Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation What we did at work today (Rawwrrrr!) by Amit Gupta on Flickr (CC BY-NC) Complexity Controlling chaos using cyberne5cs and good design P.O. Arnäs, PhD [email protected]@Dr_PO
75
Embed
Complexity - Controlling chaos using cybernetics and good design
What is a complex system? How can it be controlled? What is cybernetics and why is it so cool?
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and TransportationWhat we did at work today (Rawwrrrr!) by Amit Gupta on Flickr (CC BY-NC)
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
What is a system?
�2
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A set consists of more than one elementary part
�3
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A system differs from a set when it displays emergent proper,es
A set consists of more than one elementary part
�4
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
The whole shows proper5es that are not
found in its parts
A system differs from a set when it displays emergent proper,es
A set consists of more than one elementary part
�5
Reduc5onism is not applicable
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A set consists of more than one elementary part
A system differs from a set when it displays emergent proper,es
A system cannot be fully understood by studying its parts
separately
The whole shows proper5es that are not
found in its parts
Reduc5onism is not applicable
�6
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A set consists of more than one elementary part
A system differs from a set when it displays emergent proper,es
A system cannot be fully understood by studying its parts
separately
The whole shows proper5es that are not
found in its parts
�7
Reduc5onism is not applicable
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A system cannot be fully defined from within
Gödel´s incompleteness theorem
�8
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
A system cannot be fully defined from whithin
Gödel´s incompleteness theorem
I don´t understand
I don´t understand
Neither do I
�9
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation2009 -‐ October -‐ NodeXL Facebook Network Marc Smith FR Layout by Marc_Smith on flickr.com
I don´t understand
I don´t understand
Neither do I
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Si = (xi, yi, zi, vi)
”Trajectory” – a succession of states
”State” – The current value of the a6ributes
”A6ributes” – a collec9on of the system’s variables
An interface is a representation of a system displaying its visible
state and its visible inputs.
�37
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Image: ”MacBook Air by Jony Ive” BY marcopako on flickr
The System
Visibleattributes
Sta
te vec
torVisible
inputs
Inbound interface Outbound interface
Dimension 1: Interface direcWon
An interface can be either inbound or outbound
Ope
ration
s vect
or
�38
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Image: ”MacBook Air by Jony Ive” BY marcopako on flickr
Dimension 2: Interface width
The width of the interface consists of the number of
attributes/operations
The SystemWidthWidth
More attributes (degrees of freeedom) – wider interface
More operations – wider interface
The interface width is reduced by
encapsulation
�39
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Image: ”MacBook Air by Jony Ive” BY marcopako on flickr
EncapsulaWon
The System
The System
Encaps
ulatio
n
of co
ntent
Encapsu
lation
of functi
on
Hidden attributes
Hidden inputs
�40
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Image: ”MacBook Air by Jony Ive” BY marcopako on flickr
Dimension 3: Interface depthThe variety of each attribute and operation together constitute the
interface depth
The System
Depth DepthThe interface depth
is reduced by applying constraints
The depth can range from binary to continuous
�41
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Image: ”MacBook Air by Jony Ive” BY marcopako on flickr
Constraints
The System
The System
Stati
onary
conts
traint
s
Transitor
y
constrain
ts
Limiting states
Limiting operations
�42
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Interface WIDTH
Int
erfa
ce
DEP
TH
NARROW WIDE
SHALLOW
DEEP
Many parametersFew parameters
Binary parameters
Discreet parameters
Continous Parameters
with constraints
Continous Parameters
Constraints make interfaces more shallow
Encapsulation limits interface widthDegenerated
encapsulation
No encapsulation – open system
�43
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Reducing complexity – some ground rules
Exclude variables Diminish state space
Par99on states Group states into larger
par99ons
Break down into subsystems Create internal interfaces
Organise subsystems hierarchically Create mul9ple levels of abstrac9on
Image: ”Jam at the floaWng market” BY Stuck in Customs on flickrImage: ”Playing chess” BY Jeffrey Barke on flickrImage: ”Jan 2005 Map of the Internet” BY ma_hewje_hall on flickr
�44
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Key tools in cyberne9cs
In In In In In
Feedback loop
Syst
em s
tate,
output
Input
Disturbance
Regulated system !(transportation
system)
Regulator !(TCS)
Black boxes Encapsulate func9on Encapsulate content
Hierarchic system models
Control Theory Regulator Trajectory which needs to be controlled Finite state space
�45
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
�46
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
By construc9ng the system from a number of black boxes, complexity can be reduced
�47
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
In
By construc9ng the system from a number of classes, complexity can be reduced
�48
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
How would a programmer
design a complex system?
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation Image: ”Wheldrake sundial” BY Darwin70 on flickr
A parallel historyIn 1967, the programming
language Simula 67 was launched !
The first object-‐oriented language !
Used to build discrete event simula9on models
“…the opera9on of a system is represented as a chronological sequence of events. Each event occurs at an instant in 9me and marks a change of state in the system.” (Wikipedia)
Discrete event simula9on
Si = (xi, yi, zi, vi)
Remember
TRAJECTORY
�50
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
The canonical form of a complex system (redrawn from Booch, 1991).
�51
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Class diagrams
�52
Consignment
11..4
Waybill
Waybill item
1*
Physical
-UN number-Substance information-Handling instructions-Emergency procedures-Contact information
DG fact sheet-Statement-Name-Signature
Sender certificate Dangerous Goods Declaration
-Product name-UN number-Quantity-Packing
DGD item
1
*
Consignment documentation
Electronic
-Consignor-Consignee-Customer number
Order
-Type of goods-UN number-DG class-Substance number-Substance Name
Order Item
Weight VolumeLength Number of pallets
Size
1
*
1
*
1 *
1
*
Product
1
*
+Overpack()+Open()+Move()+Fill()+Empty()
-Weight-Size-Position
Pallet
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
+MoveTo(in NewX, in NewY, in NewZ)+LiftUp(in Distance)+PutDown()
This is the class Pallet that represents all objects that share the same data structure.
The topmost cell contains the name
The middle cell contains the attributes
The bottom cell contains the operations
�53
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
The class Box.
�54
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
An object of the class Pallet can contain several objects of the class Box. The denotations 1 and * means that one pallet (1) may contain many (*) boxes.
�55
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation Image: Connected. 362/365 by AndYaDontStop on Flickr.com
Object-‐orienta9on is applied cyberne9cs!
When looking at the core concepts of object-‐orienta9on there is a clear analogy with cyberne9cs
Very few people have made that connec9on during the last 40 years!
Why?
Different disciplines (Systems science/Mathema9cs vs Computer science) Some advances have been made towards a combina9on, but these are few
�56
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation Image: Connected. 362/365 by AndYaDontStop on Flickr.com
Systems theory Object-‐orienta9onSystem model ↔ Object model
State ↔ State
Outbound (display) interface ↔ A6ributes
Inbound (control) interface ↔ Opera9ons
Black Box ↔ Object
Trajectory ↔ Path in Statechart
Encapsula9on ↔ Encapsula9on
Abstrac9on ↔ Abstrac9on
Hierarchic architecture ↔ Hierarchic architecture
Transforma9on ↔ Object behaviour
�57
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Three approaches
123�59
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Long-‐term control scope 1Time-‐span: Years to months
Important tasks: Define the data-‐structure of the top-‐level classes of the system, i.e. the interface widths. Define acceptable data ranges for these classes, i.e. the interface depths. Define acceptable use cases. Descrip9ve complexity is reduced
by robust design
Driving ques9ons: What states should the
system be able to assume? What component types are required for the system to
assume these states?
�60
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Medium-‐term control scope 2Time-‐span: Months to weeks
Driving ques9ons: What components are needed in the system? How are the various interfaces designed?
Important tasks: Define actual use cases. Define interface width of all classes. Apply constraints to reduce interface depth.
Computa9onal complexity is reduced by good planning
�61
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Short-‐term (real9me) control scope3Time-‐span: Weeks to minutes
Driving ques9on: What state changes should
be performed and how?
Important tasks: Control the actual trajectory as it progresses through the state space.
Uncertainty-‐based complexity is reduced by crea9ng order
�62
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Examples
Focusing on interfaces
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
HETEROGENEOUS GOODS
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Arnäs, Woxenius – Approach for handling heterogeneous goods in intermodal freight networks – revisited, WCTR 2013
Heterogeneous goods leads to increased complexity
Image: http://sobar.soso.com/t/74147926?fl=29
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Heterogeneous goods leads to increased complexity
Image: http://sobar.soso.com/t/74147926?fl=29
Density Handling Stowability Liability
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
What makes some goods heterogeneous? Four dimensions:
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
Short reading list
• Klir, G. J. (1991) Facets of systems science, Plenum Press, New York.
• Booch, G. (1991) Object oriented design with applica9ons, Benjamin/Cummings Pub. Co., Redwood City, Calif.
• Ashby, W. R. (1956) An introduc9on to Cyberne9cs, Chapman & Hall Ltd, London.
• Beer, S. (1959) Cyberne9cs and management, English UniversiWes Press ltd, London.
• CasW, J. L. (1989) Alternate reali9es : mathema9cal models of nature and man, Wiley, New York ; Chichester.
�70
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
PhD-‐theses
• Hultén, L. A. R. (1997) Container logisWcs and its management, Department of Transporta9on and Logis9cs, Chalmers University of Technology, Göteborg
• Franzén, S. E. R. (1999) Public transporta9on in a systems perspec9ve : a conceptual model and an analy9cal framework for design and evalua9on, Chalmers tekniska högsk., Göteborg.
• Waidringer, J. (2001) Complexity in transporta9on and logis9cs systems : an integrated approach to modelling and analysis, Chalmers tekniska högsk., Göteborg.
• Nilsson, F. (2005) AdapWve LogisWcs -‐ using complexity theory to facilitate increased effecWveness in logisWcs, Department of Design Sciences, Lund University, Lund, Sweden
• Arnäs, P. O. (2007) Heterogeneous Goods in TransportaWon Systems -‐ A study on the uses of an object-‐oriented approach, Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie, 2625, Chalmers University of Technology, Göteborg
�71
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
+ =
The Internet first became available for Swedish consumers around 1993
A (bad and expensive) mix between Teletext and the Yellow Pages
Not very pretty
Few people with computers
And finally…
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
15 years later, in 2008, Google Flu Trends was launched
Based on what people google and where their computer is located
Ten days ahead of the official flu
tracker
www.google.org
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
How will we use the technology in 10 years?
We have no idea.
!
...and neither does she, but she will be dissatisfied with stuff that we think are
pure science fiction and almost magic.
almost
The girl and the iPad by Niclas Lindh on Flickr (CC-BY)
Per Olof Arnäs, Technology Management and Economics, Division of Logistics and Transportation
How will we use the technology in 10 years?
The girl and the iPad by Niclas Lindh on Flickr (CC-BY)