2020 Presentation at UM - TECoSA
Transcript of 2020 Presentation at UM - TECoSA
Human Robot Collaborative Assembly:The State of the Art and Latest Advancement
An Invited Talk on
Presentation Outline
2020-11-05 2
Background and definition of human-robot collaboration (HRC)
Safety standards and active collision avoidance in HRC assembly
Dynamic task planning and on-demand job dispatching
Adaptive and programming free robot control
In situ operator support in changing HRC assembly environment
Challenges and future research directions
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Why Human Robot Collaboration
2020-11-05 3
Key Requirements:
Safe environmentProgramming-freePlug-and-produceUser friendly
(Adapted from FP6 SME Robotics project’s survey on 482 SMEs)%
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 4
Scope of the Presentation
Adaptive Robot Control
Dynam
ic Task
Planning
Mobile W
orker A
ssistance
Active Collision Avoidance
Smart Sensor Network
Location-aware Task-specific Multimodal In-situ assistance Decision support
Intuitive support Portal for end users Dynamic task planning/re-planning
Supervisory control
IEC-61499 function blocks compliant Smart algorithms encapsulation Adaptation via event-driven behaviours Zero robot programming to end users Multimodal programming in real-time
3D models of structured environment Point-cloud images of human workers
SSaaffee HHRRCC EEnnvviirroonnmmeenntt
HHuummaann--RRoobboottCCoollllaabboorraattiioonn
R1
R2
R3
R4
R5
Symbiotic communication: Auditory Visual Haptic
R1
R2 R3
R4
Classification of H R Relationships
2020-11-05 5
Coexistence Interaction Cooperation Collaboration
Shared
Workspace
Direct contact
Working task
Resource
Simultaneous process
Sequential process
Reference Wang XV, Seira A, Wang L (2018) Classification, personalised safety framework and strategy for human-robot collaboration. Proceedings of CIE48, Vol. 3, pp. 2302–2311.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
2020-11-05 6
Charicteristics of HRC Assembly
The capability to track an operator’s position allows controlling a robot to support the operator as needed
Operator’s thoughts can be
converted to robot control commands
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Key Properties of HRC Assembly
2020-11-05 7
Multiplicity Initiative
Wang XV, Kemény Z, Váncza J, Wang L (2017) Human–Robot Collaborative Assembly in Cyber-Physical Production: Classification Framework and Implementation. CIRP Ann Manuf Technol 66(1):5–8.
Reference
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Definition of Symbiotic HRC
2020-11-05 8
Symbiotic human robot collaboration places the interplay of human and robot into a cyber-physical environment where human and robot interact in a shared work environment to solve complex tasks which require the combination of their best, complementing competencies.
A symbiotic HRC system possesses the skills and ability of perception, processing, reasoning, decision making, adaptive execution, mutual support and self learning through real-time multimodal communication for context-aware human-robot collaboration.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
L. Wang, R.X. Gao, J. Váncza, J. Krüger, X.V. Wang, S. Makris and G. Chryssolouris, "Symbiotic Human-Robot Collaborative Assembly," CIRP Annals – Manufacturing Technology, Vol.68, No.2, pp.701-726, 2019.
Reference
AI as the Foundation of HRC
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 9
1975:Genetic Algorithm, John Holland
1958:LISP – 1st AI Language,John McCarthy
1950:Turing Test – Can Machines Think?Alan Turing
1951:1st Neuron Computer, Marvin Minsky & Dean Edmonds
1956:1st AI Workshop, John McCarthy
1965:Fuzzy Sets, Lotfi Zadeh
1961:GPS, Allen Newell & Herbert Simon
1969:DENDRAL, Edward Feigenbaum, Bruce Buchanan, Joshua Lederberg
The Riseof AI
1943:Binary ANN Model,Warren McCulloch & Walter Pitts
1997:Hybrid AI Systems
1987:Fuzzy Logic-based electronics, Japan
1943 1950 1960 1970 1980 1990 2000
AIR&DActivities
1976:MYCIN, Stanford University
1982:Hopfield Net, John Hopfield
1970:ANN Learning, Bryson & Ho
1986:Back-Propagation and Start of DAI, D.E. Rumelhart & J.L. McClelland
1992:Genetic Programming, John Koza
1995:Intelligent Agent
Since 2002:ACO, PSO, AIO, DNA Computing …
DarkAge AI Winter
AI Becomesa Science
L. Wang, "From Intelligence Science to Intelligent Manufacturing," Engineering, Vol.5, No.4, pp.615-618, 2019.Reference
Earlier AI Applications
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 10
Garry Kasparov playing Deep Blue in 1997
Honda ASIMO walking downstairs in 2005
AlphaGo in 2016
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 11
AlphaGo vs. Lee Sedol
DeepMindServers in USA
During the legendary matches:Google cloud servers in the USA using 1920 CPUs, 280 GPUs and 64 search threads. Big data: 30 million moves. Reinforcement leaning, Monte Carlo search combined with deep neural network for decision making.
Self Learning of AlphaGo Zero
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 12
Human Tracking and Detection
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 13
Controller KRC
H
R
Reference Meguenani A, et al. (2016) Energy based control for safe human-robot physical interaction. Int. Symp. Exp. Robot, pp 809–818.
2020-11-05 14
Active Collision Avoidance
Reference Schmidt B, Wang L (2014) Depth camera based collision avoidance via active robot control. J Manuf Syst 33(4):711–7118.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 15
Four Safety Policies
Dynamic Planning and Dispatching
2020-11-05 16
Reference Michalos G, et al. (2014) ROBO-PARTNER: Seamless human-robot cooperation for intelligent, flexible and safe operations in the assembly factories of the future. Procedia CIRP 23:71–76.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
ML in HRC Assembly Planning
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 17
2020-11-05 18
Human action recognition
Part/tool identification
Part/tool identification: recognize what the human operator picks upIdentified part/tool: valid for a “picking up” + “installing” combination
Deep Learning of Assembly Context
Reference Wang P, Liu H, Wang L, Gao RX (2018) Deep learning-based human motion recognition for predictive context-aware human-robot collaboration. CIRP Ann Manuf Technol 67(1):17–20.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 19
PredictionBased on cognition and modeling of human behavioral pattern and preferenceHuman heterogeneity needs to be accounted for, in a probabilistic manner
ActionSafely adapt to human worker’s planned and unplanned interactions
Example: probabilistic modeling of human action and intention
,,
min,,
/ 1| ,
/ 1
|, :
| suf
i js ii j ii js ii j i
l sP sP s P
P s l s
P spositive constant
P s
Prediction
Prediction for Robotic Assistance
Reference Wang P, Liu H, Wang L, Gao RX (2018) Deep learning-based human motion recognition for predictive context-aware human-robot collaboration. CIRP Ann Manuf Technol 67(1):17–20.
Pred
ictio
n al
gori
thm
Context Aware Trajectory Planning
2020-11-05 20
Reference Maeda G, et al. (2017) Phase estimation for fast action recognition and trajectory generation in human–robot collaboration. Int J Rob Res 1–16.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Robot Assisting Human in Assembly
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 21
Typical Machine Learning Models
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 22
Machine learning models Supervised/ Unsupervised/ Semi-supervised
Discriminative/ Generative
Deep learning/ Not deep learning
K-Means Clustering Unsupervised Generative Not deep learningK-Nearest Neighbours Supervised Discriminative Not deep learningSupport Vector Machine Supervised Discriminative Not deep learningHidden Markov Model Supervised Discriminative Not deep learningRandom Forest Supervised Discriminative Not deep learningXGBoost Supervised Discriminative Not deep learningEnsemble Methods Supervised Discriminative Not deep learningConvolutional Neural Network Supervised Discriminative Deep learningRecurrent Neural Network Supervised Discriminative Deep learningLong Short-Term Memory Supervised Discriminative Deep learningNaive Bayes Supervised Generative Not deep learningGaussian Mixture Model Supervised Generative Not deep learningGenerative Adversarial Nets Semi-supervised Generative Deep learning
Multimodal H R Communication
2020-11-05 23
Reference Pavlovic VI, Sharma R, Huang TS (1997) Visual interpretation of hand gestures for human-computer interaction: A review. IEEE Trans PatternAnal Mach Intell 19(7):677–695.
Gesture basedBrainwave drivenHaptic guidanceVoice processing
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Gesture Based Robot Control
2020-11-05 24
Reference Lambrecht J, Krüger J (2012) Spatial programming for industrial robots based on gestures and augmented reality. IEEE/RSJ Int. Conf. IROS, pp 466–472.
Reference Makris S, Tsarouchi P, Surdilovic D, Krüger J (2014) Intuitive dual arm robot programming for assembly operations. CIRPAnn Manuf Technol 63(1):13–16.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
From Brain Signals to Robot Actions
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 25
EEG: electroencephalography
Types of Brainwaves
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 26
2020-11-05 27
Application server
Terminal
Brainwaves engine
EEG and gyro post-processing
Control logic
Emotionalstatebuffer
Instruction unit
Bra
in w
aves
API
Statusvisualizer
AI unit
Robot interface
Robotcontrol
Robotmonitor
Industrialrobot
controller
Emotional event query handling
Bra
inw
ave
h ead
set
Commandshandler
Rob
ot m
odul
esR
obot
com
man
ds
Rob
ot s
tatu
s
Brain waves
Robot control signals
ABB IRB 120 robot
Emotiv Epoc
InterfaceBrain Robot
Braintraining unit
Ethernet
1 42
3
Bluetooth dongle
Mentalcommands
enabler
Function Block
EEG: electroencephalography
Brainwave Driven Robot Control
Reference Mohammed A, Wang L (2018) Brainwaves driven human-robot collaborative assembly. CIRP Ann Manuf Technol 67(1):13–16.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Macro Micro Robot Control by FBs
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 28
AT
EMT
EI_INI
EI_UPD
EI_RUN
EO_INI
EO_RUNRDY
EO_ESS
FB_EXE
EMT
AT
Placing AF-FB
F2
EO_INI
EO_P1EI_P
EI_INI
ES-FB
EO_P3
ROUTE
EO_DONE
EO_P2
Screwing AF-FB
F10
Inserting AF-FB
F5
Riveting AF-FB
F11 F12
F13 F14
Press Fitting AF-FB
F4
Welding AF-FB
F6 F8
Chaining AF-FB
F7 F9
EI_ESR
MAC_ID
OPER
CC_UPD
ROUTE
Actuators, I/O’s
FB_EXE
Assembly Function
Block
Material Handling Function
Block
Manager Function
Block
R_INI
US US
AT(0)
FB_INFO
AC_UPD AC_UPD
Controller/HMI
US
ARNG_UPD ARNG
To another resource From Wise-Factory Cockpit
Controller/HMI
Material Handling
FB_INFO
AT
FB_EXE
Finished part conveyor
Robot
Fixture
P1
P2 P3
P4
EO_INI
EO_RUNRDY
EI_UPD
EI_RUN
EI_INI
Inserting AF-FB
INI_INFO Internal Algorithm ALG_INI
ALG_RUN ALG_UPDATE
ALG_MON
Internal Variables
EO_ESS
FB_INFO
AT
EI_ESR
AT
FB_EXE
AC_UPD
c
a
b
EO_RUNRDY 1
EI_RUN
EI_UPD
INI ALG_INI EO_INI
EI_INI
1
ALG_RUN
ALG_UPDATE
START
1
MON ALG_MON
1
EI_ESR
EO_ESS
RUN
UPDATE
Reference Wang L, Schmidt B, Givehchi M, Adamson G (2015) Robotic assembly planning and control with enhanced adaptability through function blocks. Int J Adv Manuf Technol 77(1–4):705–715.
FunctionBlocks
In Situ Support to Mobile Workers
2020-11-05 29
Reference Liu H, Wang L (2017) An AR-based Worker Support System for Human-Robot Collaboration. Procedia Manuf., pp 1–9.
VR/ARAVText
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
VR/AR Assisted HRC Assembly
2020-11-05 30
Reference Alexopoulos K, et al. (2013) ErgoToolkit: an ergonomic analysis tool in a virtual manufacturing environment. Int J Comput Integr Manuf 26(5):440–452.
Reference Makris S, et al. (2013) Assembly support using AR technology based on automatic sequence generation. CIRP Ann ManufTechnol 62(1):9–12.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Remote HRC: Monitoring and Assembly
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 31
Joints 2 5Joints 2 5Joints 2 5
Virtual Universe
BackgroundLights
T
BG Viewpoint
Control
BasePlatform
A A
G A
Joint-1Behaviour
Control
T T T
T
T
Joint 1
T
Joints 2 5
GripperBehaviour
Control
Gripper A Appearance
G Geometry
User defined codes
TransformGroup nodeT
Behaviour nodeB
BG BranchGroup node
B
B
B
B Amini robotic assembly cell
Reference L. Wang, M. Givehchi, G. Adamson and M. Holm, "A Sensor-Driven 3D Model-Based Approach to Remote Real-Time Monitoring," CIRP Annals – Manufacturing Technology, Vol.60, No.1, pp.493-496, 2011.
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 32
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 33
Future Research Directions
2020-11-05 34
Current standards do not yet address many key issues of mixed teams of humans and robots. More realisticstandards are needed.
Digital twin shall combine and align all aspects of modelling the function, structure and behaviour of the robotic cell including the worker, representing the multimodal and bidirectional channels of communication and control when effectively tuned to the real environment.
Emergency cases require fast and guaranteed mitigation to bring the situation back to normal without interruption if the situation does not endanger the integrity or safety of human.
AR-based support to workers in dynamic HRC assembly deserves more attention to be both intuitive and mental stress free. Work instructions need to be adaptive to not only the changing competence level of individual workers but also the declining focus and concentration during the day or within the week.
When adding humans to shared robotic environments, the trust of the humans on the robotic environments is unavoidably important and deserves attention.
In HRC assembly, mental stress and psychological discomfort leading to any potential accident can be monitored and diagnosed via the brainwaves of workers, collected by sensors embedded in a safety helmet.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Statistics of Cited References
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 35
Near Industrial Applications
Symbiotic HRC Assembly in Action
2020-11-05 36
Acknowledgem
entSYM
BIO-TIC project has received funding from the European Union’s H
orizon 2020 research and innovation program
me under grant agreem
ent No. 637107.
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
2020-11-05 KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang 37
Summary
Adaptive Robot Control
Dynam
ic Task
Planning
Mobile W
orker A
ssistance
Active Collision Avoidance
Smart Sensor Network
Location-aware Task-specific Multimodal In-situ assistance Decision support
Intuitive support Portal for end users Dynamic task planning/re-planning
Supervisory control
IEC-61499 function blocks compliant Smart algorithms encapsulation Adaptation via event-driven behaviours Zero robot programming to end users Multimodal programming in real-time
3D models of structured environment Point-cloud images of human workers
SSaaffee HHRRCC EEnnvviirroonnmmeenntt
HHuummaann--RRoobboottCCoollllaabboorraattiioonn
R1
R2
R3
R4
R5
Symbiotic communication: Auditory Visual Haptic
Background and definition of human-robot collaboration (HRC)
Active collision avoidance in HRC assembly
Dynamic assembly planning and on-demand job dispatching
Adaptive and programming free robot control
In situ operator support in changing HRC assembly environment
Challenges and future researchdirections
2020-11-05 38
Thank You for Listening
KTH Sustainable Manufacturing • www.kth.se • © Lihui Wang
Professor of Sustainable ManufacturingKTH Royal Institute of Technology • Stockholm • Sweden
L. Wang, R.X. Gao, J. Váncza, J. Krüger, X.V. Wang, S. Makris and G. Chryssolouris, "Symbiotic Human-Robot Collaborative Assembly," CIRP Annals – Manufacturing Technology, Vol.68, No.2, pp.701-726, 2019.
Reference
Lihui Wang