generic task routine design outline routine design definition. problem decomposition &...

21
Generic Task Routine Design

Upload: cameron-richards

Post on 26-Mar-2015

225 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Generic Task

Routine Design

Page 2: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Outline

• Routine Design Definition.

• Problem Decomposition & Specialists Hierarchy.

• Routine Design Agents.– Specialist, Plan, Task, Step, Plan Sponsor, Constraint, and Table Matcher.

• Routine Design Structure.

• Routine Design Algorithm.

• Multiple Design Definition.

• The Differences Between Single and Multiple Designs.– Design Tree, Design Limiter, Specialist Selector Interacting, and Table

Matcher.

• Case Storing & Loading.

• Mapping.

Page 3: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Routine Design Definition

• It is a problem solver that has been identified for the solution of

planning or more generally designed problems.

Routine Design

Design Constraints

Design Specification

Information processing task of Routine design

• Design Constraints: Set of input attributes.

• Design Specification: Set of output attributes.

Page 4: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Problem Decomposition & Specialists Hierarchy

• Any problem can be decomposed into a set of simpler sub-problems and each sub-problem can be also decomposed into another set of simpler sub-problems and so on.

• The Routine design approach is based on hierarchy of cooperating specialists, each responsible for an identified part of a complete design (solving a specific problem).

• The decomposition of a design problem solver into a hierarchy of specialists mirrors the decomposition of the design problem into sub-problems.

Page 5: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Problem

Specialist

Problem Decomposition

Specialist Hierarchy

Sub-Problem

Sub-Problem

Sub-Problem

Sub-Problem

Specialist

Specialist

Specialist

Specialist

More Conceptual Aspects of the design process

More Parametric Aspects of the design process

Page 6: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Routine Design Agents

• It is a place holder for the design plans and the selection mechanism for selecting among the different plans.

• One plan is chosen by a selector according to the input data and/or the current problem solving status.

• The selector consults sponsors of different plans and decides which plan to invoke.

• Each specialist has a plan selector, a set of plans to choose from, and maybe contains initial and final constraints.

Final constraints

Initial constraints

1. Specialist

SpecialistSelector Sponsor Plan2

Sponsor Plan1

Page 7: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

2. Plan: -• It specifies the course of action to be taken to

pursue the design from the current status.• It may contains other specialists, tasks,

constraints, and other Problem Solvers in an order dependent upon the problem being solved.

Plan

Problem Solvers

Tasks

Constraints

Specialists

Page 8: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

3. Task: -• It is a collection of related steps.

• It is the basic design action in Routine Design. It is responsible for setting the value of one design attribute.

• It maybe uses a table matcher or calls another problem solver.

4. Step: -

TaskStep n

Step 1 :

Step

Table Matcher

Problem Solver

OR

Page 9: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

• It determines if its plan is suitable (and how suitable).

• It tags its plan belonging to one of the following four categories: - Perfect for the current situation. Suitable for the current situation. Not sure. Not applicable for the current situation.

Plan Sponsor Table Matcher

6. Constraint: -• It is in the form of a comparison between the values of two design

attributes. If the result of this comparison is false then the current agent will fail.

5. Plan Sponsor: -

Page 10: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

• It consists of a set of rows, each row consists of condition side and action side. The condition side is a conjunction form in set of attributes and the action side assigns a value

for some output attribute.

Attribute1 Attribute2 Attribute3 Attribute4If X1 Y1 Z1 Advice1

Else If X2 Y2 Z2 Advice2Else If X3 Y3 Z3 Advice3Else If X1 Y4 Z1 Advice4Else If X2 Y1 Z2 Advice5Else X3 Y3 Z3 Advice6

Condition PartAction

Part

6. Table Matcher (Pattern Matcher): -

Page 11: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Routine Design Algorithm1. Collect the input attributes and display it.

2. Set the selected attributes in the current case with their values.

3. Invoke the top level specialist.

4. The top level specialist checks its initial constraints (if any exist).

5. The top level specialist requests its selector to select a plan.

6. The selector asks the sponsors of the different plans to assess the applicability of their respective plans.

7. The sponsors examine the current design status and the input data and assign applicability ratings to their plans.

8. The selector examines the applicability ratings and selects a plan.

9. Invoke the chosen plan that will take one or more of the following actions:-

(a) Invoke other specialists, then each one will follow a scenario similar to that followed by top level specialist.

(b) Execute tasks. These tasks consists of steps each of which a computation or uses a table matcher to assign a value to a design attribute.

(c) Check constraints.

10. The top level specialist checks its final constraints (if any exist).

11. Collect the output attributes and display it.

Page 12: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Multiple Design Definition

• It is an approach for generating multiple designs (solutions) satisfying a common set of design requirements and constraints.

Multiple Design

Requirements &Constraints Set of Designs

• Design (Solution) is a set of values for some

output attributes.

Page 13: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

The Differences Between Single and Multiple Designs

• Design Database (Tree)– For a given set inputs ( design requirements ) a multiple design system generates a group of trees, where each node represents a value for a design attribute.

–Each path from a root to a leaf node represents a complete design (solution).

– The Multiple Design deals with the design tree on behalf of the current case to satisfy each design alone.

Att1 = V1a

Att2 = V2a

Att2 = V2b

Att3 = V3d

Att3 = V3a

Att3 = V3b

Att3 = V3c

A graphic representation of a typical multiple design output

Page 14: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Design Tree Construction Algorithm

• There are two variables are used in this algorithm. They are Current_Nodes and Current_Agent to store the current nodes and current agent respectively.

1. First, the input attributes branch is built.2. Current_Nodes = last node in the above branch.While ( Agents Relation is executed ) do step 3.3. If ( Current_Agent = Specialist ) Then {Pass Current_Nodes to the selected plan in an order. } Else If ( Current_Agent = Plan OR Task )

Then { Pass Current_Nodes to its first item, then the returned nodes from this item are sent to the next item and so on. }

Else If ( Current_Agent = Step ) Then { Check each row of Current_Agent’s table matcher against each of the

paths ending at the nodes passed to it, adding the result of the row to the paths satisfying the conditions of that row. }. 4. Filter the design tree from input attributes.

Design Tree Construction Algorithm

Page 15: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Example: - This example describes how the design tree is constructed.

Specialist

Plan1

Plan2

Task11

Task12

Task21

Task22

Step111

Step112

Step121

Step211

Step221Spec. Call Plan1( [ N1 ] )

Plan1 Call Task11( [ N1 ] , [ N4 ] )

Task11 Call Step111( [ N1 ] , [ N2 , N3 ] )

Task11 Call Step112( [ N2 , N3 ] , [ N4 ] )

Plan1 Call Task12( [ N4 ] , [ N5 , N6 ] )

Task12 Call Step121( [ N4 ] , [ N5 , N6 ] )

Spec. Call Plan2( [ N1 ] )

Plan2 Call Task21( [ N1 ] , [ N7 , N8 ] )

Task21 Call Step211( [ N1 ] , [ N7 , N8 ] )

Plan2 Call Task22( [ N7 , N8 ] , [ N9 , N10 ] )

Task22 Call Step221( [ N7 , N8 ] , [ N9 , N10 ] )

Agents Hierarchy

Design Tree Construction steps

Page 16: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

The Constructed Design Tree

N1

N2

N3

N7

N8

N4

N4

N9

N10N9

N10

N5

N6N5N6

Spec. Call Plan1( [ N1 ] )

Plan1 Call Task11( [ N1 ] , [ N4 ] )

Task11 Call Step111( [ N1 ] , [ N2 , N3 ] )

Task11 Call Step112( [ N2 , N3 ] , [ N4 ] )

Plan1 Call Task12( [ N4 ] , [ N5 , N6 ] )

Task12 Call Step121( [ N4 ] , [ N5 , N6 ] )

Spec. Call Plan2( [ N1 ] )

Plan2 Call Task21( [ N1 ] , [ N7 , N8 ] )

Task21 Call Step211( [ N1 ] , [ N7 , N8 ] )

Plan2 Call Task22( [ N7 , N8 ] , [ N9 , N10 ] )

Task22 Call Step221( [ N7 , N8 ] , [ N9 , N10 ] )

Output Design Tree

Page 17: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Design Limiter

• Definition: -

- It is in the form of a table matcher and it is used to exclude the designs (solutions) that are known from experience to not be favorable if other designs exist.

• It is invoked at two levels, step and task levels.

N1

N3

N7

N4

N9

N10

N5

N6N7

Page 18: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Specialist Sponsor Interacting

• If (all returned plans within a ranking category fail)

Then { The next ranking category are used and so on }

• If (all plans fail)

Then { This specialist will fail }

Specialist SelectorSet of

Plans

Call

Set of highest

category plans

Page 19: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Table Matcher In Multiple Design

• Example

Attribute1 Attribute2 Attribute3 Attribute4If X1 Y1 Z1 Advice1If X2 Y2 Z2 Advice2: X3 Y3 Z3 Advice3If X1 Y4 Z1 Advice4

Condition Part Action Part

Soil Type

IrrigationType

PlantingType

Black Flood WetSandy Flood Wet

? ? dry

Single DesignMultiple Design

Soil Type

Soil Type

IrrigationType

PlantingType

Black ? Flood WetSandy ? Flood Wet

!=Black

!=Sandy

? dry

Page 20: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Case Storing & Loading

• Each Problem Solver accesses its current case. So, when we will be storing the Neper’s current case, we must store the other current cases of other Problem Solvers.

• We changed the case file format to avoid the compiler error and the directory structure that will be used in storing as shown in the next figure.

< selected directory>

<case name>Neper.case “Neper’s case”

<case name>_subs “New Folder”

<case name>Planting.case “Planting’s case”

. . . .

. . . .

<case name>Harvest.case “Harvest’s case”

New Case Directory Structure

• Note that:- The selected directory must be ended with.”\“

Page 21: Generic Task Routine Design Outline Routine Design Definition. Problem Decomposition & Specialists Hierarchy. Routine Design Agents. –Specialist, Plan,

Mapping• It is an important feature that allows the Problem

Solvers to interchange the values of their attributes.

Neper

Planting

Irrigation

Input

Input

output

output

Example- :