Adaptive Media Project

Keep It Simple or Complicated? A Comparative Analysis of Nexbot and GOAP with Player Modeling

Project Overview

Inspired by NextBot horror game mods for Garry's Mod, we sought to create a horror game to find out what how the AI's design would affect the player's experience. A simple Finite State Machine (FSM) AI was compared to a more complex Goal Oriented Action Planning (GOAP) AI. The GOAP AI used player modeling to adapt to the player's playstyle. So if the player spent too long in an area, the bot would attempt to nudge the player to the next area.

Screenshot of the game with the monster

I was responsible for the FSM AI, which was designed to be simple and relatively predictable. The flow of the bot was inspired by a youtube video by Gargin. A crucial part of the design was that when the bot spotted the player, it would toss a coin to decide whether to chase the player or run away. This grain of unpredictability allowed for much emergent gameplay. When the bot runs away, it goes to a random location. Sometimes this would lead it to set a point which path was next to the player. Then it would run right at the player, only to go right past them.

This approach proved to be scary, tough in a different way than the GOAP AI. Players found the FSM AI to be more aggressive and direct. While the GOAP AI was more sneaky, always being close by, but never quite in sight.

Flowchart of the finite state machine

I also designed the post processing effects, which aimed to create a dark and eerie atmosphere by imitating a low quality digital camera. The effects would increase in intensity as the monster got closer, emphasizing the danger. These effects included a vignette, chromatic aberration, and a film grain.

Maze-like sewer

Tools and Skills Used

Unity
HDRP
Post Processing
C#
C++
AI
Finite State Machine
GOAP