CSE 571: Artificial Intelligence
		 	(97725 LSC C106)

Instructor: Chitta Baral (http://www.public.asu.edu/~cbaral/)
Schedule:  T,Th 3:15-4:30 PM Class hours
           T,Th 1:30-2:30 PM Office hours 

This course will be taught as a follow-on to CSE 471. (For people
who are taking this course at the same time as 571, i will try
my best to co-ordinate with the 471 schedule.)
We will use a common theme and a common application to bind 
several subfields of AI which appear disconnected in most AI books
and pay special attention to topics related to uncertainity that
are not covered in CSE 471.
Due to several important recent developments in the field which 
have not made to the textbooks yet,
the course will be based on several (seem to be well written) recent
papers and handouts and will fall back on two books for the basic material.

The books are:

(1) Artificial Intelligence: a modern approach.
    Russell and Norvig.

(2) Computational Intelligence: a logical approach.
    Mackworth, Goebel and Poole.

Some of the authors of the papers to be covered in the
class are mentioned below.

* Introduction. 
  .Goal: Development of autonomous agents.
  .Such agents need to plan. But to plan they have to represent 
   knowledge about their abilities and the world, reason with this knowledge, 
   learn, and need to smartly weed through the possibilities (i.e., 
   search).

* STRIPS -- a simple language for planning.
  .Semantics of STRIPS. (State-space defined by STRIPS, entailment
   between STRIPS descriptions and plan queries)
  
* Planning as search
  .Quick overview: forward, backward and partial order planning.
  .Heuristic planner. (Geffner and Bonet.)
  .Quick introduction of Graph plan.
  .Casting Graphplan as a heuristic planner.

* Planning as constraint satisfaction. (Peter Van Beeks)
   (?? May be skipped)

* Logic, Knowledge representation: Using planning as a test case
  .Propositional Logic; SATplan, Blackbox. (Selman and Kautz)
  .Logic programming; Smodels, DLV based planners. (Lifschitz, Erdem)
  .Temporal Logic; TLPlan (Bacchus and Kabanza) 
  .Temporal goals. (Kabanza et al.)
  .Causal Logic; CCalc (McCain and Turner)
  .First-order logic, Situation Calculus; Golog (Reiter et al.)
  .Reasoning about sensing; Knowledge goals. (Baral and Son)
  .Planning with sensing ;Using logic program and DLV 
  .Complexity and sound approximations; w.r.t. planning with sensing 
		(Baral, Kreinovich and Trejo)

* Agents
  .Deliberative agents (Baral, Gelfond, and Provetti)
	-- planning from the current situation, knowledge assimilation,
	   abductive reasoning, counterfactuals.
  .Diagnostic agents (Baral, McIlraith and Son; Deep Space from NASA Ames)
	-- diagnosis, diagnostic planning, repair planning.
  .Reactive Agents (Brooks; Baral and Son)
  .Congolog (Levesque et al.)

* Uncertainity
  .Topics on uncertainity from Russel and Norvig that are not covered
   in CSE471. (Chapter 17)
  .Using Bayes' nets to reason about actions and plan (Boutilier)
  .Probabilistic calculus of actions. (Pearl)
  .Causality and Bayes' nets. (Pearl)
  .Planning with MDPs. (Geffner) Policy Iteration.
  .Planning with POMDps. (Geffner)
  	(Role of learning, and RTDP in planning with MDPs and POMDPs)

* Additional topics on learning
  .Explanation based learning; its role in planning. (Kambhampati et al.)
  .Learning Bayes' nets.
  .Learning causation.

* Robotics (robots = agents in a  physical world)
  .Hands on robotics using Lego robots.
  .Navigating using Sonars.
  .Mapping an office environment using sonars.
  .RAP and other robot control architectures

Class mailing list: cse571-f99@asu.edu (not active yet!)

Class home page: http://www.public.asu.edu/~cbaral/cse571-f99/ 

Grading:

I Class Participation: (~10%)

  -- Actively keeping up with readings, answering questions posed in
     the class 

  -- Being the designated note-taker for one or more classes

II Homeworks, Programming Assignments and Research Reports: (~45%)

  Programming assignments will involve using several programs
  mentioned above, and recreating and even improving parts of those
  programs. You may be asked to prove a new result, or to devlop a new
  formalism, or enhance an existing formalism; and write a report based 
  that. We also plan to use Nomad scout robots and Lego robots
  for the robotics part of the class. 
  

III Examinations: (~45%)

   Midterm and Final. The exams may be Take-Home exams.

(IMPORTANT: The instructor reserves the right to make minor
modifications in the grading weights with advance warning.)

*****HONOR CODE

Unless otherwise stated, all homeworks, projects and take-home tests
should be done by the students alone without consultation with 
any other students, guides, web-sites or experts (with the exception
of the instructor).

*****LATE SUBMISSION GUIDELINES:

By default, late submissions will not be entertained unless through
prior arrangements. In cases where the instructor provides permission
for late submission, a 10% late submission penalty/day will be levied.