You are on page 1of 19

PLANNING AND LEARNING

CONTENT

• Planning
• Planning with State Space
Search
• Partial Order Planning
• Hierarchical Planning
• Conditional Planning
RECAP
• Uncertainty
• Representing Knowledge in an uncertain domain
• Reasoning in Uncertainty
• Probabilistic models
• Bayesian Network
• Semantics of belief network
• Simple inference in belief network
PLANNING
• Planning is about decision making tasks
performed by robots or computer
programs to achieve a specific goal.
• Execution of planning is about choosing
a sequence of actions, a high likelihood
to complete the specific task.
PLANNING
• Planning is the process of generating a sequence of actions to achieve a
desired goal or objective.
• It involves determining a series of steps or operations that, when executed,
lead from an initial state to a desired state or goal state.
• Planning is essential for autonomous agents and intelligent systems to make
decisions and accomplish tasks in various domains.
TYPES OF PLANNING TYPES OF PLANNING

1. Classical Planning
Classical Planning
2. Reactive Planning
Reactive Planning
3. Partial Order Planning
4. Temporal Planning Partial Order Planning

5. Hierarchical Planning Temporal Planning

6. Probabilistic Planning
Hierarchical Planning
7. Multi-Agent Planning
Probabilistic Planning

Multi-Agent Planning
PLANNING STATE SPACE SEARCH
• Imagine you're trying to solve a puzzle.
• You have different states the puzzle can be in, like when
some pieces are in the right place and others are not.
• State space searching is like exploring all the possible
moves you can make from your current state to
eventually reach the solution.
• You keep track of where you've been and where you can
go next until you find the solution.
• It's like navigating through a maze, trying different
paths until you find the way out.
EXAMPLE
• Classic problem of finding the shortest path from one location to another on a map.
1. Initial State: You start at a particular location on the map.

2. Goal State: Your destination is another specific location on the map.

3. State Space: The state space consists of all possible locations you can reach from your current location by moving along roads or
paths on the map.

4. Actions: Actions correspond to moving from one location to another along a road or path.

5. Transition Model: The transition model specifies which locations you can reach from your current location by taking specific actions
(moving along roads).

6. Cost Function: Each action has an associated cost, which could be the distance traveled or the time taken to reach the destination.

7. Search Algorithm: You use a search algorithm, such as breadth-first search or A* search, to explore the state space and find the
shortest path from the initial location to the goal location.

8. Solution: The solution is the sequence of actions (i.e., the path) that leads from the initial location to the goal location with the
minimum cost.

In this example, state space searching involves systematically exploring the map, considering different paths and their associated costs,
until you find the shortest path from your starting location to your destination.
PARTIAL ORDER PLANNING
• Partial order planning is a type of planning approach used
in artificial intelligence (AI) where the order of executing
actions is not strictly determined. Instead of requiring a
linear sequence of actions, partial order planning allows
for flexibility in the ordering of actions, enabling
concurrent or interleaved execution of certain actions.
• It's like planning a party where you don't need to follow a
strict schedule. You can do some things whenever you
want, as long as they don't clash with each other. It's
about being flexible and efficient in achieving your goals
without being tied down to a specific order of events.
IN PARTIAL ORDER PLANNING:
1.Flexibility: Actions can be executed in any order as long as they do PARTIAL ORDER PLANNING
not conflict with each other or violate any constraints.
2.Dependency Relations: Actions are related to each other by
dependency relations, indicating which actions must precede or Flexibility
follow others.
3.Constraints: Constraints may include temporal constraints (e.g., Dependency Relations
action A must precede action B) or resource constraints (e.g., only
one action can use a particular resource at a time). Constraints
4.Efficiency: Partial order planning can lead to more efficient plans
by allowing for parallel execution of independent actions or by Efficiency
avoiding unnecessary sequential constraints.
5.Search Algorithms: Partial order planning algorithms typically use Search Algorithms
techniques such as constraint propagation, constraint satisfaction, or
graph-based algorithms to explore the space of possible action
sequences and identify valid partial order plans.
PARTIAL ORDER PLANNING
Partial order planning provides a
flexible and efficient approach to
planning by relaxing strict ordering
constraints and allowing for more
concurrency and parallelism in action
execution. It is particularly useful in
domains where actions can be
performed concurrently or where the
exact order of actions is not critical to
achieving the desired goal.
HIERARCHICAL PLANNING
• Hierarchical planning is a problem-solving
approach in artificial intelligence where
complex tasks are broken down into smaller,
more manageable sub-tasks. This is akin to
dividing a big project into smaller tasks,
making it easier to tackle each part
individually.
• In short, Hierarchical planning is breaking
big tasks into smaller ones, like organizing a
project into manageable steps, making it
easier to achieve the overall goal efficiently.
IN HIERARCHICAL PLANNING:
HIERARCHICAL PLANNING
1.Decomposition: The main task is decomposed into a hierarchy of sub-tasks,
with higher-level goals broken down into lower-level goals.
2.Abstraction: Each level of the hierarchy abstracts away details, focusing on Decomposition
achieving a specific objective without considering the implementation details
of lower-level tasks.
Abstraction
3.Modularity: Hierarchical planning promotes modularity by encapsulating
functionality within different levels of the hierarchy, making it easier to
Modularity
understand and manage complex systems.
4.Reusability: Sub-tasks can be reused across different plans, allowing for
efficient planning and execution of similar tasks. Reusability

5.Coordination: Coordination between different levels of the hierarchy


ensures that the overall goal is achieved while respecting dependencies and Coordination
constraints between sub-tasks.
HIERARCHICAL PLANNING

Hierarchical planning provides a structured


approach to problem-solving by organizing
complex tasks into a hierarchy of
manageable sub-tasks, facilitating efficient
planning and execution in various domains.
CONDITIONAL PLANNING
• Conditional planning is a problem-solving approach in artificial intelligence
where plans are generated based on conditions or contingencies.
• Instead of following a fixed sequence of actions, conditional planning
considers different possible scenarios and selects appropriate actions based
on the current state of the environment or available information.
• It involves generating plans that are contingent upon specific conditions or
events, allowing for adaptive and flexible decision-making in dynamic
environments.
CONDITIONAL PLANNING INCLUDES THE FOLLOWING KEY COMPONENTS:

CONDITIONAL PLANNING
1. Conditions: Conditional planning involves considering different conditions or
states of the environment that may influence the execution of actions or the
selection of plans.
Conditions
2. Actions: Plans in conditional planning are composed of actions that can be
executed depending on the conditions present in the environment.
3. Contingencies: Plans may include contingencies or alternative courses of Actions
action to be taken based on the observed conditions or changes in the
environment. Contingencies
4. Decision Making: Conditional planning requires decision-making algorithms
or strategies to select appropriate actions or plans based on the current state of
Decision Making
the environment and the specified conditions.
5. Flexibility: Conditional planning allows for flexibility in plan execution,
enabling adaptation to changing conditions or unforeseen events during Flexibility
execution.
6. Optimization: Conditional planning algorithms may aim to optimize plans Optimization
based on specific criteria, such as minimizing costs, maximizing rewards, or
CONDITIONAL PLANNING

Conditional planning is like preparing for


different scenarios. Instead of having a fixed
plan, we consider different conditions or
situations that might arise and then decide
on the best course of action based on those
conditions. It's about being flexible and
adaptive in our planning process to achieve
our goals effectively, even when faced with
uncertainty or changes in the environment.
CONCLUSION
The planning module has taught us diverse
problem-solving techniques, from state
space searching to hierarchical planning.
We've learned to adapt strategies with
conditional and partial order planning,
enhancing our ability to tackle real-world
challenges efficiently in AI.
EDUCATE

EMPOWER

ILLUMINATE

You might also like