You are on page 1of 17

BHARAT INSTITUTE OF ENGINEERING AND TECHNOLOGY

Ibrahimpatnam - 501 510, Hyderabad, Telangana State

DEPARTMENT OF ELECTRONICS AND COMMUNICATION


ENGINEERING
Presentation
On

Unit – IV – Planning - Classical Planning


by

CH. KRANTHI REKHA,


Assistant Professor,
ECE Dept, BIET

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 1


Problem Solving by Search –II
• Classical Planning: Definition of Classical
Planning, Algorithms for Planning with State-
Space Search, Planning Graphs, other
Classical Planning Approaches, Analysis of
Planning approaches.
• Planning and Acting in the Real World: Time,
Schedules, and Resources, Hierarchical
Planning, Planning and Acting in
Nondeterministic Domains, Multi agent
Planning.

Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 2


1/6/2021
PLANNING
• Planning is one of the classic AI problems
• It has been used as the basis for applications like controlling
robots and having conversations
• A plan is a sequence of actions
• An action is a transformation of a state, so a plan can be
thought of as a series of transformations of some initial state
• we usually specify an initial state and a goal state, and then try
to find (a short!) plan that transform the initial state state in the
goal state

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 3


• In more realistic planning problems, we might not
know exactly what initial state we are in!
• This states and actions model of planning is quite
general, and so can be applied to many different
kinds of problems
• In general, planning is an extremely difficult
problem, and planners are usually out-performed by
solvers customized specifically for a particular
problem

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 4


Example Planning Problem:
Making Tea
• suppose you have a robot that can serve tea
• think about what the robot must do to make tea, e.g.:
• it must put water in the kettle
• heat the kettle
• get a cup
• pour hot water into the cup (after the water is hot enough)
• get a tea bag
• leave the tea bag in the water for enough time
• remove the tea bag
• add milk
• add sugar
• mix the tea
• serve the tea
1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 5
Example Planning Problem:
Getting Dressed
 suppose you want to put your socks and shoes on
 suppose you have:
• a left sock
• a right sock
• a left shoe
• a right shoe
 one possible plan for putting these on is:
• put left sock on left foot
• put right sock on right foot
• put left shoe on left foot
• put right shoe on right foot

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 6


 Another possible plan is:
• put left sock on left foot
• put left shoe on left foot
• put right sock on right foot
• put right shoe on right foot
 For this planning problem, what matters is that the
left sock be put on before the left shoe, and the right
sock be put on before the right shoe

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 7


• For many actions, the exact order doesn’t matter
• Plus, it might be possible to do some actions in
parallel, e.g. maybe you could somehow hop into
both shoes at the same time
• Note that we made the somewhat unusual choice
that there are “left socks” and “right socks”
• More realistically, we might say we have different
socks, sock1 and sock2
• It doesn’t matter on which foot sock1 and sock2 go,
so even more plans are possible
– i.e. you could swap sock1 and sock2 in any correct
plan to get another correct plan

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 8


Planning vs. Problem Solving

Planning agent is very similar to problem


solving agent
• Constructs plans to achieve goals, then
executes them
Planning agent is different from problem
solving agent in:
• Representation of goals, states, actions
• Use of explicit, logical representations
• Way it searches for solutions

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 9


Planning vs. Problem Solving

Planning systems do the following:


 divide-and-conquer
 relax requirement for sequential
construction of solutions

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 10


Many AI Planners in History

• Well-known AI Planners:
• STRIPS (Fikes and Nilsson, 1971): theorem
proving system
• ABSTRIPS (Sacerdoti, 1974): added hierarchy of
abstractions
• HACKER (Sussman, 1975): use library of
procedures to plan
• NOAH (Sacerdoti, 1975): problem
decomposition and plan reordering

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 11


STRIPS-Based Approach to Robot Control
• Use first-order logic and theorem proving to
plan strategies from start to goal
STRIPS language:
• “Classical” approach that most planners use
• Lends itself to efficient planning algorithms
Environment: office environment with
specially colored and shaped objects
STRIPS planner: developed for this system to
determine the actions of the robot should
take to achieve goals

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 12


STRIPS
• STRIPS (STanford Research Institute Problem Solver)
 A restrictive way to express states, actions and goals, but leads
to more efficiency
 States: conjunctions of ground, function-free, and positive
literals, such as At(Home) ^ Have(Banana)
 Closed-world assumption is used
 Goals: conjunctions of literals, may contain variables
(existential), hence goal may represent more than one state
E.g. At(Home) ^ Have(Bananas)
E.g. At(x) ^ Sells(x, Bananas)
 Actions: preconditions that must hold before execution and the
effects after execution

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 13


An Example of STRIPS World Model
• A world model is a set of facts
• The robot’s knowledge can be represented by the following
predicates:
• INROOM(x, r), where x is a movable object, r is a room
• NEXTTO(x, t)
• STATUS(d, s), where d is a door, s means OPEN or CLOSED
• CONNECTS(d, rx, ry) The world model in this figure can be
represented with the above predicates, with the initial state and
goal state

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 14


Challenges of AI and Planning

• Closed world assumption: assumes that


world model contains everything the
robot needs to know: there can be no
surprise
• Frame problem: how to represent real
world situations in a manner that is
computationally tractable

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 15


Planning with State-Space Search

• Planning algorithms:
The most straightforward approach is to
use state-space search
• Forward state-space search
(Progression)
• Backward state-space search
(Regression)

1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 16


1/6/2021 Ch. Kranthi Rekha, Asst. Prof, BIET, ECE 17

You might also like