School of Computing

1
School of Computing Mathematics Research
Conference 2004

• Environmental Scanning And Its Role In Adaptive
  Software Systems
• Henry Forsyth
• Supervisory Team
• Prof. Taleb Bendiab
• Mr. Andy Laws
• Dr. Carl Bamford

2
Todays Presentation

• Introduction to Adaptive Software
• Basic Steps In Adaptation
• The Role Of The Environment In Adaptive Software
• Why Are Environments So Problematic?
• Some Issues For Consideration
• Problems Encountered So Far
• Environment Scanning An Organisational Approach
• Characteristics For Successful Environmental
  Scanning
• Closed Loop Adaptive Test System
• Next Research Steps
• Conclusions

3
Introduction To Adaptive Software

• Self-adaptive software evaluates its own
  behaviour and changes behaviour when the
  evaluation indicates that it is not accomplishing
  what the software is intended to do, or when
  better functionality or performance is possible.
  DARPA 1997
• Self-adaptive software modifies its own
  behaviour in response to changes in its operating
  environment. Oreizy 1999
• Therefore the benefits of adaptive software
  should be-
• Increased robustness and reliability
• Less requirement for human maintenance
• Increasingly agile software systems

4
The Role Of The Environment In Adaptive Systems

• The environment is likely to play an
  increasingly important role in software
  development in the future
• Ubiquitous Devices will increasingly utilise
  sensors which will provide a view into the
  physical world around us
• The development of GRID computing
• Sensor Based Human Machine Interaction
  (J.Allanson)
• Environments are becoming increasingly complex,
  populated and unknown
• If we could define / model our environment
  completely there would be no need for adaptive
  software
• Without effective environmental awareness we will
  find it difficult to develop robust reliable
  software systems

5
Basic Steps In Adaptation and Why Are
Environments So Problematic ?

• Adaptation basically involves three tasks
• 1) An ability to recognise a change in the
  environment
• 2) An ability to determine the changes to be made
  to the system in response to environmental
  changes
• 3) An ability to effect the change in order to
  generate the new system
• From an adaptive software viewpoint many research
  groups are concentrating on other elements which
  also need solving-
• Learning
• Evaluating
• How can we successfully model an environment if
  it
• is so complex and subject to dynamic change?

6
Some Issues For Consideration

• What kind of information needs to be collected
  from the environment to make effective adaptation
  decisions?
• How Accurate ? (Precise, Approximate, Sampled
  observations)
• How current ?
• The most successful animals are those that can
  adapt quickly to their environment as opposed to
  always selecting the optimum solution
• How do we represent the raw data so that the
  system can perform reasoning on it
• This will require modelling our environment in
  order to be able to
• Diagnose program failures and performance
  problems
• Provide a basis for choosing new strategies
• Can we develop filtering techniques to reduce
  background noise from the environment which
  will improve
• the efficiency of our adaptive systems
• This is an overhead associated with monitoring
  any
• environment and therefore is it important to
  achieve the
• appropriate monitoring balance

7
Problems Encountered So Far

• Difficulty in obtaining self-adaptive research
  material
• The self-adaptive material available whilst
  mentioning the importance of the environment was
  tending to concentrate on other areas
• Too slow in finding an application in which to
  ground the research
• Instead of looking for killer applications I
  should have been looking for something more
  limited in scope but which I could get up and
  going much sooner
• Getting easily discouraged
• Not looking at other domains which could possible
  provide guidance in successfully developing
  Environmental
• Scanning policies and techniques which could be
• applied to the field of adaptive software

8
Environment Scanning A Strategic Management
Approach

• Environmental Scanning defined as A kind of
  radar to scan the world systematically and signal
  the new, the unexpected, the major and the minor
  (Brown 1985)
• Literature search extended into business
  organisation in which research concluded that
  companies with more effective scanning and
  assessment measures are more successful (Newgren
  1984, Dollinger 1984)
• However guidelines on how to scan are few
• The scanning process should be structured to
  minimise the chances of being blind-sided by an
  environmental change you should have seen coming

9
Characteristics For Successful Environmental
Scanning

• An ability to measure environmental uncertainty
  and devise scanning measures to compensate
• Therefore if we perceive the environment to be
  uncertain / hostile we should increase scanning
  resources
• Environmental complexity could possibly be
  calculated by
• The number of environmental factors considered in
  decision making
• The degree to which these factors change over
  time
• An ability to alter scanning techniques depending
  on the environment
• This might mean scanning changing from
• Information Searching (Actively search the
  environment)
• Passive Information Monitoring
• An ability to perform sufficiently comprehensive
• scanning
• Data is often inherently scattered, vague, and
  imprecise and
• come from a host of sources (Morrison 1992)

10
More Characteristics For Successful Environmental
Scanning

• An ability to adjust scanning frequency to reduce
  overhead
• Irregular Scanning (ad-hoc and tends to be crisis
  initiated)
• Periodic Scanning (an area which to have a low
  environmental uncertainty)
• Continuous Scanning (Key indicator of system
  performance)
• An ability to make a risk assessment in terms of
  severity and probability of its occurrence
• This will help to focus scanning using risk
  analysis techniques
• The ability to recognise time critical
  environmental occurrences

11
Next Research Steps (12 Months)

• Continue to widen the research search to other
  areas which may be applicable to environmental
  scanning and modelling
• Build a closed adaptive test system
• Simple Application is all that is required at
  this stage
• Develop a environmental scanning framework which
  may well itself include adaptive behaviour
• Ability to alter scanning in relation to
  environmental measures discussed earlier
• Test the framework / systems with increasingly
  complex / turbulent environments to measure
  performance and further improve functionality

12
Closed Adaptive Test System

• The difference between a open / closed adaptive
  test system would be that
• A system is open-adaptive if new application
  behaviours and adaptation plans can be introduced
  during run-time
• A system is closed-adaptive if it is
  self-contained (limited number of adaptive
  behaviours) and is not able to support the
  addition of new behaviours
• Andy Laws has developed an early closed adaptive
  test system which involves sort algorithms
• His work is initially dealing with a well behaved
  environment
• As open adaptive systems are considerably more
  difficult to build (how do we trust new
  behaviours, learning, evaluation etc) I am going
  to build a closed adaptive test system but with
  an increasingly turbulent environment
• This will allow me to develop environmental
• scanning strategies without getting dragged into
  
• wider adaptive issues

13
Conclusions

• The environment will play an increasingly
  important element in all software development
  into the future
• Its certainly not easy !!
• However there are lessons to be learned from
  other disciplines which may be applicable in
  helping us understand the factors behind
  successful environmental scanning
• Starting to ground some of my finding in terms of
  developing some testing software will help
  greatly in allowing progress to be made
• The next 12 months is likely to be critical in
  determining the success of this research

14

• Any Questions ?