Please enable JavaScript.
Coggle requires JavaScript to display documents.
Goal-based agents - pathfinding (Final State Machine (Requires an initial…
Goal-based agents - pathfinding
Final State Machine
Requires an initial state
Goals help organise behaviour
by limiting objectives and therefore actions that agent needs to consider
Problem solving
Goal Formulation
Based on current situation and performance measure.
Search
Looks for sequence of actions to reach goal.
If environment is observable, discrete, known solution to problem is a fixed sequence of actions
Well defined problem
Components
Initial state
Set of possible actions
Transition model
State Space
Historically from representing semantic networks
State : Node
Each node : Integer
Each edge connects two nodes
Weighted Edge
There may be a cost involved eg. time, distance, effort
Digraph
Nodes specifying a directed edge = ordered pair
Where edges have different costs (eg. can't go backwards, uphill v downhill)
With directed edges
Reflex and goal-based agents - making decisions
Agent
types
Simple reflex
Model-based reflex
Goal-based
Utility-based
Modelling a system
Representation of knowledge
System / agents have to make decisions
Finite State Machine
State depends on old state + input