Robotics, Agents, and e-Work:
The Emerging Future of Production
Shimon. Y. Nof
PRISM Center
Production, Robotics, and Integration Software for Mfg. & Management
Purdue University, W. Lafayette, IN, USA
INCOM’06, St. Etienne, France
May 17-19, 2006
1
PRISM Lab
/Purdue
Outline
1. The influence of e-Work on enterprises and on
society
2. Six key design principles and collaborative
control theory
3. Network models and their emerging redesign
counterparts
4. IMPACT: What does it mean? What do we
really expect out of production?
2
PRISM Lab
/Purdue
e-Work Definitions
•
•
[PRISM Center, 1999]
Collaborative , computer -supported , and
communication - enabled work operations in highly
distributed organizations of humans / robots /
autonomous systems
Our goal: Augment human abilities at work, and organizations’ abilities
to accomplish missions
v - Design
e - Business
e-Commerce
v - Factory
i-Robotics
e-Mfg.
e- Logistics
i -Transp.
v-Enterprise
...
e-W o r k
• Challenges:
•
•
Complexity • Scalability • Dependence • Integrity
• Communication • Coordination • Noise • Mismatch …
•
•
•
Problem: Potential promise of emerging e-work cannot materialized
/Purdue
without
collaboration
support
PRISM Lab
3
The transformative influence of e-Work
‹
We know: power fields, e.g., magnetic fields and gravitation
ƒ
‹
Influence bodies to organize and stabilize
So do cyber fields, e.g., IT and communication
ƒ Envelop us and influence us to organize our work
systems in a different way
ƒ Purposefully, stabilize work to effectively produce
desired outcomes
4
PRISM Lab
/Purdue
Survey of e-collaborative production - sample
(Jeong, 2005)
Multi-agent system architecture
for machine tool automation
Fault Diagnosis of CNC-machines
(Marzi & Timpe, TU-Berlin)
Virtual Labs (H. Erbe et al.)
5
collaborating robots:
Master-slave model
“ROBOTLINK simultaneous
motion” www.fanuc.co.jp
PRISM Lab
/Purdue
e-Collaborative design, engineering and manufacturing
Distributed actuation & measurement
Petin, Iung, Morel; France
Collaborative CAD / CAE / CIM
Y.S. Wong; Singapore
6
PRISM Lab
/Purdue
End-to-end processes: Collaborative, inter-operable, harmonized
Supply network harmonization
e-Collaborative e-Manufacturing
L. Korba, “e-Manufacturing”
Canada
http://www. Apriso.com, FlexNet™
7
PRISM Lab
/Purdue
The 15 e-dimensions of collaborative e-Work
8
ee
uu
rdrd
uu
/P/P
ab
ab
MMLL
ISIS
RR
PP
Scope of e-Production challenges and solutions
‰ e-X ≠X. e-Work and e-Production /
e-Business are not the same as work,
production, and business.
‰ In general, e-Work enables
e- Production / e-Business, and the
latter require e-Work.
‰ Design objectives and benefits must
respond to increasing needs of
sustainability and dependability
with a growing world population.
‰ Emerging enablers: Networks of
demand & supply; smart teams,
workflow interactions, decentralized
decisions and automation, and
collaborative control and
management.
9
CHALLENGES
SOLUTION
TRENDS
[IFAC-CC5 Milestone, 2005]
PRISM Lab
/Purdue
e-Work enables e-Production
e-Work
e-Work
FUNCTIONS
TOOLS
¾ e-PRODUCTION
¾ e-Logistics
•
e-Operations
• Human-Computer Interaction
• Human-Robot Interaction
• Integration
• Collaboration
• Coordination
• Networking
10
•
•
•
•
•
•
•
¾ e- Design
Agents
Protocols
¾ e-Factory
Enables
Workflow
¾ ERP
Middleware
Parallelism
¾ e-Manufacturing
¾ CRM
Requires
¾ SCM
Teamwork
¾ Robotics
Groupware
¾ Nano-systems
PRISM Lab
/Purdue
Emerging principles of collaborative control theory
• Collaborative control theory has been developed to support
effective design of e-collaborative systems
• Network models applied to analyze/ optimize design
• A network model is defined as N = {Vk, Eij} where V is a
vertex, or node; E is an edge, or channel.
• Network flow can be uni- or bi-directional; Nodes and/or
edges (channels) can be active
N = {Vk , Eij }
Flow through edges
11
Activation of edges and/or vertices
PRISM Lab
/Purdue
(1) The Principle of Collaboration Requirement Planning, CRP
[Rajan and Nof, 1996]
•Effective e-collaboration requires advanced planning and
on-going re-planning
• CRP- I, plan “who does what, how, and when”
• CRP-II, during execution, revise plan in real time, adapting to temporal
and spatial changes and constraints
N = {Vk , Eij }
Eij
Eij
Vk
R + N 0 → N1 ; Static
12
N = {Vk , Eij }
Vk
R (t ) + N 0 → N 2 (t ); Dynamic
Network models for CRP on initial network N0:
Static requirements R; Dynamic R (t).
PRISM Lab
/Purdue
(2) The Principle of Collaborative e-Work Parallelism
[Ceroni and Nof, 1999]
• Optimally exploit the fact that work in software work-spaces
and human work-spaces can and must be allowed to
advance in parallel
• To be effective, e-Work systems cannot be constrained by
•
13
linear
(sequential) precedence of tasks – delegate!
“KISS”: Keep It Simple, System!
Tasks(t ) + N0 →{Tasks(t ), na (t )} + {Tasks(t ), nb (t )}
N 0 = {na (t ), nb (t )}
PRISM Lab
/Purdue
PIEM (centralized optimization algorithms) and DPIEM (optimization with
distributed protocols) for planning the communication and coordination tradeoffs in e-Work of design, mfg., logistics with parallelism
Start
F
1A
0.800
Summary: Local and Integrated Scenarios [Ceroni, 2000]
1B
1.000
No. of Sub-
J
Φ
Π
Τ
tasks
Local
Scenario
(A+B)
1.984
6
1.4239
0.5607
52
Enterprise
A
1.390
8
1.0350
0.3558
16
Enterprise
B
0.593
8
0.3889
0.2049
36
Integrated
Scenario
1.807
1
1.1666
0.6404
26
F
2A
0.600
2B
0.800
3A
1.000
J
4A
0.700
F
5A
0.400
3B
0.800
J
F
6A
0.700
4B
0.900
J
Optimize the DOP, Degree of Parallelism
End
14
PRISM Lab
/Purdue
(3) The Principle of Conflict Resolution in Collaborative e-Work
[Huang and Nof, 1999]
• Minimize the cost of resolving conflicts among collaborating
e-Workers by automated EWSS (e-Work support systems)
• Beyond reducing information and task overloads, e-Work
must be
designed to automatically prevent and overcome as many errors and
conflicts as required to be effective
{eij (t )} + {cij (t )}
Conflict & Error Detection Agents (CEDA) and Protocols
(CEDP) are assigned to Network N0(t)
15
PRISM Lab
/Purdue
Critical Cost of Error Recovery / Conflict Resolution
(C-Y Huang, 2001)
1000
900
s=0.1
800
s=0.2
Cost
700
s=0.3
600
s=0.4
500
s=0.5
400
s=0.6
s=0.7
300
s=0.8
200
s=0.9
100
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Number of Iterations
Increases
exponentially when
human
communications
and operations are
applied (assuming
q=0.2)
100
90
s=0.1
80
s=0.2
Cost
70
s=0.3
60
s=0.4
50
s=0.5
40
s=0.6
s=0.7
30
s=0.8
20
Reaches an
upper bound
when IT is
Applied
(assuming q=0.0)
s=0.9
10
0
1
2
3
4
5
6
7
8
9
10
Number of Iterations
16
11
12
13
14
15
q = % of human involvement
S = rate of conflicts
PRISM Lab
/Purdue
Solutions for e-diagnostics, e-recovery, e-resolution
Help Request with
Advisory Action
Sensor
Knowledgebase
Computer-Supported Conflict
Resolution: FDL’s Field-of-View
function [Lara, 2001]
Recovery
Strategy
Assistance
Recovery
Order
Controller
Sensor
Information
Sensing
R
o
b
o
t
Operations and
Recovery Actions
Production Process
NEFUSER: Neuro-Fuzzy Error Recovery
with human-robot recovery interactions
[Avila, 2002]
17
PRISM Lab
/Purdue
(4) The Principle of Collaborative Fault-Tolerance
[Jeong and Nof, 2005]
• Fault-tolerant collaboration can yield better results from a
team of weak agents, relative to a single optimized and
even flawless agent
•
Ancient synergy lesson: “A collaborative team is better than the sum of
the individuals”; this principle: How to achieve this advantage by smart
automation, even with some faulty agents/channels
N (t1 )
Eij
Vk
Eij
Fault (t1 + ∆t )
Vk
N (t1 ) = {Vk , Eij };1 − 2 − 5 − 6 − 9
N (t2 ) = {Vk , Eij }
Original flow 1-2-5-6-9 adapts to link or node failures Æ backup flow 1-47-8-9 or 1-2-3-6-9, depending on energy and response-time constraints
18
PRISM Lab
/Purdue
Collaborative fault-tolerance TAP design
50
45
40
35
Y(m)
30
25
20
15
10
5
0
Principles 1-4 at work:
Alternative MEMS and nano sensor
arrays / networks optimized along an
artery for measurement and control
19
0
5
10
15
20
25
30
35
40
45
50
X(m)
6 Faulty sensor routed
communication by a timebased control [Jeong, 2006]
PRISM Lab
/Purdue
(5) The Join/Leave/Remain (JLR) Principle in Collaborative
Organizations
[Chituc and Nof, 2005]
• Individual
organizations: Decide repeatedly when and why to
JLR a given CNO based on measured total participation
gains and costs, 3-D {Agility; Payoff; Cost}
• For a CNO: Same (including increased coordination) relative
to each member organization
(a)
Eij
Vk
Vn +1 (t )
N = {Vk , Eij }
(b)
N1
Eij
Eij
Eij
Vk
N2
N3
Vk
Vk
N 3 ≠ N1 + N 2
The JLR model: (a) Two nodes join a network. (b) Larger and small
/Purdue
organizations
unite,
transform
the
original
network.
PRISM Lab
20
(6) The Principle of Emergent Lines of Collaboration
and Command, LOCC
[Velasquez and Nof, 2006]
• Design evolutionary mechanisms of interaction and
organizational learning for better ad-hoc decisions, effective
improvisation, on-the-spot contact creation, and best
matching protocols to pair decision makers and executors
•
•
Critical under all hazards emergency situations
Ex. real-time KBS for monitoring and diagnosing dynamic processes
N (t1 )
Eij
Vk
Eij
Disturb(t1 + ∆t )
N (t1 ) = {Vk , Eij };1 − 2 − 5 − 6 − 9
Vk
N (t2 ) = {Vk , Eij };1 − 2 − 5 − 8 − 9
LOCC model: Line 1-2-5-6-9 evolves after disturbance (emergency)
to 1-2-5-8-9
21
PRISM Lab
/Purdue
TIE, Teamwork Integration Evaluators: Parallel/distributed
interactions and engagements
New measures needed: XXXability – e.g., viability, interoperability, detectability, scalability, dependability, reachability, integration-ability, recover/Purdue
ability,
learn-ability,
resolvability,
integrity,
trust-ability,
…..
PRISM Lab
22
Challenge:
Bio-inspired collaborative node and channel behaviors
(b)
F2 (t )
(a)
F1 (t )
F (t1 )
F1 (t ) = F2 (t )
F2 (t )
F1 (t )
(e)
F (t2 )
(f)
(d)
F1 (t ) > F2 (t )
(c)
Physical, electrical, and biochemical dynamic changes
in neural/brain activities
enable survival,
collaboration, interactions
(h)
N1 = M a
(g)
N2 = M v
R1* ( Route)
R2*− (adaptive)
Bio-inspired network models: (g)
Survival maps; (h) propagation; (i)
same, both nodes & channel variations
23
(i)
PRISM Lab
/Purdue
Summary 1/2
Collaborative e-Work and e-Production – Our Future
Emerging: Smart parallel teams will be able to interact better;
Collaborative control EWSS will enable production
effectiveness, harmony
Trend 1.
Collaborative Coordination Control Theory
Why? Optimized coordination of e-Work interactions is critical
Smart protocols will prevent and manage errors,
conflicts, reorganization, and interactions
Trend 2.
Why? Complexity, dependency among distributed teams increases
Trend 3. Smart protocols for fault-tolerant collaboration
Why? Future e-production will depend on cheaper, redundant,
disposable arrays/networks/teams
24
PRISM Lab
/Purdue
Summary 2/2
IMPACT: What does it mean? What do we really expect out of
production?
•
Over half the world population does not yet enjoy the
benefits of production – important changes are emerging
•
Food, transportation, security, medical & healthcare
services, education, civilization -- all depend on production
•
Emerging production analogous to dramatic productivity
transformation in agriculture
•
The emerging future of production enterprises depends on our
understanding of how effective e-Work can be designed
•
Interesting questions: dynamic reorganization? optimal
clustering? profitability vs. sustainability? Topology-based
performance? Bio-inspired network behavior?
•
Most important: Let’s collaborate wisely!
25
PRISM Lab
/Purdue
Acknowledgement
‹ Research
reported in this article has been
developed at the PRISM Center with NSF, Indiana
21st Century fund for Science and Technology,
and industry support.
‹ Special thanks to my colleagues, Visiting Scholars
and students at the PRISM Lab and the PRISM
Global Research Network, and in IFAC Committee
CC5 for Manufacturing and Logistics Systems, who
have collaborated with me to develop the e-Work
and collaborative control knowledge.
26
PRISM Lab
/Purdue