The course  was offered in the Ph.D. program in Computing Science and Engineering of the Sapienza University in Rome, Department of Computer, Control, and Management Engineering.  Part of the material covered was also presented as invited lectures in the Artificial Intelligence in Games and Videogames M.Sc. course taught by Prof. Marco Schaerf.

The course aims to provide an introduction to the techniques currently used for the decision making of non-player characters (NPCs) in commercial video games, and show how a simple technique from academic artificial intelligence research can be employed to advance the state-of-the art.

The course is fairly introductory mostly aimed for first-year Ph.D. students and M.Sc. students, and is also appropriate for last-year undergraduate students.

All the material presented in lectures is available at tinyurl.com/NPC-AI-LaSapienza.

Abstract: In this article I will be covering an artificial intelligence (AI) technique for decision making that can be used in various parts of game development to account for “thinking before acting”. The technique is called Classical Planning in academic AI research, and is one of the most basic approaches for deliberating about the effects of actions and the way the properties of a given domain change under these effects. Variants of this technique have been used successfully in game development under the name Goal Oriented Action Planning (GOAP), each time focusing on a different aspect of this technique, adopting it for the particular needs of the game.

More information: In the article I go over a simple example inspired from real-time strategy games. I focus on the behavior of a peasant that can be instructed to bring food or handle other resources, and extend the character’s functionality so that they can take more advanced commands which may need a series of actions to be realized. Using this example I go over one of the simplest forms of planning, that is propositional STRIPS planning, and present an implementation using Python.

The following screenshots are code listings included in the article, which show how the planning algorithm can help the peasant find ways to realize the given commands. 

The self-contained Python code can be found here: PyPlan. The file goap.py contains all the necessary classes and methods for implementing planning, and example.py includes a simple example based on the peasant scenario that is explained in detail in the article.

On Friday March 16, I gave a talk at the IEEE Patras Games Expo 2012, which was organized by the IEEE Computer Society Student Chapter in collaboration with the IEEE Student Branch of University of Patras.

My talk was related to artificial intelligence (AI) techniques for specifying the behavior of non-player characters (NPCs) in video games. As a quick introduction to the main ideas behind the most common techniques that have been used in commercial video games, I went over three approaches: Finite State Machines (FMSs), Behavior Trees (BTs), and Goal-Oriented Action Planning (GOAP). The intention was to introduce students to the ideas and highlight what I consider to be the past, the present, and the future of techniques for NPC behavior.

The slides of my talk (in Greek) can be found  here. Some examples in the slides are taken from the book Artificial Intelligence for Games by  Ian Millington, and John Funge, an excellent resource for all types of AI methods for game development.

The 8th International Cognitive Robotics Workshop (CogRob-2012) will be held on July 22-23, 2012 in Toronto, Canada, as part of the of the Twenty-Sixth Conference on Artificial Intelligence (AAAI-12) workshop programme.

CogRob is a well-established forum that aims to bring together researchers involved in all aspects of the theory and implementation of cognitive robots, to discuss current work and future directions. While the emphasis of the workshop is on the methods and techniques developed in the field of Artificial Intelligence, we welcome work in related cognitive science disciplines investigating computational/cognitive models of behavior. Also, we especially welcome discussions and demonstrations of implemented systems.

For more information visit the website of CogRob-2012 or the ResearchGate page of CogRob-2012.

The talk was based on the following workshop paper I presented at AIIDE-2011: The SimpleFPS Planning Domain: A PDDL Benchmark for Proactive NPCs.

The details of the talk can be found here. The slides of my presentation can be found here.

Abstract:
In this paper we study the progression of situation calculus action theories that are able to handle a class of actions that, while extremely simple conceptually and common in many settings, cannot be handled by previous approaches. Specifically, based on the notion of safe-range queries from database theory and just-in-time action histories, we present a new type of action theories that ensures that actions have bounded effects over a restricted range of objects. Such theories may represent incomplete information and can be progressed by directly updating the knowledge base in an algorithmic manner.

Bibtex:

@inbook{vassos11hectorfest,
author = {Vassos, Stavros and Sardina, Sebastian},
booktitle = {Knowing, Reasoning, and Acting: Essays in Honour of Hector J. Levesque},
editor = {Lakemeyer, Gerhard and McIlraith, Sheila},
publisher = {College Publications},
title = {A Database-type Approach for Progressing Action Theories with Bounded Effects},
year = {2011}
}

Bibtex:

@inproceedings{vassos11simplefps,
author = {Vassos, Stavros and Papakonstantinou, Michail},
booktitle = {AIIDE 2011 Workshop on Non Player Character AI},
location = {Stanford, California, USA},
month = oct,
title = {The {SimpleFPS} Planning Domain: A {PDDL} Benchmark for Proactive {NPCs}},
year = {2011}
}

The slides of my talk (in Greek) can be found  here and the PDDL files mentioned in the talk can be found here. The planner used for the demo is BlackBox which can be found here (for windows use this executable and install the Cygwin DLL or make sure that cygwin1.dll is at the same folder as the BlackBox executable).

The abstract of the talk follows (in Greek).

Πρώτο μέρος: Εισαγωγή στην αναπαράσταση προβλημάτων σχεδιασμού (planning) με βάση τη γλώσσα STRIPS. Προέλαση, οπισθοχώρηση, και ευρετικές συναρτήσεις για την εύρεση λύσης σε προβλήματα σχεδιασμού με βάση την αναζήτηση. Αναπαράσταση προβλημάτων σχεδιασμού στην τυπική γλώσσα PDDL και χρήση του planner BlackBox για την επίλυση προβλημάτων σχεδιασμού στο πεδίο του puzzle game Sokoban.

Δεύτερο μέρος: Εισαγωγή στην ανάπτυξη τεχνητής νοημοσύνης για χαρακτήρες (non-player characters) σε video games και εφαρμογές τεχνικών σχεδιασμού σε εμπορικά video games. Αναπαράσταση των βασικών στοιχείων ενός First-Person Shooter game σε PDDL από την οπτική ενός αντίπαλου χαρακτήρα στον κόσμο του παιχνιδιού, και χρήση του planner BlackBox για την επίλυση προβλημάτων σχεδιασμού που σχετίζονται με τις επιλογές του χαρακτήρα στο παιχνίδι. Σύντομος σχολιασμός επιπλέον τεχνικών όπως η παρακολούθηση εκτέλεσης και ο επανασχεδιασμός.

The 9th International Workshop on Nonmonotonic Reasoning, Action and Change (NRAC-2011) will be held as part of the workshop programme of the 2011 International Joint Conference on Artificial Intelligence (IJCAI-2011) at Barcelona, Spain, in July 2011.

NRAC is a well-established forum for researchers interested in sharing their experiences in work in the areas of Nonmonotonic Reasoning, Reasoning about Action, and Belief Revision. An intelligent agent exploring a rich, dynamic world, needs cognitive capabilities in addition to basic functionalities for perception and reaction. The abilities to reason nonmonotonically, to reason about actions, and to change one’s beliefs, have been identified as fundamental high-level cognitive functions necessary for common sense. Research in all three areas has made significant progress during the last two decades of the past century. It is, however, crucial to bear in mind the common goal of designing intelligent agents. Researchers should be aware of advances in all three fields since often advances in one field can be translated into advances in another. Many deep relationships have already been established between the three areas and the primary aim of this workshop is to further promote this cross-fertilization.
This workshop will bring together researchers with the aim to:

• Compare and evaluate existing formalisms.
• Report on new developments.
• Identify the most important open problems and research questions.
• Identify possibilities of solution transferral between the areas.
• Identify important challenges for the advancement of the areas.
• Discuss challenges encountered when applying techniques in applications.

For more information visit the website of NRAC-2011 and the ResearchGate page of NRAC-2011.

Abstract:
In this thesis we study a reasoning module for agents that have cognitive abilities, such as memory, perception, action, and are expected to function autonomously for long periods of time. The module provides the ability to reason about action and change using the language of the situation calculus and variants of the basic action theories. The main focus of this thesis is on the logical problem of progressing an action theory.

First, we investigate the conjecture by Lin and Reiter that a practical first-order
definition of progression is not appropriate for the general case. We show that Lin and Reiter were indeed correct in their intuitions by providing a proof for the conjecture, thus resolving the open question about the first-order definability of progression and justifying the need for a second-order definition.

Then we proceed to identify three cases where it is possible to obtain a first-order progression with the desired properties: i) we extend earlier work by Lin and Reiter and present a case where we restrict our attention to a practical class of queries that may only quantify over situations in a limited way; ii) we revisit the local-effect assumption of Liu and Levesque that requires that the effects of an action are fixed by the arguments of the action and show that in this case a first-order progression is suitable; iii) we investigate a way that the local-effect assumption can be relaxed and show that when the initial knowledge base is a database of possible closures and the effects of the actions are range-restricted then a first-order progression is also suitable under a just-in-time assumption.

Finally, we examine a special case of the action theories with range-restricted effects and present an algorithm for computing a finite progression. We prove the correctness and the complexity of the algorithm, and show its application in a simple example that is inspired by video games.

Bibtex:

@phdthesis{vassos09phdthesis,
address = {Toronto, Canada},
author = {Vassos, Stavros},
school = {University of Toronto},
title = {A Reasoning Module for Long-Lived Cognitive Agents},
year = {2009}
}

Typos:

• Page 38,  point 5, last equation:  ”s less-than do(a,s’)” should be  ”s less-or-equal-than do(a,s’)”.
• Page 175, second paragraph, last sentence:  ”Unfortunately we can no longer then assume that D_una” should be “Unfortunately we can no longer then assume that D_alpha”.