Access,2020,8:191617–191643.
[3]
NAGATANI K, KIRIBAYASHI S, OKADA Y, et al. Emergency
responseto the nuclear accidentat the Fukushima Daiichi Nuclear
PowerPlantsusingmobilerescuerobots.Journal of Field Robotics,
2013,30(1):44–63.
[4]
DELMERICOJ,MINTCHEVS,GIUSTI A,etal.Thecurrentstate
and future outlook of rescue robotics. Journal of Field Robotics,
2019,36(7):1171–1191.
[5]
ATKESONC G, BABUBPW, BANERJEEN,etal. No falls,no
resets: Reliable humanoid behavior in the DARPA robotics
challenge[C]. 2015 IEEE-RAS 15th International Conference on
HumanoidRobots(Humanoids),2015:623-630.
[6]
FENGS,WHITMANE,XINJILEFUX,etal.Optimization‐based
full body control for the darpa robotics challenge. Journal of field
robotics,2015,32(2):293–312.
[7]
SPENKOM,BUERGERS,IAGNEMMAK.TheDARPArobotics
challengefinals: humanoid robots to the rescue[M]. 121. Springer,
2018.
[8]
SHEHR, SCHWERTFEGER S,VISSERA.16 YearsofRoboCup
Rescue.KI-Künstliche Intelligenz,2016,30(3):267–277.
[9]
KARUMANCHI S, EDELBERG K, BALDWIN I, et al. Team
RoboSimian: semi‐autonomous mobile manipulation at the 2015
DARPA robotics challenge finals[J]. Journal of Field Robotics,
2017,34(2):305-332.
[10]
SCHWARZ M, BEUL M, DROESCHEL D, et al. DRC team
nimbro rescue: perception and control for centaur-like mobile
manipulation robot momaro [M]. The DARPA robotics challenge
finals:humanoidrobotstotherescue.Springer.2018:145-190.
[11]
STENTZ A, HERMAN H, KELLY A, et al. CHIMP, the CMU
highlyintelligent mobileplatform.Journal of Field Robotics,2015,
32(2):209–228.
[12]
HUTTER M, GEHRING C, LAUBER A, et al. Anymal-toward
leggedrobots forharshenvironments. Advanced Robotics,2017,31
(17):918–931.
[13]
KRUIJFF-KORBAYOVá I, FREDA L, GIANNI M, et al.
Deployment of ground and aerial robots in earthquake-struck
amatrice in italy (brief report)[C]. 2016 IEEE international
symposium on safety, security, and rescue robotics (SSRR), 2016:
[14]
Zhangetal.
–11 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted
278-279.
Autonomous robot for gas and oil sites challenge [EB/OL]. http://
www.argos‐challenge.com/en/challenge.
[15]
TAUROB‐ARGOSwinner[EB/OL].http://taurob.com/text‐argos‐
gewinner/.
[16]
ENDOD,NAGATANIK.Assessmentofatrackedvehicle’sability
totraversestairs.ROBOMECH Journal,2016,3(1):1–13.
[17]
YAMAUCHI G, NAGATANI K, HASHIMOTO T, et al. Slip-
compensatedodometryfortrackedvehicleonlooseandweakslope.
Robomech Journal,2017,4(1):1–11.
[18]
ROUČEKT,PECKAM,ČíŽEK P, et al. System for multi-robotic
explorationof undergroundenvironmentsCTU-CRAS-NORLAB in
the DARPA Subterranean Challenge. arXiv preprint arXiv:2110,
0591,1:2021.
[19]
AGHAA,OTSUK,MORRELLB,etal.Nebula:Quest forrobotic
autonomy in challenging environments; team costar at the darpa
subterraneanchallenge.arXiv preprint arXiv:2103,1147,0:2021.
[20]
CHENX, ZHANGH,LU H,etal. RobustSLAMsystem basedon
monocular vision and LiDAR for robotic urban search and
rescue[C].2017 IEEE International SymposiumonSafety,Security
andRescueRobotics(SSRR),2017:41-47.
[21]
ROUČEK T, PECKA M, ČíŽEK P, et al. Darpa subterranean
challenge: Multi-robotic exploration of underground environ-
ments[C].InternationalConferenceonModellingandSimulationfor
AutonomousSystems,2019:274-290.
[22]
TRANZATTO M, MASCARICH F, BERNREITER L, et al.
Cerberus:Autonomouslegged and aerial robotic exploration in the
tunneland urbancircuitsofthedarpa subterraneanchallenge.arXiv
preprint arXiv:2201,0706,7:2022.
[23]
SCHWARZ M, RODEHUTSKORS T, DROESCHEL D, et al.
NimbRoRescue:Solvingdisaster‐responsetaskswith the mobile
manipulation robot Momaro. Journal of Field Robotics,2017, 34
(2):400–425.
[24]
LIY, LIM,ZHUH, etal.Development andapplicationsofrescue
robots for explosion accidents in coal mines. Journal of Field
Robotics,2020,37(3):466–489.
[25]
LöSCHR,GREHLS,DONNERM,etal.Designofanautonomous
robot for mapping, navigation, and manipulation in underground
mines[C]. 2018 IEEE/RSJ International Conference on Intelligent
RobotsandSystems(IROS),2018:1407-1412.
[26]
SZREK J, ZIMROZ R, WODECKI J, et al. Application of the
infrared thermography and unmanned ground vehicle for rescue
action support in underground mine —The amicos project. Remote
Sensing,2020,13(1):69.
[27]
BHATIA R, LI L. Throughput optimization of wireless mesh[28]
networks with MIMO links[C]. IEEE INFOCOM 2007-26th IEEE
InternationalConferenceonComputerCommunications,2007:2326-
2330.
PEARSONK. LIII. Onlinesand planes ofclosestfit to systemsof
pointsinspace. The London, Edinburgh, and Dublin philosophical
magazine and journal of science,1901,2(11):559–572.
[29]
MURRAYRM,LI Z, SASTRY S S. A mathematical introduction
toroboticmanipulation[M].CRCpress,2017.
[30]
KRIEGEL H P, KRöGER P, SANDER J, et al. Density‐based
clustering. Wiley interdisciplinary reviews: data mining and
knowledge discovery,2011,1(3):231–240.
[31]
ZHANGJ,SINGHS.LOAM:LidarOdometryandMappinginReal-
time[C].Robotics:ScienceandSystems,2014:1-9.
[32]
BESL PAUL J, MCKAY N D. A method for registration of 3-D
shapes. IEEE Transactions on Pattern Analysis and Machine
Intelligence,1992,14(2):239–256.
[33]
FORSTER C, CARLONE L, DELLAERT F, et al. IMU
preintegrationon manifold for efficient visual-inertial maximum-a-
posterioriestimation[C],2015.
[34]
SIMANEK J, REINSTEIN M, KUBELKA V. Evaluation of the
EKF-basedestimationarchitecturesfordatafusioninmobilerobots.
IEEE/ASME Transactions on Mechatronics,2014,20(2):985–990.
[35]
BIRCHERA,KAMELM,ALEXISK,etal.Recedinghorizon"next-
best-view" planner for 3d exploration[C]. 2016 IEEE international
conferenceonroboticsandautomation(ICRA),2016:1462-1468.
[36]
move_base- ROS Wiki [EB/OL].(2020)[http://wiki.ros.org/move_
base.
[37]
EDSGER D, MISA T J. An interview with edsger w. dijkstra.
Communications of the ACM,2010,53(8):41–47.
[38]
A-TEC Official Website [EB/OL]. (2020)[https://atec.leaguer.com.
cn/index/index/championshipsjj.
[39]
DEPETRISP, NGUYEN H, DHARMADHIKARIM,etal.RMF-
Owl: A Collision-Tolerant Flying Robot for Autonomous
Subterranean Exploration[J]. arXiv preprint arXiv: 2202.11055,
2022.
[40]
HUDSON N, TALBOT F, COX M, et al. Heterogeneous Ground
and Air Platforms, Homogeneous Sensing: Team CSIRO Data61's
Approach to the DARPA Subterranean Challenge. arXiv preprint
arXiv:2104,0905,3:2021.
[41]
OTSU K, TEPSUPORN S, THAKKER R, et al. Supervised
autonomyfor communication-degraded subterraneanexplorationby
arobotteam[C].2020IEEEAerospaceConference,2020:1-9.
[42]
OHRADZANSKYMT,RUSHER,RILEYDG,etal.Multi-agent
autonomy: Advancements and challenges in subterranean
exploration[J].arXivpreprintarXiv:2110.04390,2021.
[43]
Earthshaker:AMobileRescueRobotforEmergenciesandDisastersthroughTeleoperationandAutonomous
Navigation Zhangetal.
–12 DOI:10.52396/JUSTC-2022-0066
JUSTC,2022,52(X):
Just Accepted