| Crédit : 3 ECTS |
| Langue du cours : anglais
|
|
|
|
Volume horaire
- CM :
36 h
- Volume horaire global (hors stage) :
36 h
Compétences à acquérir
By the end of this course, students will be able to:
-
Model a variety of decision and planning problems using appropriate representations (state spaces, constraints, planning formalisms, etc.).
-
Understand and select suitable algorithmic approaches—such as search, optimization, or constraint-based methods—to compute solutions.
-
Use existing algorithms and solvers effectively, and assess their strengths, limitations, and suitability for a given problem.
Description du contenu de l'enseignement
This class introduces the main ideas and algorithms that let an artificial agent plan and carry out sequences of actions to reach a goal. We explore several families of techniques:
-
Search algorithms:
How an agent explores a state space. We look at uninformed methods like breadth-first and depth-first search, and informed methods such as greedy best-first search and A*, which use heuristics to guide the search.
-
Local search and optimization:
Techniques for improving a solution step by step, including hill climbing, simulated annealing, local beam search, and genetic algorithms.
-
Constraint satisfaction problems (CSPs):
A framework for modelling problems using variables and constraints. We cover AC-3 for constraint propagation and backtracking search for finding consistent assignments.
-
Nondeterministic and partially observable environments:
How agents plan when actions have uncertain effects or when they cannot see the full state of the world. We introduce AND–OR tree search and belief states.
-
Multi-agent environments:
Basic ideas from game playing, including minimax and alpha–beta pruning, where agents must reason about opponents.
-
Classical planning:
An introduction to PDDL and planning-graph techniques for encoding and solving high-level planning tasks.
Mode de contrôle des connaissances
50% Project - 50% Exam
Bibliographie, lectures recommandées
Russell, S., & Norvig, P. (2009). Artificial Intelligence: A Modern Approach (3rd ed.). Upper Saddle River, NJ: Prentice Hall.
Enseignant responsable
JULIEN LESCA