Programme

The course provides an introduction to Automated Planning as a core area of Artificial Intelligence concerned with the automatic generation of action sequences to achieve desired goals. After presenting the motivations, application domains, and key challenges of planning systems, the course introduces planning problem representations, focusing on states, actions, goals, and domain specification languages. The main techniques for classical planning are then examined, including state-space search, plan-space search, planning graph methods, and heuristic planning approaches. The course subsequently covers more advanced topics such as Hierarchical Task Network (HTN) planning, temporal planning, and the management of temporal constraints. Particular attention is devoted to the use of PDDL for modeling planning domains and problems, as well as to the analysis of practical case studies and applications in robotics, autonomous systems, and decision-support environments.

Prerequisiti

To effectively attend the course, students are expected to possess a basic background in Artificial Intelligence and Computer Science, particularly regarding knowledge representation, search algorithms, and problem-solving techniques. Fundamental notions of propositional and first-order logic, together with elementary programming and data structure skills, are also recommended. No prior knowledge of Automated Planning is required, as the course progressively introduces the main planning models, languages, and algorithms.

Core Documentation

Main References

Malik Ghallab, Dana Nau, Paolo Traverso, Automated Planning: Theory and Practice, Morgan Kaufmann, 2004.
This is the primary reference textbook for the course. It provides a comprehensive introduction to planning models, algorithms, temporal planning, hierarchical planning, and practical applications.

Stuart Russell, Peter Norvig, Artificial Intelligence: A Modern Approach (3rd or 4th edition).
Provides the broader AI background needed to understand planning, search, and knowledge representation techniques.

Optional Reading

Dana Nau, Malik Ghallab, Paolo Traverso, Automated Planning and Acting, Cambridge University Press, 2016.
Covers more advanced topics related to the integration of planning and execution in intelligent systems.

Michael Wooldridge, An Introduction to MultiAgent Systems.
Useful for students interested in planning within distributed and multi-agent environments.

Teaching Material
Lecture slides and notes.
Selected research papers provided during the course.
Exercises and case studies based on PDDL modeling and automated planners.

Reference Bibliography

Vedi sopra

Attendance

Facoltativa a scelta dello studente

Type of evaluation

Student learning will be assessed through a written examination designed to evaluate both theoretical understanding and the ability to apply the main concepts and methodologies of Automated Planning. The examination may include open-ended questions, planning problem modeling exercises, PDDL domain and problem specification tasks, analysis of planning algorithms and techniques, as well as questions related to plan generation and evaluation. Assessment will consider the correctness of the answers, the quality of the reasoning process, the appropriate use of technical terminology, and the level of mastery of the planning methods and tools covered in the course.