Automated Planning and Decision Making
Updates:
Instructor:
Prof. Ronen Brafman
Contact
information:
Office: 209 Bldg. 37.
Email: Brafman@cs
Phone: 642 8041
Content:
The goal of this course is to gain familiarity with models and algorithms
developed mostly in the field of Artificial Intelligence for automating the
process of decision making and planning. The two main motivations are to help
build autonomous systems, such as the rovers NASA landed on Mars or artificial
characters in a computer game, as well as provide technology for
decision-support systems. Given a model of a system, such as a robot on Mars,
we could write a program that tells it how to behave, but we would prefer to
simply tell it what we want to accomplish and have it automatically decide what
actions to take. This is the basic planning problem, and the course will
consider various models of planning problems and various solution methods.
These include: heuristic search algorithms and method for generating good
heuristics, reachability and relevance analysis, partial order planning, Bayesian
networks, influence diagrams, reinforcement learning, utility functions, and
more.
Workload
and grade: There will be a few theoretical and practical assignments, one
programming assignment, and two quizzes (bchanim). The quizzes will form 25% of
the grade, each. The contribution of each assignment will depend on its
difficulty.
The
following lectures are narrated, and so it is recommended to view them as slide
shows. The files are very large (10,13 and 16 MB). Please e-mail me if you find
some part difficult, so that I can go over it in class.
I will be available on Wednesday to answer questions.
Assignment 1 - Due 12/3 in the morning class
K-Subset relaxation by Domshlak
Monotonic relaxation by Domshlak
Partial Order Planning by Domshlak
Planning-as-Satisfiability by Domshlak
Conformant planning with audio - part 1
Conformant planning with audio - part 2
Conformant planning with audio - part 3