1 WORK PLAN 1.1 Tasks 1.1.1 Alpha Version 1.1.2 Beta Version 1.1.3 Final Version 1.2 Schedule (by team-member if applicable)Kingdom: Creating Dynamism through Quest Generation and Adaptive AI Participants: Sriraman Subbaraman: [email protected] Adam Porroni: [email protected] Spencer Miller: [email protected] Faculty Advisor: Professor Norman Badler Background and Abstract: Computer games hold a special place as a key area that spurs the development of better artificial intelligence and algorithms and structures that all go towards providing a more engaging and challenging experience for the player. In many games, the aspect of simulation with respect to active AI is overlooked, due to a lack of emphasis from game studios, with only a few notable exceptions. F.E.A.R., one such exception, implements adaptive AI in the player’s opponents, who react to strategies the player employs and then counter with strategies that are effective in defeating the player. In addition, this system allows difficulty to be solely based on the player’s first-person shooter (in this case) playing ability, as detection of novice playing can be also be adapted to by toning down the level of the AI. This makes for a much more accessible game that also feels much more realistic, as enemies dive behind cover and initiates circling maneuvers in order to kill the player. The other game that comes to mind is The Sims, a game in which the lives of the Sims are computer generated but player guided. Each Sim character has a set of needs and a life of its own, and the player must deal with those needs. This creates a dynamic world that the player must react to, allowing a player to get much more involved, than in other games. However, neither of these games taps the true potential of AI controlled NPCs or objects, or of adaptive AI. This is the basis for our research. In addition, games revolve around a set of central concepts, one of which is suspension of disbelief, the barrier between a player being truly involved in the game and distancing themselves from it because they cannot suspend their disbelief. Suspension of disbelief requires that a game follow common sense, the common sense of real life. In the field of role-playing games, especially Massively Multiplayer Online Role-Playing Games, the presence of static non player characters, or NPCs, will erode away any initial suspension of disbelief. This is the key factor in a problem called “grinding”, which is essentially the fact that most RPG’s have you do the same quests for different static NPCs in order to get experience. Only a small subset of game players is willing to stand the grind, and no solution has surfaced.The goal of this project is to add dynamic qualities to a gameworld in order to establish a greater level of depth and fundamentally redefine how gameplay works. We will design and implement “Dynamic” NPCs (non-player characters). A Dynamic NPC or DNPC, will operate as a thinking entity in the gameworld. Each DNPC will have an AI that attempts to accomplish the tasks generated by the role of the character. In attempting to fulfill a role, the DNPC may need to accomplish a task well outside its skill set, and at that point it would spawn a player quest and the content related to that quests. This system creates a dynamic world for the player to play in, one that changes with and more importantly without the players direct input. In addition, because quests become things necessary for the success of the DNPCs, they make much more sense to the player, and can provide a greater sense of accomplishment. However, the greatest consequence of such a system is the evolution of game world simulation, from a fundamentally static environment to a dynamic one. Indeed such a product would not only bring games to the next level, it would provide an interesting platform in which any number of social simulations could be run using those DNPCs. In fact, the AI controlling those DNPCs can undergo almost infinite revisions as the level of machine learning algorithms and adaptive AI techniques increases, as those two fields combine to form the character of a DNPC. In sum, our task is to make a “Kingdom” of DNPCs, a self-contained and yet self-sufficient community of dynamic characters within which a human player can immerse himself or herself. This community will not require the player as implied above, but rather will be augmented, complemented by a single or many human players interacting among the DNPCs. Related Work: Toward the goal of developing a fully immersive environment could be argued as the eternal trajectory of game evolution and design. With the fantastic advances in CPU processing potential, memory storage schemes and even GPU – graphical processing unit – technology, computer, console and arcade games have never before been as exciting and realistic as they are today. Alongside these developments comes desire, for, as Dr. Alexander Nareyek notes in the January 2007 edition of IEEE Intelligent Systems AI magazine, game-players fully appreciate these advances and “want ever-increasing free-roaming environments” at their disposal (Nareyek, A., 2007). These demands have not fallen on deaf ears, for industry Goliaths and Davids alike are searching for better and better storytellers to write their plotlines and adept artists to design their landscapes. However, except for some well-respected outliers in the game development industry and academia, research and production of realistic artificial intelligence systems to augment potentially immersive games barely reaches the mark. Among the exceptional minds researching interactive, proactive game AI is Dr. Pieter Spronck, a Dutch computer scientist who has produced dozens of papers and experiments on AI best practices in game environments. Currently working at the University of Maastricht in their esteemed Institute for Knowledge and Agent Technology, Dr. Spronck has set his focus mostly on dynamic adversarial AI models for real-time strategy and, just recently, role-playing games (see Spronck, P., Phase-Dependent Evaluation in RTS Games, (BNAIC 2007) 3-10). Among his most exhaustive research experiments has been in dynamic scripting, where he has explored diverse schemes for autonomous decision making by the game AI. His major paper in this research is Automatic Rule Ordering for Dynamic Scripting, which tests the