lund september 2010 robotics and automation software€¦ · lund september 2010 robotics and...
TRANSCRIPT
Lund September 2010
Robotics and Automation software
[email protected], et.al., Lund University, Sweden
cs.LTH.se/RSS
Outline
1. Intro and projects
2. SMErobot results
3. Flexible software systems
4. Rosetta and future robots
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 2
Robotics Observations & Motivations
1. We humans want safe and sustainable (assistance for) a good quality of life !
2. Resources are limited => Efficiency !
3. The world is changing => Flexibility !
• Sustainable quality of life:
– Transformations of natural resources
(robot-assisted manufacturing and recycling)
– Scalable solutions required
– Human aspects….
• Efficiency:
– Resource-aware implementations
– Reuse in engineering
• Flexibility:
– Interoperable and open systems? No, not sufficient in robotics!
– Applications and technologies evolve rapidly: New approach needed…..
• Safety and Security: Key issues also calling for new FLEXIBLE approaches…
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 3
New/running robot projects
• MONROE – Hyper-Modular Open Networked RObot systems with Excellent performance$: EU.FP7.ICT, within www.echord.info, with Güdel AG, Switzerland.
• COMET – Plug-and-produce COmponents and METhods for adaptive control of industrial robots enabling cost effective, high precision manufacturing in factories of the future.$: EU.FP7.NMP/FoF, coordinated by Delcam PLC Ltd., UK
• ROSETTA - RObot control for Skilled ExecuTion of Tasks in natural interaction with humans; based on Autonomy, cumulative knowledge and learning$: EU.FP7.ICT, coordinated by ABB Corporate Research, Västerås
• ENGROSS – Enabling Growing Software Systems$: SSF with LUCAS&Math partners @LTH, mobile robots
• ProFlexA – Productive Flexible Automation$: SSF ProViking, foundry applications, informally with ABB Robotics, Västerås
• INROSY – Intelligent Networked RObotics SYstems with reconfigurable exogenous system sensing. $: STINT KOSEF, with Hanyang university, Seoul, Korea
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 4
„Re-Inventing Industrial Robotics“
“established”Vision: The „SME-Robot“
Automotive industries
Changeover time < once/year
Programming offline (>90%)
Robot unit price ~ 25 T€
Workcell cost ~ 4 * robot unit price
Sensor equipped < 5% of installations
Planning simultaneous engineering
Maintenance trained staff
< once/day
• shop-floor
• 1/3 of today‘s robot price
• ~ 1 * robot unit price
• 100%
• shop floor
• by worker
Manufacturing SMEs
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 5
SMErobot video
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 6
Example industrial collaborationULUND-ABB work (F/T control)
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 7
EURON TechTransfer Award 05: Open platformfor skilled motions in productive robotics
Work
Robot
Language
Mas
ter
Contr
ol
Motor
control
Trajectory
generation
Arm
control Arm
Application ……
Task description
ABB controller
Cel
l co
ntr
oll
er
RPL/XMLPathsync
250Hz<1ms5Hz
F/T-sens.8kHz
← kernel space←
VxWorks
Linux
Windows(any OS)
userspace
The European Robot Initiative for Strengthening the Competitiveness of
SMEs in Manufacturing
Klas Nilsson [Martin Hägele]www.smerobot.org
Outlook and Lessons learned from SMErobot™
Klas Nilsson, Lund University – LTH, Sweden; [email protected]
ICRA2010, Anchorage, 3 May 20109Outlook and lessons learned in SMErobot™
Project-Partner
Project-lead
Robotics-industries
End-users
Suppliers
Transfer
Research
ICRA2010, Anchorage, 3 May 201010Outlook and lessons learned in SMErobot™
SMErobot: A Family of New Robots
Three Major Innovations:
1. Robot capable of understanding human-like instructions
2. Safe and productive human-aware space-sharing robot
3. Three-day-deployable integrated robot system
• FP6 Integrated Project• 17 partners, major European robot manufacturers• Project runtime March 2005 - May 2009• Coordinator: Fraunhofer IPA, Stuttgart
ICRA2010, Anchorage, 3 May 201011Outlook and lessons learned in SMErobot™
The SMErobot Initiative
Intuitive instruction of fettlingof castings for the foundry
Fast installation, small batchsize production change (forgery)
The SME welding robot
Automation of manualwoodworking processes
Robot capable ofunderstanding human-likeinstructions
Safe and productive human-aware space-sharing robot
Three-day-deployableintegrated robot system(install-configure-instruct)
• Training and education
• Socio-economics (new business models, LCC)
• Standardization
• Exploitation, IPR
Research & Development Demonstrations (Focus)
InnovationRelatedActivities
2010-09-15 12Outlook and lessons learned in SMErobot™
Industrial research perspective
Present Robot Applications
Future Robot Applications
Future Robot Product Development
Present Robot Product Development
Industry Segments
R&D in Robotics
New Technology
Interplay and feedback from other areas, also from present robot applications…
2010-09-15 13Outlook and lessons learned in SMErobot™
User Interface
Intelligent gripper
Bin picking process
What isshown:
“the objects”
What was invented forintegrating “the
objects”
Robot guide device
XIRP
Comm.
protocol
UPnP
Comm.
protocol
Config
.module
PC -Cell
contro
ller
Research
End User
Industrial
Industry linking Science and Users
• An example from P’n’P demonstration:
2010-09-15 14Outlook and lessons learned in SMErobot™
The SMErobot InitiativeInnovations:1. The robot capable of understanding human-like instructions2. The safe and productive human-aware space-sharing robot3. The three-day-deployable integrated robot system
Applications in SME manufacturing Research & DevelopmentImplementation
• Training and
Education
• Socio Economics
• Dissemination
• Exploitation
• Result exploitation
• Intellectual
property rights
• Management
Demonstrations
Coo
pera
tive
flexi
ble
Rob
ot W
orkc
ell
Intu
itive
Inst
ruct
ion
of F
ettli
ngC
astin
gs
Rob
ot a
ssis
tant
at
man
ualw
orkp
lace
s
Objectives
The
robo
t un
ders
tand
ing
hum
an-li
kein
stru
ctio
ns
The
safe
hum
an- w
are
spac
esh
arin
gro
bot
The
thre
e-da
yde
ploy
able
inte
grat
edro
bot s
yste
m
Innovations
Fou
ndry
Co-
oper
ativ
e sm
all
batc
has
sem
bly
Met
al p
rodu
cts
Mac
hine
ry
Woo
dwor
king
SM
E e
nd
- use
rg
rou
pS
ME
sys
tem
inte
gra
tors
Science and technology (S&T) progress beyond state of the art urgently needed!
Interplay between core sciences and SME needs, by integration project SMErobot…
2010-09-15 15Outlook and lessons learned in SMErobot™
Innovations – beyond state of the art
1. Robot capable of understanding human-like instructions
# New Low-Cost Human-Robot-Interaction (HRI)
→ Declarative knowledge for interaction fusion
2. Safe and productive human-aware space-sharing robot
# New Mechatronics for High-Performance Robotics
→ Safety with awareness of human intension
3. Three-day-deployable integrated robot system
# Increased modularity & Managed geometries/motions
→ Agile technology upgrading; Technologies and
SME-needs change faster than standards evolve
Legend: # =accomplished, → = results but deserves further research
ICRA2010, Anchorage, 3 May 201016Outlook and lessons learned in SMErobot™
Intuitive Instruction
New interaction devices and their use
Tactile guidanceGraphicsSpeech
ICRA2010, Anchorage, 3 May 201017Outlook and lessons learned in SMErobot™
Safe, Human-aware Space-sharing
Physically harmless and low-cost robot mechanics
Safe robot working without fences
ICRA2010, Anchorage, 3 May 201018Outlook and lessons learned in SMErobot™
The “3-day deployable” robot workcell
“Plug-and-Produce” Robot program generation bydistributed product-process data
Ethernetnetwork
Ethernetnetwork
ICRA2010, Anchorage, 3 May 201019Outlook and lessons learned in SMErobot™
Training, Socio-Economics
2010-09-15 20Outlook and lessons learned in SMErobot™
Conclusions
• Force control and lead-though programming
established for usage in actual products:
– Worked (for ideal cases) in labs 20 years ago,
– was difficult to fit into industrial systems,
– SMErobot enhancements and proof of concept in SMEs!
• Wide variety of competences needed for S&T
progress; SME business models needed for impact!
• Very successful barrier breakthroughs (S&T
results),
but industrial impact not completed!
• SMErobot as a model and inspiration for extended
Robotics and Automation research….
ICRA2010, Anchorage, 3 May 201021Outlook and lessons learned in SMErobot™
Video and Questions
• The SMErobot vision:The coffee-break video [SMErobot on YouTubeor on smerobot.org]
• Summary of results: This video……
www.smerobot.org
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 22
Robot control: R&D Situation
Problems (not just to write another algorithm/procedure):
• Reuse not even the case within a lab
• Few shared platforms, hampering scientific values
• Industrial development re-implementing controls
• Public funding goes to re-implementation
• Technologies missing for efficient reuse
q Current situation hampers the entire area of robotics….
Approach:
• Separation of concerns
• Multiple reuse perspectives (not layers)
• Middleware and frameworks in thin layers
• Bottom-up definitions and implementations
• Separation of mechanism and policy
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 23
Interoperable and Open Systems
Interoperability
• The standard driving force for standardization
• Very valuable when it works (USB, IP, SCXML, etc., etc.)
– Evolving apps and techs: “Standards playing catch-up”
Open Systems
• Public interfaces/APIs
• Modules (using those APIs) replaceable by third parties
– Fixed APIs/abstractions and data/control-flows: Not evolvable!
Standardization based on local attempts to reuse:
q We built a system X according to the architecture Y including principles Z. It worked very well, and hence we conclude that everyone should use XYZ as the standard robot control platform.
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 24
Interoperability
Interoperability by standardization [NIST.gov]:
• Interface standards and specifications that are
– Correct
– Complete
– Unambiguous
– End-user directed
– Vendor written
• Implementations that are
– Compliant
– Interoperable
– Adopted broadly and worldwide
Insight: Takes too long to agree in a global and evolving world![Compare mobile telecommunication applications]
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 25
Technical Insights / Expert comments
• Reuse: Source code, binary components, algorithms, data, definitions;
different perspectives.
• Primary separation of concerns:
– Coordination
– Configuration
– Computation
– Communication
• Loose coupling: In code and configurations, but also in assumptions,
definitions, service interaction, etc.
• Technology slicing; keep standards small and replaceable.
• Open source and proprietary systems deserve special attention.
• Service-oriented message-based interfaces to functionality.
• Compositionality; no need to know if components are composed.
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 26
Standards, modularity, and applications
• Market demands
packaged products.
• Useful components key
for time to market, or to
enable research focus,
• Price, performance,
maintainability, etc
based on tradeoffs and
optimization.
• Product development
and research need
engineering principles
in common
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 27
Robot System Architectures
Reflecting:
• Product lines
• Application areas
• User interaction
• Technologies
• Resource tradeoffs
• Variation points
• Implicit knowledge
• Business cases
• Certification
• Engineering taste
Taste/architecture is personal!
Thus, standardize components/techniques for robot architectures!
Product line architectures:
Domainunderlyingknowledge,human needs,domaincharacteristics
Business Finance,
accounting,marketing,
sales
TechnologyGeneric Tools,OTS apps,
computing/communicationsinfrastructure
CoreCompetencies
Application-Family
Architecture
Domain-IndependentInfrastructure
Idealized/context-non-specific knowledgeand architecture, not
shaped/driven/informedby business insights
An organization’sdomain-independent
technical assets
Domain expertise and knowledgethat is not captured or implemented
Domain-Specific Engineering
Product-LineArchitectures
Domainunderlyingknowledge,human needs,domaincharacteristics
Business Finance,
accounting,marketing,
sales
TechnologyGeneric Tools,OTS apps,
computing/communicationsinfrastructure
CoreCompetencies
Application-Family
Architecture
Domain-IndependentInfrastructure
Idealized/context-non-specific knowledgeand architecture, not
shaped/driven/informedby business insights
An organization’sdomain-independent
technical assets
Domain expertise and knowledgethat is not captured or implemented
Domain-Specific Engineering
Product-LineArchitectures
From http://csse.usc.edu/~neno
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 28
ENGROSS: Platform and Mobile Manipulation
• DLR, KUKA, Schunkarms depicted here.
• We will use future ABBlight-weight robot arms
• Mobile platform fromwww.care-o-bot.de(mid and right pict.)
Justin:
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 29
ENGROSS HW
Classified ABB technologies formanipulation gohere as body &arms [Rosetta]
Modules
of current
platform
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 30
ENGROSS: The safety-flexibility tradeoff
The extremes are less appropriate….
LUCAS: Robotics and Semantic Systems, Lund, Sept. 2010 31
Force and vision – reaction time crucial
• Applications….
External Sensing and Control – Reconfiguration of ABB Industrial Robot Controllers, ICRA2010 32
Integrated High- and Low-level Control
Computer(Pentium)
PanelSafety
Power &UPS
PowerDriveSafety
Computer(PowerPC)
Drives(DSP)
Sensor/Process Interfaces
LAN
Control module
Drive modules
SMB
Motor
Perception, safety-watch, filtering, …
Engineering system
Skill-level ctrl
Soft/no RT
Hard RT ext_ctrl ext_ctrl
IRC5
RT ctrl
Skill exec
Task &WM
Task & WM
Overallknowledge-
base
RobAPIClient
Virtualcontrol
Sensing
YourFunc.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 33
ROSETTA
RObot control for Skilled ExecuTion of Tasks in natural interaction with humans; based on
Autonomy, cumulative knowledge and learning
ROSETTA develops “human-centric” technology for industrial
robots that will not only appear more human-like, but also
cooperate with workers in ways that are safe and
perceived as natural. Such robots will be programmed in
an intuitive and efficient manner, making it easier to
adapt them to new tasks when a production line is changed to
manufacture a new product.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 34
Research Areas
• Intuitive ways of instructing the
robot
• Task-level instruction
• Knowledge and skill representation
• System support (engineering tool)
• Robot control
• Sensor integration
• Assembly operations
• Learning
• Skill-based architectures
• Semantic acquisition and
interpretation
• Safety
• Physical human-robot interaction
• Injury criteria
• Workspace supervision
Four major efforts§ Knowledge Integration Framework
§ Human-Centric Aspects
§ Safety Criteria
§ Robot Kinematics
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 35
Rosetta
RObot control for Skilled ExecuTion of Tasks in natural interaction with
humans; based on Autonomy, cumulative knowledge and learning
Research on:
• Natural Language Processing
• Knowledge
representation
• Machine vision
• Real-time control
• Learning
• Human-cooperative
robots
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 36
Overall Architecture
Engineering ToolStation Configuration Tool
Task Definition Tool
Simulation Environment
Knowledge Integration FrameworkKnowledge about devices
Knowledge about device capabilities (Skills)
Knowledge about injury criteria
ROSETTA ControllerRuntime Controller
Safety Watcher
Task Sequencing
Robot ControllerSafety Sensors
The engineering tool gets knowledge about devices from KIF and can feed back task descriptions The ROSETTA Controller gets the
Task Description from KIF and feeds back knowledge learned.
The ROSETTA Controller acts upon sensor input. In the project new safety sensors are developed.
The ROSETTA Controller tells the Robot Controller what to do, and reads sensor values from the Robot.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 37
Related or Prior Work in other Projects
• Baseline: The semantic web, a medium to exchange
data and knowledge
• DBPedia is a project that extracts in the form of RDF
triples and links RDF repositories
• Data sources can be accessed through a SPARQL
endpoint
• It served as main inspiration to the KIF
• The main repository of accessible knowledge in robotics
i the form of XML, databases (tables), text, etc., needs
conversion/triplification when imported.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 38
Wikipedia: Resource_Description_Framework (RDF)
The Resource Description Framework (RDF) is a family of World Wide Web
Consortium (W3C) specifications originally designed as a metadata data
model. It has come to be used as a general method for conceptual
description or modeling of information that is implemented in web
resources, using a variety of syntax formats.
A collection of RDF statements intrinsically represents a labeled, directed
multi-graph. As such, an RDF-based data model is more naturally suited
to certain kinds of knowledge representation than the relational model
and other ontological models. However, in practice, RDF data is often
persisted in relational database or native representations also called
Triplestores, or Quad stores if context (i.e. the named graph) is also
persisted for each RDF triple. As RDFS and OWL demonstrate, additional
ontology languages can be built upon RDF.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 39
Wikipedia: RDFS
RDFS: RDF Schema is an extensible knowledge representation language,
providing basic elements for the description of ontologies, otherwise called
Resource Description Framework (RDF) vocabularies, intended to structure
RDF resources.
rdf:type is a property used to state that a resource is an instance of a class.
rdfs:subClassOf allows to declare hierarchies of classes.
For example, the following declares that 'Every Person is an Agent':
foaf:Person rdfs:subClassOf foaf:Agent
Hierarchies of classes support inheritance of a property domain and range
(see definitions in next section) from a class to its subclasses.
rdfs:subPropertyOf is an instance of rdf:Property that is used to state that all
resources related by one property are also related by another.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 40
Predicate-argument structures for the Open world
An open world is one in which we must assume at
any time that new information could come to light,
and we may draw no conclusions that rely on
assuming that the information available at any
one point is all the information available.
Basis for the open (human, not co-engineered) world;
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 41
Architecture
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 42
XML (left) to RDF (right) Conversion
This RDF tree is additive such that it automatically forms a graph when another tree is imported or inferred.
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 43
Client Integration
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 44
Client Visualization
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 45
Browsing the Integration server
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 46
Conclusions
XML widely used in the normative/engineered world,
suitable for tree-structured data.
Semantic web and DBPedia our inspiration sources
RDF (serialized as XML or N3 or…) basis for the open
world, with triples as graph elements.§ Data additive, even when coming from unrelated sources
§ Predicate-argument structure for natural/human language semantics
§ Dedicated graphs for specific needs (performance, algorithms, etc.)
AutomationML import and presentation, as example of
incorporation of predefined context/application data.
Next: Supporting Learning, Control, Safety, …
based on the actual scenarios (data-first manner)…
ROSETTA @ ISR/Robotik 2010 http://www.fp7rosetta.euA KIF for Robotics FP7 ICT 230902 ROSETTA - Slide 47
Thank you for your attention!
Questions?
The research leading to these results has received funding from the European Community’s Seventh Framework ProgrammeFP7/2007-2013 – Challenge 2 – Cognitive Systems, Interaction, Robotics – under grant agreement No 230902 - ROSETTA.
Please visit our webpage for further information and contact information:
http://www.fp7rosetta.eu/
Klas Nilsson ([email protected])
http://rss.cs.lth.se