Jonathan Jin
John Reppy
CMSC23800
05/29/14
One of the more established trends in the video-game industry is the strive towards realism in video games. This goal has benefitted immensely from advancements in the field of computer graphics; visual design is, after all, one of the more immediately-noticed apparent aspects of any video game, and oftentimes weighs heavily into consumers' first impression of the game. As a result of the current state of the field of computer graphics, modern-day triple-A game titles are almost consistently visually stunning, with highly detailed landscape and character models lending credibility to the in-game world and narrative.
However, there are inherent difficulties associated with the use of computer graphics to achieve realism in an interactive medium such as video games. Firstly, graphical realism is handicapped by diminishing returns. While the difference between a 2000-polygon character model and a 20,000-polygon model is stark and immediately apparent, that between a 20,000- and 20,000-polygon model is less so; and the trend propagates forward. Second is the infamous Uncanny Valley -- a problem with serious implications regarding the extent to which computer graphics can emulate and depict human characters.
These, among other, problems seriously impede future advances in conveying realism in video games by graphical means; as such, game designers have begun to find it necessary to explore alternative means of conveying realism.
The recently-released Bioshock Infinite is one instance of the aforementioned trend. The designers developed a unique pathfinding algorithm for the NPC Elizabeth -- one specifically tailored to as closely emulate human curiosity, autonomy, and exploratory behavior as mathematically possible.
Elizabeth's pathfinding AI consists of several key points.
Firstly is the notion of Elizabeth being both follower and leader. The distance between her and the player-character (PC) must be managed and balanced so that she never appears to be explicitly in-tow to the PC. Similarly, while Elizabeth is given "free reign" to explore the level map, her nature as a "companion" character means that her movement must nevertheless be tied to that of the PC.
Elizabeth's pathfinding AI essentially boils down to an augmented version of a leader-follower model of NPC pathfinding, whereby Elizabeth can move freely between the PC's current location and the next level-specific goal location.
Similarly, to create a sense of curiosity in her character, the designers chose to give Elizabeth the ability to freely interact with various objects in the game level (what I call, in this project, "Points of Interest", or"POIs"). Examples of such interactions include using an item, providing commentary, or setting a new objective for the player.
For the sake of this course project and presentation, I've chosen to implement the aforementioned algorithm in ActionScript 3 using the Flixel game-making library.
The implementation consists of the following main classes:
- Leader, defined in
demo/src/com/jjin/Leader.as; - Follower, defined in
demo/src/com/jjin/Follower.as; and - Map, defined in
demo/src/com/jjin/state/PlayState.as.
The demo is located in the demo/ directory.
Building the demo requires the Flex SDK. With it installed, one can, while inside the demo directory, build with:
mxmlc
-static-link-runtime-shared-libraries src/Elizabeth.as -output ./demo.swf
There is also a Makefile; make will run the above command.
Alternatively, this project and presentation is also hosted here; the demo is on the last slide.
Corriea, Alexa. "The long road to building AI for BioShock Infinite's Elizabeth." Polygon. N.p., 18 Mar. 2014. Web. 27 May 2014. http://www.polygon.com/2014/3/18/5522450/the-long-road-to-perfecting-ai-for-bioshock-infinites-elizabeth.