The Integration of AI and Level Design in Halo, Jaime Griesemer & Chris Butcher, Bungie/MGS, Game Developer Conference 2002
Annotated Powerpoint presentation describes the interaction between level design and AI in Halo. Also describes the AI architecture for Halo, the influence of hit points on the perceived intelligence and the hints for tactical positioning.
Terrain Analysis for Realtime Strategy Games, Dave Pottinger, Game Developer Conference 2000
Discusses terrain analysis in RTS games, and its relation to path finding, influence maps, areas. The concepts are illustrated using the Age of Empires I and II.
Formation-Based Pathfinding With Real-World Vehicles, Jim Van Verth, Red Storm Entertainment, in Game Developer Conference, 2000
Hierarchical formation based path finding combined with vehicle steering.
The Basics of Team AI, Clark Gibson, John O'Brien, Red Storm Entertainment, Game Developer Conference, 2001
Overview of the issues in implementing team AI, emphasizing the player's awareness of AI, the use of hierarchies, simplicity, and tuning.
AI getting a grasp of geography, M. Everson, design ideas for "The Clash of Civilizations", 1999?
Uses path sampling to determine geographical 'hot spots' in a strategy game
Tactical Analysis, Lars Lidén, slides 50 through 97 of the "The Use of Artificial Intelligence in the Computer Game Industry" talk, 2001
Great set of heuristics to interpret tactical properties of 3D terrain, given a waypoint graph and line-of-sight information.
The Making of the Official Counter-Strike Bot, Michael Booth, Turtle Rock Studios, Game Developer Conference 2004
Powerpoint presentation and movies of a solid FPS AI design with state-of-the-art pathfinding, tactical reasoning, view behavior and human-like flaws. In line with the human-only Counter-Strike multi-player games, the bots exhibit limited amount of team coordination.
Human Level Artificial Intelligence for Computer Simulations and Wargames, Ezra Sidran, 2003
Collection of ideas, notes and pointers about AI for wargames and military simulations. Also includes comparisons betweens today's game AI (actually: how the academic world sees game AI) and Ezra's previous work on wargames, such as the Universal Military Simulator (1989).
A Modular Framework for Artificial Intelligence Based on Stimulus Response Directives, Charles Guy, 1999
Older yet still relevant description of the AI architecture of SpecOps II inspired by the human biological nervous system. Apparently SpecOps II lacked terrain annotation for the AI, causing the AI to perform a massive amount of raycasting to sample their surroundings.
Learning Goals in Sport Games, Jack van Rijswijck, University of Alberta, Game Developer Conference, 2003
Despite the broad title, the document (and presentation) primarily deal with the use of force-fields to improve the defensive and offensive position picking in soccer games. The force-fields adapt to successful and failing ball passing combinations, thereby preventing repeated exploitation of AI positional weaknesses.
Terrain Reasoning for 3D Action Games, William van der Sterren, Game Developer Conference 2001
Discusses tactical evaluation of terrain positions (and how to improve the results using reinforcement learning). Includes examples of picking sniper spots, squad maneuver, and path planning. (Presentation also available).
AI for Tactical Grenade Handling, William van der Sterren, 2000
Discusses design and implementation of AI capable of judging when to employ (scarce) grenades, and fast precomputed table based approach to tossing grenades around corners.
Using COTS Software to Capture Deliberate and Reactive Weapons Firing Behavior: Lessons Learned in Knowledge Acquisition, A. Henniger, G. Taylor, J. Surdu, C. Jacquet,
Interesting paper (and presentation) because it presents ready-to-use information on target acquisition doctrine by means of several examples (illustrated by Red Storm's tactical shooter game Ghost Recon).
Tactical Movement Planning for Individual Combatants, D. Reece, M. Kraus, P. Dumanior, in Proceedings of the 9th Conference on Computer Generated Forces and Behavioral Representation, 2000
Describes the relation between various terrain representations and tactical path finding, and focuses on individual and squad AI in urban environments. One of the few papers dealing with tactical path finding.
Issues Involved With Integrating Live and Artificial Virtual Individual Combatants, D. Ourston, D. Reece, presented at the 1998 Spring Simulation Interoperability Workshop, Orlando, FL
Illustrates the problems of human - CGF interaction (using visual/aural cues), modeling direct and collateral damage, and realistic ways of movement.
Soldier Agents in a Virtual Urban Battlefield, D. Reece, presented at the First Workshop on Simulation and Interaction in Virtual Enviroments, University of Iowa, July 1995 (postscript reader required)
Illustrates the problems of dealing with terrain features and obstacles, the need for modeling realistic behavior, and the corresponding computing costs.
An Efficient Representation of Spatial Data For Terrain Reasoning By Computer Generated Forces, A.J. Reich, in Proceedings of the ELECSIM95, SCS, 1995
Uses topological information to reason about cross-country mobility of ground maneuver units, resulting in automatically generated 'cognitive maps' to support avenue of approach planning.
Terrain Reasoning Challenges in the CCTT Dynamic Environment, C.E. Campbell, G. McCulley, in Proceedings of the 5th annual conference on AI, Simulation and Planning in High Autonomy Systems, IEEE, 1994
Discusses the issues in designing a terrain representation supporting realistic close-combat tactical trainers.
A Hierarchical Distributed Planning Framework for Simulated Battlefield Entities, Jeremy Baxter, Richard Hepplewhite, at 19th Workshop of the UK Planning and Scheduling SIG, 2000
Hierarchical planning, illustrating different styles of planning and using some terrain analysis.
Executing Group Tasks Despite Losses and Failures, Jeremy W. Baxter, Graham S. Horn, at 10th Conference on Computer Generated Forces and Behavioral Representation, 2001
Multi-agent based framework to plan, execute and monitor group tasks. Accompanied by a good, clear presentation.
GI Agent: Multi-Agent Combat Simulation of Company Level Infantry Units, CPT Joel Pawloski, Naval Postgraduate School, February 28, 2001
Multi-agent simulator to evaluate various Company organization, using a RELATE agent design. Provides good overview of multi-agent combat simulators, and its use in evaluating Company organizations. Peculiar: With as "Lessons learned": "Unusual sources of information: game sites/books". A presentation is also available.
A Terrain Reasoning Algorithm for Defending a Fire Zone, Mikel D. Petty, Robert W. Franceschini, Amar Mukherjee, Information and Security, Vol. 3, 1999
Heuristic to position a hierarchy of units based on intervisibility, concealment, observation, spacing and organization.
Command Decision Model Technology Assessment, Army Model & Simulation Office (eds.), 1997?
US Army funded AI technology assessment by various authors, introducing and discussing rule based systems, fuzzy technologies, case-based reasoning, genetic algorithms and evolutionary programming, neural networks and bounded neural nets, lattice automata, planning systems, petri nets and colored petri nets. Also includes more information on command agents is availabe in the same publication.
Spatial Plans, Communication, and Teamwork in Synthetic MOUT Agents, Bradley J. Best, Christian Lebiere, CMU, 2003, in Proceedings of the 12th conference on Behavior Representation In Modeling and Simulation (paper and presentation).
AI spatial representation and reasoning for urban combat, prototyped using the Unreal Tournament game.
Modeling Synthetic Opponents in MOUT Training Simulations, Brad Best, Carnegie Mellon University, 2002
ACT-R (agent framework) based opponents in Unreal Tournament based game engine providing MOUT simulation.
Towards Human-like Adversaries for MOUT Training, J.E. Laird, R. Wray, in 23rd SOAR Workshop
Overview of issues in developing adversaries for urban combat.
Agent-based Soldier Behavior in Dynamic 3D Virtual Environments, David N. Back, MOVES Institute / NPS, 2002
Nicely documented design of AI adversaries in Unreal Tournament based combat simulator.
Realistic Evaluation of Terrain by Intelligent Natural Agents (RETINA), René Burgess, MOVES Institute / NPS, 2003
Solid overview of tactical terrain assessment, route and deployment planning, with examples and comparisons of the different techniques (force fields, A*, fluid flow).
Algorithmic Approaches to Finding Cover in Threedimensional, Virtual Environments, David J. Morgan, MOVES Institute / NPS, 2003
Overview, experiments and evaluation of cover in 3D environments. Prototyped in America's Army: Operations, and accompanied by code and test maps for AA:O (1.6).
Integrated On- And Off-line Cover Finding And Exploitation, Gregory H. Paull, Christian J. Darken, in Game On 2004, November 2004
Discussion of a waypoint-based and (positioned around the NPC) sensor-grid approaches to finding cover, acknowledging the expensive nature of ray casts. Prototyped using Unreal 2004 engine / America's Army.
Determining possible avenues of approach using ANTS, Pontus Svenson and Hedvig Sidenbladh, in the 6th International Conference on Information Fusion, pp 1110-1117, Cairns, Australia 2003
Uses ant simulation to compute (worst-case) potential enemy avenues of approach.
Artificial Intelligence in Games:Food for Thought Series, Aleks Jakulin, and the related application project Artificial Intelligence for Tactical Games, C(ibej, Jekovec, Leban, Lutrek, nidaric, 2003 (English presentation, demo movies, source code)
Nice observations (food for thought) and project/demonstrator of tactical behavior in 2D environment, with lessons learned (at the end of the presentation).
How qualitative spatial reasoning can improve strategy game AIs, K. Forbus, J. Mahoney, K. Dill, Spring Symposium on AI and Interactive Entertainment, March, 2001
Defines qualitative spatial reasoning, and sketches the benefits for tactical and strategic game AI.
Enabling and recognizing strategic play in strategy games: Lessons from Sun Tzu, Gordon, Andrew S., in the 2002 AAAI Spring Symposium on Artificial Intelligence and Interactive Entertainment, Stanford University, March 25-27, 2002.
Analysis of strategies in Sun Tzu's The Art of War, resulting in a list of the enabling AI actions necessary to exhibit these strategies in a RTS game.
It knows what you're going to do: Adding anticipation to a Quakebot, J.E. Laird, in AAAI 2000 Spring Symposium Series: Artificial Intelligence and Interactive Entertainment, March 2000, AAAI Technical Report SS-00-02
Adding threat modeling and prediction to a Quake bot, a based on the SOAR-agent architecture. Shows the flexibility of the generic rule and operator based agent architecture, but also the costs of implementing AI in such a generic way (as compared to, for example, the Quake III Arena AI, which also features limited prediction).
Related to Laird's work:
Towards Flexible Teamwork, M. Tambe, Journal of Artificial Intelligence Research, Volume 7, Pages 83-124
Agent based "joint intentions" team work, illustrated for rotary wing attack and escorting, and for RoboCup.
FC Portugal Team Description: RoboCup 2000 Simulation League Champion, Luis Paulo Reis, Nuno Lau, 2001
Analysis and description of the 2000 RoboCup Simulation League winner. The AI has a very interesting strategic/tactical positioning system, and great positional awareness. The AI's positioning and marking capability prevented all of the opponents to score.
Layered Learning in Multi-Agent Systems (A Winning Approach to Robotic Soccer), Peter Stone, Ph.D thesis / book, CMU, 2000
Analysis and design of AI for the real and simulated robots playing RoboCup soccer. Discusses the architecture of individual and team AI, and an approach to learning at individual and team level. All of this is illustrated using practical examples from the RoboCup competition.
On Emergence of Scalable Tactical and Strategic Behaviour, Mikahil Prokopenko, Marc Butler, Thomas Howard, 2000
Despite the theoretical jargon and thin examples, a good analysis and classification of the various kinds of AI agents, and what it takes to exhibit tactics. Based on RoboCup Simulation League experiences.
Game AI and game design
AI challenges in Entertainment and Player Expression, Doug Church, presentation at the 2005 AIIDE, June 2005
Focusing on single player games, Doug sketches how game AI is likely to develop and lists challenges for game AI. Again (see below), Doug puts the role of game AI in perspective.
Game Design, Game AI, and, sadly, reality, Doug Church, lecture at University of Michigan's Game Lecture Series, October 2001
Explains AI in games, and the role and contributions of game AI in the different genres (racing, action, RTS, RPG, sports, shooter).
AI and Design: How AI Enables Designers, Brian Reynolds, Big Huge Games, Game Developer Conference, 2004
From primarily an RTS (with random map generators) background, discusses the key uses for AI and where these overlap with design goals, illustrated using Civilization, Alpha Centauri and Rise of Nations.
The Use of Artificial Intelligence in the Computer Game Industry Lars Lidén, talk at University of Michigan, October 19, 2001
Using Half-Life as an example, Lars discusses the components of an AI system, decision making, tactical analysis and artificial stupidity.
Chap. 6 "Design Techniques and Ideals" in Chris Crawford's "The Art of Computer Game Design", 1982
Sketches the game design requirements for AI.
Game AI architecture and game AI mechanisms
The Quake III Arena Bot, Jan Paul van Waveren, Masters Thesis, Delft University of Technology, June 2001
One of the few documents describing an advanced commercial game AI implementation in full. The thesis provides insight in the overal architecture of the QuakeIIIArena bot and team AI, including its highlights such as automated area/waypoint creation, fast pathfinding using hierarchical routing, AI goal stacks and puzzle solving, and human-bot communication.
Dude, where's my Warthog: From Pathfinding to General Spatial Competence, by Damian Isla, Bungie Studios/MGS, AIIDE, 2005,
Illustrated by examples from Halo2, Damian explains needs for spatial reasoning in today's FPS AI, and suggests a pragmatic approach: the representation that solves your problem is the right one.
Agent Architecture Considerations for Real-Time Planning in Games (How to Plan in Real-Time and Keep Your Job), by Jeff Orkin, Monolith Productions, AIIDE, 2005,
Jeff's most detailed description of the Goal Action Planning (GOAP) implementation in NOLF2 and F.E.A.R. Accompanied by a paper. Very interesting to learn how Jeff prevent CPU spikes and uses beliefs (qualitative data).
Building AI Sensory Systems: Lessons from Thief and Half-Life, by Tom Leonard, Valve, Game Developer Conference 2003,
Powerpoint presentation on the sensory system design, as applied in two games famous for their AI response to player activity.
A Modular Framework for Artificial Intelligence Based on Stimulus Response Directives by Charles Guy, 1999
Elegant AI design after the (human) biological signal paths, in combination with heuristics.
Simulation Level-Of-Detail and Culling by Stephen Chenney, Game Developer Conference, 2001
Presentation addressing efficient approaches to convincing large complex environments populated by many AIs.
Towards More Realistic Pathfinding by Marco Pinter, Gamasutra feature, 2001
Advanced A* pathfinding, taking into account unit turn radius and facing.
Unreal Tournament AI documented in Wiki format by the Unreal mod community
Documentation of Unreal Tournament's AI, with pointers to information from Steven Polge.
AI Programming Wisdom (Steve Rabin, ed.), Charles River Media, 2002
Finally, a 600 page book solely devoted to game AI, offering over 71 chapters on how to build solid game AI. The contents of this book is described here.
Game Programming Gems I (Mark Deloura, ed.), Charles River Media, 2000
First in a series, providing over 50 practical approaches and solutions to game programming. The AI contents of this book is described here.
Game Programming Gems II (Mark Deloura, ed.), Charles River Media, 2001
Second in a series, providing dozens of more advanced solutions and techniques for game programming to game programming. The AI contents of this book is described here.
Commercial AI toolkits with freely available documentation
SimBionic offers an editor to rapidly define behaviors using FSMs and an engine to execute these behaviors.
Kynogon for games and simulation.
(Kynogon also supplied the RenderWare AI). Apparently provides some static terrain analysis ("access ways" represent potential hostile avenues of approach"), but the documentation made available is light on details.
PathEngine, Thomas Young's company offering middleware aimed at pathfinding
Extensively documented points-of-visibity mesh-based solution for pathfinding.
Game AI lecture notes
Artificial Intelligence for Interactive Entertainment lecture notes (Spring 2005) by Ken Forbus and Greg Dunham, Northwestern University, 2005
Great lecture notes illustrated with a non-trivial game: FreeCiv (a Civilization clone).
AI Game Programming lecture notes (Fall 2005) by Héctor Muñoz-Avila and his students, Lehigh University, 2005
Lecture notes with strong emphasis on planning (by Héctor Muñoz-Avila) and student presentations, the latter inspired by the AI Game Programming Wisdom 2 book.
Computer Game Design notes,
Spring 2001, by Ken Forbus, Northwestern University, 2001
Ken makes available of good collection of notes, many of them originating from the 2001 GDC tutorial by John Laird and Mike van Lent, and the 2001 GDC lecture on The Sims by Will Wright and Jamie Doornbos.
Forums and Blogs
GameDev.Net AI forum, GameDev.Net
Visited by both amateur game developers, professional game developers and academics.
Joint blog by Damian Isla (Halo2 AI), Greg Alt (AI for The Suffering, Drakan), Jeff Orkin (AI for F.E.A.R. and No One Lives Forever 2), Adam Russell (villager's AI for Fable), Paul Tozour (AI for MechWarrior 4), Robert Zudek.