Problem Frames: part 2

1
Problem Frames part 2

• Frame Concerns
• Model Domains
• Variant Domains
• Particular Concerns
• Frame Flavors

2
Concern

• If I have this kind of problem frame, I should be
  concerned about
• Commanded behavior
• Are the commands sensible and viable?
• Information display
• From what we know about the real world, can the
  machine ensure the display is correct given the
  limited information it receives?

3
Local Traffic Monitoring
Vehicles sensors
c
a
C
Monitor Computer
Printout - Vehicles
Printout
d
b
C
a VS!SensorOni b MC!WtVehLine, WtTotLine
c VS!Bike, Car, Truck d SP!InformationPrinted

4
Concern

• If I have this problem frame, I should be
  concerned about
• Simple workpieces
• Are commands sensible and viable?
• You probably get to design commands
• Translation
• As the machine traverses the inputs and outputs
  domains simultaneously, it correctly calculates
  the values to be written to the outputs domain
  from the values encountered in the inputs domain.

5
Information Display Problem Frame
Real world
C3
RW!C1
C
Information machine
Display - Real world
Display
Y4
IM!E2
C
6
Transformation Problem Frame
Inputs
Y3
IN!Y1
X
Transform machine
IO Relation
Outputs
Y4
TM!y2
X
7
Payroll System

• Inputs Time cards, New employee form, benefit
  choice, raise, W4 form
• Outputs Pay checks, W1 form, W2 form, check and
  forms to insurance company

8
Payroll problem diagram fitted to information
display frame
Payroll forms
c
a
C
Requirements for payroll
Payroll System
Output
d
b
C
a Pf! time cards, new employee form, benefit
choice, raise, W4 C1 b PS! paychecks, W1, W2,
checks and forms to insurance E2 c Pf! time
cards, new employee form, benefit choice, raise,
W4 C3 d O! paychecks, W1, W2, checks and forms
to insurance Y4
9
A model

• Lets the machine remember phenomena
• Lets the machine carry out some calculations
  incrementally
• Can treat defined terms as if they corresponded
  to separate phenomena
• Can treat processes of a conceptual domain as if
  they were physical entities

10
Real world
C3
RW!C1
C
Modeling machine
Model - Real world
Model
Y6
MM!E5
X
Model
Y6
MD!Y7
X
Display machine
Display - Model
Display
Y4
DM!E2
C
11
Real world
C3
RW!C1
C
Payroll input
DB - Real world
Payroll DB
Y6
PI!E5
X
Payroll DB
Y6
MD!Y7
X
Reports machine
Reports -DB
Reports
Y4
RM!E2
C
12
Payroll DB
Y6
PD!Y7
X
Checkwriting machine
Paychecks - DB
Paycheck
Y4
CM!E2
C
Payroll DB
Y6
PD!Y7
X
W2 machine
W2 - DB
W2
Y4
WM!E2
C
13
Decomposing Problem

• Separate a complex information display problem
  into a separate model building problem and model
  display problem.
• Model is part of the solution.
• Model building problem and model display problem
  can be decomposed into parallel model building
  and model display problems.

14
Model Domains are common

• Compilers (Transformation)
• symbol table, abstract syntax tree
• Robotics (controlled behavior)
• map, where we are on the map
• Commanded behavior, Workpieces
• undo, selections

15
Costs of Model Domains

• An extra domain
• Can be similar to description of real world
  which is which?
• Many possible models picking one makes the
  problem description less general

16
Advantages of Model Domain

• Keep track of history
• Leads to more efficient machines
• Leads to simpler descriptions for requirements
  and machines
• defined terms implemented in one place
• conceptual entities defined in one place

17
Other variant frames

• Description
• Lexical domain that describes a part of the
  requirement, or part of some other domain.
• Operator
• Add an operator, like Commanded Behavior adds an
  operator to Controlled Behavior.
• Connection
• A domain between the machine and the central
  domain.

18
One-way lights with Description
b
a
Lights units
Conformity to regime
Lights controller
c
d
Encoded regime
cER! (Char) d ER!(Phase,Duration)
a LC! (RPulsei, GPulsei) b
LU!(Stopi,Goi)
19
Patient Monitoring with Connection
Nurses station
Periods ranges
c
a
Monitor machine
Medical staff
ICU patient
Analog devices
b
e
f
c Notify d RegisterValue f FactorEvidence
a Period, Range, PatientName, Factor c
EnterPeriod, EnterRange EnterPatentName,
EnterFactor
Patent monitoring annotated context diagram
20
Particular Concerns

• Concerns about a particular problem, not about a
  particular problem frame.
• Overrun
• Initialization
• Reliability
• Identities
• Completeness

21
Overrun

• A domains ability or inability to respond to
  each externally controlled event before the next
  event occurs.
• Occurs when there is a mismatch of speeds at a
  domain interface.
• Machine too fast
• Machine too slow

22
Strategies for slow machines

• Simple inhibition - the machine inhibits the
  shared events whenever it is not ready
• Ignoring - the machine ignores shared events
  whenever it is not ready
• Buffering - the machine buffers shared events
  whenever it is not ready, and participates later

23
Initialization

• Machine has initial state
• Problem has initial state
• only start machine when domain is in proper state
• machine sets state of the domain
• machine learns state of the domain
• Model has initial state

24
Reliability

• What happens when problem domain violates its
  description?
• Importance depends on
• Likelihood of failure
• Cost of failure

25
Reliability

• Separate problem of normal operation from problem
  of reliability
• Detection
• Diagnosis
• Repair

26
Identities

• Identities concern is when the machine has an
  interface of shared phenomena with a set of
  individuals that
• are not connected into any structure that
  identifies them and
• do not identify themselves.

27
Identities

• Add a structure that identifies individuals
• Model
• must be initialized
• must be maintained

28
Completeness

• Does the machine do everything it is supposed to?
• Does a finite-state machine define what happens
  for every possible event?
• Does the description cover a large enough scope?

29
Increasing Description Scope

• Widen participation
• X can do Y
• What else can it do?
• Can anything else do Y?
• Complementary events
• For each event, can the opposite event happen?

30
Particular Concerns

• Probably more
• Knowledge of experts
• Experience is important

31
Domain Flavors

• Static
• Physical domains, like maps for a flight
  simulator
• Data structures like trees or graphs

32
Dynamic Flavors

• Tolerance
• Capacity of a causal domain to tolerate events
  that do not cause changes in the domain state
• Example shut the door to a closed door.
• Robust, inhibiting, fragile
• Inhibition
• Safer to inhibit in a biddable domain than a
  causal domain

33
Dynamic flavors

• Discrete vs. continuous
• Active vs. passive state of domain
• Passive domain causes no shared events or
  shared state changes except in response to an
  externally caused event
• Event-active domain can generate events
• State-active domain can cause state changes

34
Formality

• Lexical domains are formal
• Causal domains are never formal
• Sometimes the informality of a problem domain
  can be treated quite lightly the flavor of
  informality is barely detectable, and you can get
  away with ignoring it. But sometimes the flavour
  is quite strong, and you have to be very careful
  in making descriptions.

35
Conclusion

• Problem frames and problem analysis is important.
• We need to be able to separate analysis of
  problem from analysis of solution.
• Much more work to do.