Home Environment Themes Projects Core Curriculum Organization Events Publications

The purpose of the course is to give an overview of planning and optimal control, two central research areas for CADICS, as well as relate them to each other. The format is a full day workshop with lectures and discussions and written exercises for course credits. The course leaders and examiners are Jonas Kvarnström, IDA and Torkel Glad, ISY. The course will be given Tuesday May 8th 2012 9-16 in Systemet.

Automated Planning

  1. What is automated planning?
  2. The basics: A comparatively in-depth explanation of a couple of fundamental planning techniques
  3. An overview of the field: What can be done that goes beyond fundamental "classical" planning, and where are the limitations?
What is Automated Planning?
  • An informal overview of what the artificial intelligence community means by "automated planning".
  • A set of illustrating example domains where planning techniques have been or are being applied.  

Introduction to Automated Plan Generation

  • Basic concepts and terminology required for modeling planning domains and specifying planning problems.
  • An introduction to common formal models for planning, as well as higher-level concepts built on top of these models.
  • Planning as search: A general method for plan generation.
  • Domain-independent heuristics: How the structure of a planning domain can be used to automatically extract admissible or non-admissible heuristic functions. A general overview as well as one or more concrete examples of state-of-the-art heuristics.
  • Using rich domain knowledge for efficient planning.

Plan Generation: Extended Models

  • An overview of the breadth of the field of planning: What can we do that goes beyond the standard "classical" definition of planning, and where are the current limitations in terms of state-of-the-art planning?

Dynamic Programming

  1. The basic problem: to optimize a criterion over a number of stages. 
    • Some simple examples.
    • The different types: deterministic/stochastic, discrete/continuous state space, discrete/continuous time.
  2. The solution: the dynamic programming algorithm 
    • Discrete state space. Connection to shortest path problems.
    • Continuous state space. Connection to model predictive control (MPC).
  3. Some extensions and applications.