Goaldriven adaptation semantics

1
Goal-driven adaptation semantics
Gregory Brown
Advisor Dr. Betty H.C. Cheng
Problem Description Adaptive systems can modify
their behavior through dynamic adaptation at
runtime this adds inherent complexity Complexity
can lead to requirements and design
errors errors can lead to errant
behavior Formally describe the adaptation
semantics3 this can be used to verify
correctness Goal-oriented specification can aid
in correct design1 provides a graphical
representation of adaptation semantics and
graphical wrapper to formal definitions KAOS
goal-oriented modeling is easy to use has good
tool support1,2

• Common Adaptation Semantics3
• One-point adaptation
• Guided adaptation
• Overlap adaptation

Guided Adaptation KAOS Model The overall
goal, 1, is the same as in one-point semantics
but is refined to the goal to convert from the
restricted source program to the target program.
This is further refined into a goal to change
from the source program to a restricted source
program. Initially, the Change Agent, 7, monitors
for the request to change, AReq, 8, that triggers
the Change Agent performing the adaptation. When
the request is seen, the Change Agent changes the
system from the source program to the restricted.
When the safe state on the source program is
reached, the change agent then adapts to the
target program. The overlap adaptation semantics
are similar. Instead of there being a restricted
source program stage, the target program is
started immediately and both source and target
are restricted. When the source safe state is
reached, the source is simply shut off.
One-point Adaptation At a point, there is a
request to adapt, AReq At some point after that,
it executes the adaptation Adaptation occurs when
the source, SSpec, reaches a safe state, can be
turned off and target, TSpec, can be turned
on
KAOS Objects Goals represent system
goals Requirements a goal under the
responsibility of an automated Agent AND
Refinement all subgoalsmust be fulfilled for
parent goal to be OR Refinement at least one
subgoalmust be fulfilled for parent goal to be A
used to show temporal sequencefrom left to
right
One-point Adaptation KAOS Model The overall
goal, 1, is to convert from the source program,
or state, to the target program, or
state. Initially, the Change Agent, 4, monitors
for the request to change, AReq, 5, that triggers
the Change Agent performing the adaptation. This
goal is fulfilled by the sequence of the system
being in the source state, 2, and then in the
target state, 3, after adaptation.

• Assumptions for adaptation semantics
• Adaptive systems act like non-adaptive systems
  for periods of time
• Process of adaptation
• originally in source configuration
• identifies need to adapt
• undergoes adaptation
• in target configuration
• Assume the adaptive system has moderate
  computational reflection to observe internal
  behavior and state
• The source and target programs are expected to be
  formally defined with LTL or A-LTL3
• The adapt operator extension

Guided Adaptation At a point, there is a request
to adapt, AReq The system then enters a
restricted source state, with restriction
condition, RCOND Eventually, a safe state for
adaptation is reached Adaptation occurs when the
restricted source reaches a safe state and can be
turned off The target is then turned on
References 1 A. Dardenne, A. van Lamsweerde, and
S. Fickas. Goal-directed requirements
acquisition. Science of computer Programming,
20350, 1993. 2 M. Feather, S. Fickas, A. van
Lamsweerde, and C. Ponsard. Reconciling system
requirements and runtime behavior. In IWSSD
Proceedings of the 9th international workshop on
Software specification and design, page 50,
1998. 3 J. Zhang and B. H. C. Cheng. Using
temporal logic to specify adaptive program
semantics. Journal of Systems and Software
Special Issue on Architecting Dependable
Systems, 2006. 4 Greg Brown, Betty H.C. Cheng,
Heather Goldsby, and Ji Zhang, Goal-oriented
Specification of Adaptation Semantics in
Adaptive Systems, to appear in ICSE 2006
Workshop on Software Engineering for Adaptive and
Self-Managing Systems (SEAMS), Shanghai, China,
May 2006
Acknowledgements This work has been supported in
part by NSF grants EIA-0000433, EIA-0130724,
CCR-9901017, CNS-0551622, CCF-541131,
CCF-0541131. Department of the Navy, Office of
Naval Research under Grant No. N00014-01-1-0744,
Eaton Corporation, Siemens Corporate Research,
and a grant from Michigan State Universitys
Quality Fund for High Assurance Systems.
April 20, 2006