15-887*: Planning, Execution, and Learning
MW 1:30-2:50pm, GHC 5222
Fall 2016

Planning is ubiquitous in everyday life -- from planning how to make dinner to planning how to graduate from University with the least amount of work. Researchers in AI have studied planning problems for many years, and many techniques exist for automating planning processes.

This course will explore both classical and modern approaches to planning. Issues to be discussed include: how to represent actions and world state, how to search for plans efficiently, how to deal with uncertainty in actions and the world state, how to find "good" or "optimal" plans,how to represent time, and how to dynamically combine planning and execution.

Specific planning techniques to be covered include: means-ends analysis, linear and non-linear planning, partial-order planning, graph-based planning, heuristic planning, BBD-based planning, hierarchical planning, temporal planning, conditional and conformant planning, probabilistic planning and learning using Markov models (MDPs and POMDPs), integration of planning, perception and execution, execution monitoring and replanning, path planning, multi-agent planning, and scheduling.

There are no explicit prerequisites, but a basic knowledge of AI is assumed.

Course Information
This is a lecture course. There is no textbook, but students will study research papers and use existing planners.
There will be three homeworks (15% each), a project (25%), and a take-home final exam (30%).

July 23, 2016