CS 536 - Advanced Artificial Intelligence

Spring 2004

Bob Wall



Class Web site .

The Lisp reference manual, Common Lisp the Language, 2nd Edition, by Guy L. Steele.


Partially Observable Markov Decision Processes (POMDPs)

Brief PowerPoint presentation I did on POMDPs.

Very good POMDP tutorial site.

Another page with pointers to POMDP information.

And one more general info site.



Moore, Brett L, Todd M. Quasny, Larry D. Pyeatt, and Eric D. Sinzinger, Performance of a Single Action POMDP in a Recognition Task.

Describes the implementation of a simple POMDP to recognize words in text with additive noise. The POMDP outperformed a recognizer based on Hidden Markov Models.


Schesvold, Doug, Jingpeng Tang, Benzir Md Ahmed, Karl Altenburg, and Kendall E. Nygard, POMDP Planning for High Level UAV Decisions: Search vs. Strike.

Describes the use of POMDPs for high-level robotic mission planning, specifically in UAVs (Unmanned Air Vehicles).


Foka, Amalia F. and Panos E. Trahanias, Predictive Autonomous Robot Navigation.

Describes the use of POMDPs to control robot movement in a crowded environment, to predict the trajectories of other objects and obstacles. Introduces a hierarchical representation of the POMDPs to maintain small state spaces and thus reduce computational complexity.


Forbes, Jeff, Tim Huang, Keiji Kanazawa, and Stuart Russell, The BATmobile: Towards a Bayesian Automated Taxi.

Describes the use of POMDPs to control the driving behavior of an automated taxi. Introduces several approximation methods used to implement the POMDP with sufficiently high performance. Simulations show that the POMDP control produces very good results.



Source code and results for Assignments

Results for each of the class assignments.

First assignment - Prolog.
program1.html Assignment description
date_diff.pl Source code for solution to first part - difference between two dates.
poker.pl Source code for solution to second part - poker hand classifier.
sql_db.pl Source code for solution to third part - SQL database in Prolog.
Second assignment - Machine Learning.
program2.html Assignment description
Third assignment - Dealer's Choice.
program3.html Assignment description
Imputation of Data for Input to SVMs We were to conduct experiments and write a paper for mock submissioin to the ANNIE 2004 conference. We chose to examine how data with missing attribute values could be preprocessed so it could be used as input for a Support Vector Machine (SVM) learning algorithm.



Mail me at: bwall@cs.montana.edu

Last modified: Apr. 22, 2004