Posted December 11, 2025 by SteveShack
#techpost #release
After another exciting semester at Champlain College, I was delighted to take AI for Games, a class run by Alekandre Tolstenko. In the class we were given the liberty to choose whatever project we wanted with the prospect of implementing various AI behaviors. I chose to work with my good friend, Brandon Cherry on a Five Nights at Freddy’s fangame which utilized both Behavior Trees and Waypoint Tactics. However, for the sake of this post, I will be focusing solely on Waypoint Tactics. I believe that this demo provides a diligent job at displaying the use and importance of Waypoint Tactics.
Upon researching Waypoint Tactics, I found various sources, and I believe Timothy Carbone breaks down the system very well, saying, “Waypoints, in terms of pathfinding, are positions in a level that a pathfinding system uses to connect areas of the level” (Carbone 2019). While waypoints are only the half of it, they are incredibly important because they inform the AI where they can and should go next. However, it’s when the tactics come into play that it brings further strategy into the equation. It suddenly takes the system from randomization to a thinking process that determines what it perceives to be the best output and acts upon it. Carbone further emphasizes the usefulness of this system stating, “This type of system is insanely helpful for designers because it allows them to place key interest points wherever they’d like in the level without having to have the programmers implement new systems for each of them” (Carbone 2019). As a designer myself, this is incredibly important because it allows me to design levels easier, and think about certain scenarios because I can better anticipate the behavior from the AI within the environments.
In our game, we have various points set up for our sole animatronic to travel to, and while we use randomization for certain aspects of the game, there are factors that better determine what the AI does next. For example, if the player has their door open, and The Mangle is sitting down the hall, they are more likely to attack the player than enter the vents trying to attack them from behind. While these factor a bit into our behavior tree, it is still these tactics which are thought about during every playthrough which leads to the AI playing differently each time the game is played, making it feel fresh as the player plays.
To better understand the ease of access and use of this AI, I looked toward Guerrilla Games and their presentation on Killzone. The presentation was created by Remco Straatman and Arjen Beij where they note that the system allows developers to “Share/reuse behavior across games” enabling “Quick level prototyping”, “Can be combined with scripting” (Guerrilla). Since we have our AI set to utilize these waypoints, we can simply add in additional points, or even create different maps utilizing the same system, and the animatronic will still be able to navigate around our abandoned pizzeria. However, I will note that the developer did say that it can also be “Harder to control in detail” “Harder to test” this AI which is absolutely valid once there becomes more and more factors to consider (Guerrilla). This can lead to unexpected behaviors which we ourselves encountered in our game, but were able to fix before our final due date. A prime example of this would be when the power was out, The Mangle would pick a hallway to come down and kill the player, however, if they were in a hallway they could end up choosing the other which led to them walking past the player to then turn around once they reached the other hallway.
Additionally, William van der Sterren wrote that as an AI calculates values, an “ambush needs cover and shadow and exposure within a certain ray -> we can decide which locations are worth” (Van der Sterren). While considering these certain applications, we were able to ensure that our animatronic should have more sense than picking random spots, but should take advantage of its surroundings and for example attack the player when they appear more vulnerable.
While our Waypoint Tactic system is being used in a Horror game, Van der Sterren did note that our technique is “Primarily present in Real-time strategy (RTS) (as opposed to turn-based strategy) games” (Van der Sterren). This is due to RTS games commonly needing many units to automatically move to specific locations without unnecessary interaction from the player. However, the system is still useful for other genres including FPS games, for example enemies seeking cover, Stealth games doing patrolling of areas, and lastly Horror which we’ve utilized in our project for AI aggression, level interaction, and trying to scare the player.
This technique is very flexible, and can be utilized on various platforms including PC and Console due to modern CPUs being able to handle thousands of functions with quick computation times. In addition, this technique can also be utilized with mobile as well, but cannot be utilized too heavily for the sake of power consumption on mobile devices despite their ever increasing CPU power.
Lastly, our implementation of Waypoint Tactics has allowed us to create a horror experience along the lines of those found in FNAF games with randomization, and specific behaviors which the player can begin to learn even if they are subverted. For a small proof of concept project, Brandon and I have been able to make a really neat project which we are both proud of, and I appreciate the liberty of the project choice because it allowed us to create something we are fans of, and would enjoy working with in the future.
Works Cited:
Link to Game: https://brandoncherry166.itch.io/fnaf-the-mangle