IAI Chapter 6 Planning

Questions about the lecture 'Introduction to Artificial Intelligence' of the RWTH Aachen Chapter 6 Planning

Questions about the lecture 'Introduction to Artificial Intelligence' of the RWTH Aachen Chapter 6 Planning


R. B.
This flashcard set delves into the intricacies of planning algorithms and problem-solving techniques at a university level. It covers key concepts such as the POP algorithm, STRIPS operators, and the differences between planning and search. The flashcards explore the definitions of actions, goal states, and initial states, as well as the characteristics of partial and total order plans. This set is invaluable for computer science students and researchers looking to understand the foundational principles and practical applications of planning in artificial intelligence.
Flashcards
20
Students
1
Language
English
Category
Computer Science
Level
University
Created / Updated
04.02.2017 / 14.02.2017

Flashcards

What is the idea of planning?

Planning is looking for plan to reach goal state from initial state

What are the differences of planning towards search? [3]

1. States need to be completely specified

2. Search is interested in all successor states while planning generates some

3. Search searches for sequence of actions while planning for a plan

What are the components of a plan in planning? [3]

1. Actions or plan steps which are paritally ordered // Order

2. Set of variable assignments V(var)=const // Bindings

3. Set of causal relations Si →c Sj with Si satisfies precondition c of Sj // Links

What are different names for actions of a plan? [2]

1. Plan steps and 2. STRIPS operators

What holds for partially ordered actions?

Actions are partially ordered Si < Sj with Si must be executed before Sj

What holds for a complete plan? [2]

1. If forall c in Precond(Sj) exists Si with Si<Sj and c in Effects(Si)

2. If forall Sk with Si<Sk<Sj \neg c not in Effects(Sk)

What holds for a consistent plan? [2]

1. If Si<Sj then Sj!<Si

2. If x=A then x!=B for distinct A and B // Unique Names Assumption

When is a plan a solution?

If it is complete and consistent

What is the idea of a total order plan?

Every step after each other

What is the idea of a partial order plan? [2]

1. Parallel step sequences are found

2. Decisions about order and variable assignment are delayed

How can causal relations be protected in a partial order plan? [3]

1. Conflict is S1 →c S2 and S3 \negc→ not parallel possible

 

2. Demotion is first S3 \negc→ then S1 →c S2

3. Promotion is first S1 →c S2 then S3 \negc→

What are the characteristics of Standford research institute problem solver (STRIPS) operators? [2]

1. Developed in the early 70s

2. Solver not in use anymore but operators

What is the definition of an action for STRIPS operators? [3]

1. Action name with parameters

2. Preconditions as positive literals

3. Effects as positive and negative literals

What is the definition of initial states for STRIPS operators? [3]

1. Set of ground literals

2. No function symbols

3. Only constants

What is the definition of goal states for STRIPS operators? [2]

1. Set of literals

2. Possible free variables implicitly existentially quantified

What is the definition of a problem in planning? [3]

An initial plan with

1. Operator S1 Start

2. Operator S2 Finish with set of preconditions

3. S1<S2

What is the definition of the POP Algorithm? [4]

1. Make minimal plan out of initial and goal state

2. Return plan if solution else perform Select-Subgoals with Sneed and c

3. Perform Choose-Operator

4. Perform Resolve-Threats of plan and jump to 2.

What is the definition of the Select-Subgoal Algorithm?

Return a step from steps of plan with not yet achieved c

What is the definition of the Choose-Operator Algorithm? [3]

1. Choose operator Sadd with effect c and if no such operator return fail

2. Add link Sadd →c Sneed to links of plan

3. Add Sadd<Sneed to orderings of plan

What is the definition of the Resolve-Threats Algorithm? [3]

1. Forall Sthreat that contradicts Si→cSj in links of plan

2. Add Sthreat<Si OR Sj<Sthreat to orderings of plan

3. If not consistent anymore return fail

Study