3202
peS
62
]OR.sc[
1v78641.9032:viXra
ON THE EFFECTS OF THE VARIATIONS IN NETWORK CHARACTERISTICS
IN CYBER PHYSICAL SYSTEMS
G´eza Szabo´, Sa´ndor Ra´cz J´ozsef Pet˝o
Ericsson Research, Budapest, Hungary Budapest University of Technology and Economics, Hungary
Email: {geza.szabo,sandor.racz}@ericsson.com Email: pjoejoejoe@gmail.com
Rafael Roque Aschoff
Pernambuco Federal Institute of Education, Science, and Technology, Brazil
Email: rafael.aschoff@gprt.ufpe.br
KEYWORDS izedlocationse.g.,possiblegraspingposesofvarious3D
network characteristics,cyber-physics, Gazebo objects.
Another aspect of cloud robotics is the way in which
ABSTRACT
therobotcontrolrelatedfunctionalityismovedintothe
cloud. Thesimplestwayistoruntheoriginalrobotspe-
The popular robotic simulator, Gazebo, lacks the fea- cifictaskinacloudwithoutsignificantchangesinit. For
ture of simulating the effects of control latency that example, in a Virtual Machine (VM), in a container, or
would make it a fully-fledged cyber-physical system in a virtualized ProgrammableLogic Controller (PLC).
(CPS) simulator. The CPS that we address to mea- Anotherwayistoupdate,modifyorrewritethecodeof
sure is a robotic arm (UR5) controlled remotely with robot related tasks to utilize existing services or APIs
velocity commands. The main goalis to measure Qual- ofthe cloud. The third way is to extend the cloudplat-
ity of Control (QoC) related KPIs during various net- form itself with new features that make robot control
work conditions in a simulated environment. We pro- more efficient. These new robot-aware cloud features
pose a Gazebo plugin (OurPlugin 2017) to make the can be explicitly used by robot related tasks (i.e. new
above measurement feasible by making Gazebo capable robot-awareservicesorAPIsofferedbycloud)orcanbe
to delay internal control and status messages and also transparent solutions (e.g., improving the service pro-
to interface with external network simulators to derive videdbythecloudtomeettherequirementoftherobot
even more advanced network effects. Our preliminary control).
evaluationshows that there is certainly an effect onthe Designingcyber-physicalsystemsischallengingbecause
behavior of the robotic arm with the introduced net- ofa)the vastnetworkandinformationtechnologyenvi-
work latency in line with our expectations, but a more ronmentconnectedwithphysicalelementsinvolvesmul-
detailed further study is needed. tiple domains such as controls, communication, analog
anddigitalphysics,andlogicandb)theinteractionwith
INTRODUCTION
thephysicalworldvarieswidelybasedontimeandsitu-
ation. ToeasethedesignofCPS,robotsimulatorshave
A cyber-physical system (CPS) is a mechanism con- been used by robotics experts. A well-designed simula-
trolled or monitored by computer-based algorithms, tor makes it possible to rapidly test algorithms, design
tightlyintegratedwiththeinternetanditsusers. Unlike robots, perform regression testing, and train AI system
more traditional embedded systems, a full-fledged CPS using realistic scenarios.
istypicallydesignedasanetworkofinteractingelements Therearevariousalternatives,setsoftoolsthatmakeit
withphysicalinputandoutputinsteadofasstandalone possibletoputtogetheraCPSsimulationenvironment,
devices. For tasks that require more resources than are but it is very difficult, needs a lot of interfacing with
locally available, one common mechanism is that nodes various tools and impractical. The requirements of a
utilizethenetworkconnectivitytolinkthesensororac- widely applicable CPS are the following:
tuator part of the CPS with either a server or a cloud
environment,enablingcomplexprocessingtasksthatare • Should be modular in terms of interfacingwith the
impossible under localresourceconstraints. Among the CPS
wide diversity of tasks that CPS is applied we focus on
robot control in this paper. • Shouldbemodularintermsofinterfacingwithnet-
Currently,oneofthemainfocusofcloudbasedrobotics work simulator, realization environment
is to speed up the processing of input data collected
frommanysensorswithbigdatacomputation. Another • Should be able to cooperate with widely applied
approachistocollectvariousknowledgebasesincentral- environments